Initial revision

Originally committed as revision 5 to svn://svn.ffmpeg.org/ffmpeg/trunk

52 files changed, 23294 insertions, 0 deletions

diff --git a/libavcodec/Makefile b/libavcodec/Makefile
new file mode 100644
index 0000000000..a418516b16
--- /dev/null
+++ b/libavcodec/Makefile
@@ -0,0 +1,59 @@
+include ../config.mk
+CFLAGS= -O2 -Wall -g
+LDFLAGS= -g
+
+OBJS= common.o utils.o mpegvideo.o h263.o jrevdct.o jfdctfst.o \
+      mpegaudio.o ac3enc.o mjpegenc.o resample.o dsputil.o \
+      motion_est.o imgconvert.o imgresample.o msmpeg4.o \
+      mpeg12.o h263dec.o rv10.o
+
+# currently using libac3 for ac3 decoding
+OBJS+= ac3dec.o \
+       libac3/bit_allocate.o libac3/bitstream.o libac3/downmix.o \
+       libac3/imdct.o  libac3/parse.o
+
+# currently using mpglib for mpeg audio decoding
+OBJS+= mpegaudiodec.o \
+       mpglib/layer1.o mpglib/layer2.o mpglib/layer3.o \
+       mpglib/dct64_i386.o mpglib/decode_i386.o  mpglib/tabinit.o
+
+# i386 mmx specific stuff
+ifdef CONFIG_MMX
+OBJS += i386/fdct_mmx.o i386/fdctdata.o i386/sad_mmx.o i386/cputest.o \
+	i386/dsputil_mmx.o
+endif
+
+LIB= libavcodec.a
+TESTS= imgresample-test dct-test
+
+all: $(LIB) apiexample
+
+$(LIB): $(OBJS)
+	rm -f $@
+	$(AR) rcs $@ $(OBJS)
+
+dsputil.o: dsputil.c dsputil.h
+
+%.o: %.c
+	$(CC) $(CFLAGS) -c -o $@ $< 
+
+%.o: %.s
+	nasm -f elf -o $@ $<
+
+clean: 
+	rm -f *.o *~ *.a i386/*.o i386/*~ \
+           libac3/*.o libac3/*~ \
+           mpglib/*.o mpglib/*~ \
+           apiexample $(TESTS)
+
+# api example program
+apiexample: apiexample.c $(LIB)
+	$(CC) $(CFLAGS) -o $@ $< $(LIB) -lm
+
+# testing progs
+
+imgresample-test: imgresample.c
+	$(CC) $(CFLAGS) -DTEST -o $@ $^ 
+
+dct-test: dct-test.o jfdctfst.o i386/fdct_mmx.o i386/fdctdata.o fdctref.o
+	$(CC) -o $@ $^
diff --git a/libavcodec/ac3dec.c b/libavcodec/ac3dec.c
new file mode 100644
index 0000000000..7332a11d8e
--- /dev/null
+++ b/libavcodec/ac3dec.c
@@ -0,0 +1,189 @@
+/*
+ * AC3 decoder
+ * Copyright (c) 2001 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include "avcodec.h"
+
+#include <inttypes.h>
+#include "libac3/ac3.h"
+
+/* currently, I use libac3 which is Copyright (C) Aaron Holtzman and
+   released under the GPL license. I may reimplement it someday... */
+typedef struct AC3DecodeState {
+    UINT8 inbuf[4096]; /* input buffer */
+    UINT8 *inbuf_ptr;
+    int frame_size;
+    int flags;
+    ac3_state_t state;
+} AC3DecodeState;
+
+static int ac3_decode_init(AVCodecContext *avctx)
+{
+    AC3DecodeState *s = avctx->priv_data;
+
+    ac3_init ();
+    s->inbuf_ptr = s->inbuf;
+    s->frame_size = 0;
+    return 0;
+}
+
+stream_samples_t samples;
+
+/**** the following two functions comes from ac3dec */
+static inline int blah (int32_t i)
+{
+    if (i > 0x43c07fff)
+	return 32767;
+    else if (i < 0x43bf8000)
+	return -32768;
+    else
+	return i - 0x43c00000;
+}
+
+static inline void float_to_int (float * _f, INT16 * s16) 
+{
+    int i;
+    int32_t * f = (int32_t *) _f;	// XXX assumes IEEE float format
+
+    for (i = 0; i < 256; i++) {
+	s16[2*i] = blah (f[i]);
+	s16[2*i+1] = blah (f[i+256]);
+    }
+}
+
+static inline void float_to_int_mono (float * _f, INT16 * s16) 
+{
+    int i;
+    int32_t * f = (int32_t *) _f;	// XXX assumes IEEE float format
+
+    for (i = 0; i < 256; i++) {
+	s16[i] = blah (f[i]);
+    }
+}
+
+/**** end */
+
+#define HEADER_SIZE 7
+
+static int ac3_decode_frame(AVCodecContext *avctx, 
+                            void *data, int *data_size,
+                            UINT8 *buf, int buf_size)
+{
+    AC3DecodeState *s = avctx->priv_data;
+    UINT8 *buf_ptr;
+    int flags, i, len;
+    int sample_rate, bit_rate;
+    short *out_samples = data;
+    float level;
+
+    *data_size = 0;
+    buf_ptr = buf;
+    while (buf_size > 0) {
+        len = s->inbuf_ptr - s->inbuf;
+        if (s->frame_size == 0) {
+            /* no header seen : find one. We need at least 7 bytes to parse it */
+            len = HEADER_SIZE - len;
+            if (len > buf_size)
+                len = buf_size;
+            memcpy(s->inbuf_ptr, buf_ptr, len);
+            buf_ptr += len;
+            s->inbuf_ptr += len;
+            buf_size -= len;
+            if ((s->inbuf_ptr - s->inbuf) == HEADER_SIZE) {
+                len = ac3_syncinfo (s->inbuf, &s->flags, &sample_rate, &bit_rate);
+                if (len == 0) {
+                    /* no sync found : move by one byte (inefficient, but simple!) */
+                    memcpy(s->inbuf, s->inbuf + 1, HEADER_SIZE - 1);
+                    s->inbuf_ptr--;
+                } else {
+                    s->frame_size = len;
+                    /* update codec info */
+                    avctx->sample_rate = sample_rate;
+                    if ((s->flags & AC3_CHANNEL_MASK) == AC3_MONO)
+                        avctx->channels = 1;
+                    else
+                        avctx->channels = 2;
+                    avctx->bit_rate = bit_rate;
+                }
+            }
+        } else if (len < s->frame_size) {
+            len = s->frame_size - len;
+            if (len > buf_size)
+                len = buf_size;
+            
+            memcpy(s->inbuf_ptr, buf_ptr, len);
+            buf_ptr += len;
+            s->inbuf_ptr += len;
+            buf_size -= len;
+        } else {
+            if (avctx->channels == 1)
+                flags = AC3_MONO;
+            else
+                flags = AC3_STEREO;
+
+            flags |= AC3_ADJUST_LEVEL;
+            level = 1;
+            if (ac3_frame (&s->state, s->inbuf, &flags, &level, 384)) {
+            fail:
+                s->inbuf_ptr = s->inbuf;
+                s->frame_size = 0;
+                continue;
+            }
+            for (i = 0; i < 6; i++) {
+                if (ac3_block (&s->state))
+                    goto fail;
+                if (avctx->channels == 1)
+                    float_to_int_mono (*samples, out_samples + i * 256);
+                else
+                    float_to_int (*samples, out_samples + i * 512);
+            }
+            s->inbuf_ptr = s->inbuf;
+            s->frame_size = 0;
+            *data_size = 6 * avctx->channels * 256 * sizeof(INT16);
+            break;
+        }
+    }
+    return buf_ptr - buf;
+}
+
+static int ac3_decode_end(AVCodecContext *s)
+{
+    return 0;
+}
+
+AVCodec ac3_decoder = {
+    "ac3",
+    CODEC_TYPE_AUDIO,
+    CODEC_ID_AC3,
+    sizeof(AC3DecodeState),
+    ac3_decode_init,
+    NULL,
+    ac3_decode_end,
+    ac3_decode_frame,
+};
+
+/* register codecs which could clash with mplayer symbols */
+/* XXX: rename all symbols to avoid clashed */
+void avcodec_register_more(void)
+{
+    register_avcodec(&mp3_decoder);
+    register_avcodec(&ac3_decoder);
+}
+
diff --git a/libavcodec/ac3enc.c b/libavcodec/ac3enc.c
new file mode 100644
index 0000000000..2bd4179aab
--- /dev/null
+++ b/libavcodec/ac3enc.c
@@ -0,0 +1,1460 @@
+/*
+ * The simplest AC3 encoder
+ * Copyright (c) 2000 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <math.h>
+#include "avcodec.h"
+
+#include "ac3enc.h"
+#include "ac3tab.h"
+
+//#define DEBUG
+//#define DEBUG_BITALLOC
+#define NDEBUG
+#include <assert.h>
+
+#define MDCT_NBITS 9
+#define N         (1 << MDCT_NBITS)
+#define NB_BLOCKS 6 /* number of PCM blocks inside an AC3 frame */
+
+/* new exponents are sent if their Norm 1 exceed this number */
+#define EXP_DIFF_THRESHOLD 1000
+
+/* exponent encoding strategy */
+#define EXP_REUSE 0
+#define EXP_NEW   1
+
+#define EXP_D15   1
+#define EXP_D25   2
+#define EXP_D45   3
+
+static void fft_init(int ln);
+static void ac3_crc_init(void);
+
+static inline INT16 fix15(float a)
+{
+    int v;
+    v = (int)(a * (float)(1 << 15));
+    if (v < -32767)
+        v = -32767;
+    else if (v > 32767) 
+        v = 32767;
+    return v;
+}
+
+static inline int calc_lowcomp1(int a, int b0, int b1)
+{
+    if ((b0 + 256) == b1) {
+        a = 384 ;
+    } else if (b0 > b1) { 
+        a = a - 64;
+        if (a < 0) a=0;
+    }
+    return a;
+}
+
+static inline int calc_lowcomp(int a, int b0, int b1, int bin)
+{
+    if (bin < 7) {
+        if ((b0 + 256) == b1) {
+            a = 384 ;
+        } else if (b0 > b1) { 
+            a = a - 64;
+            if (a < 0) a=0;
+        }
+    } else if (bin < 20) {
+        if ((b0 + 256) == b1) {
+            a = 320 ;
+        } else if (b0 > b1) {
+            a= a - 64;
+            if (a < 0) a=0;
+        }
+    } else {
+        a = a - 128;
+        if (a < 0) a=0;
+    }
+    return a;
+}
+
+/* AC3 bit allocation. The algorithm is the one described in the AC3
+   spec with some optimizations because of our simplified encoding
+   assumptions. */
+void parametric_bit_allocation(AC3EncodeContext *s, UINT8 *bap,
+                               INT8 *exp, int start, int end,
+                               int snroffset, int fgain)
+{
+    int bin,i,j,k,end1,v,v1,bndstrt,bndend,lowcomp,begin;
+    int fastleak,slowleak,address,tmp;
+    INT16 psd[256]; /* scaled exponents */
+    INT16 bndpsd[50]; /* interpolated exponents */
+    INT16 excite[50]; /* excitation */
+    INT16 mask[50];   /* masking value */
+
+    /* exponent mapping to PSD */
+    for(bin=start;bin<end;bin++) {
+        psd[bin]=(3072 - (exp[bin] << 7));
+    }
+
+    /* PSD integration */
+    j=start;
+    k=masktab[start];
+    do {
+        v=psd[j];
+        j++;
+        end1=bndtab[k+1];
+        if (end1 > end) end1=end;
+        for(i=j;i<end1;i++) {
+            int c,adr;
+            /* logadd */
+            v1=psd[j];
+            c=v-v1;
+            if (c >= 0) {
+                adr=c >> 1;
+                if (adr > 255) adr=255;
+                v=v + latab[adr];
+            } else {
+                adr=(-c) >> 1;
+                if (adr > 255) adr=255;
+                v=v1 + latab[adr];
+            }
+            j++;
+        }
+        bndpsd[k]=v;
+        k++;
+    } while (end > bndtab[k]);
+
+    /* excitation function */
+    bndstrt = masktab[start];
+    bndend = masktab[end-1] + 1;
+    
+    lowcomp = 0;
+    lowcomp = calc_lowcomp1(lowcomp, bndpsd[0], bndpsd[1]) ;
+    excite[0] = bndpsd[0] - fgain - lowcomp ;
+    lowcomp = calc_lowcomp1(lowcomp, bndpsd[1], bndpsd[2]) ;
+    excite[1] = bndpsd[1] - fgain - lowcomp ;
+    begin = 7 ;
+    for (bin = 2; bin < 7; bin++) {
+        lowcomp = calc_lowcomp1(lowcomp, bndpsd[bin], bndpsd[bin+1]) ;
+        fastleak = bndpsd[bin] - fgain ;
+        slowleak = bndpsd[bin] - s->sgain ;
+        excite[bin] = fastleak - lowcomp ;
+        if (bndpsd[bin] <= bndpsd[bin+1]) {
+            begin = bin + 1 ;
+            break ;
+        }
+    }
+    
+    end1=bndend;
+    if (end1 > 22) end1=22;
+    
+    for (bin = begin; bin < end1; bin++) {
+        lowcomp = calc_lowcomp(lowcomp, bndpsd[bin], bndpsd[bin+1], bin) ;
+        
+        fastleak -= s->fdecay ;
+        v = bndpsd[bin] - fgain;
+        if (fastleak < v) fastleak = v;
+        
+        slowleak -= s->sdecay ;
+        v = bndpsd[bin] - s->sgain;
+        if (slowleak < v) slowleak = v;
+        
+        v=fastleak - lowcomp;
+        if (slowleak > v) v=slowleak;
+        
+        excite[bin] = v;
+    }
+
+    for (bin = 22; bin < bndend; bin++) {
+        fastleak -= s->fdecay ;
+        v = bndpsd[bin] - fgain;
+        if (fastleak < v) fastleak = v;
+        slowleak -= s->sdecay ;
+        v = bndpsd[bin] - s->sgain;
+        if (slowleak < v) slowleak = v;
+
+        v=fastleak;
+        if (slowleak > v) v = slowleak;
+        excite[bin] = v;
+    }
+
+    /* compute masking curve */
+
+    for (bin = bndstrt; bin < bndend; bin++) {
+        v1 = excite[bin];
+        tmp = s->dbknee - bndpsd[bin];
+        if (tmp > 0) {
+            v1 += tmp >> 2;
+        }
+        v=hth[bin >> s->halfratecod][s->fscod];
+        if (v1 > v) v=v1;
+        mask[bin] = v;
+    }
+
+    /* compute bit allocation */
+    
+    i = start ;
+    j = masktab[start] ;
+    do {
+        v=mask[j];
+        v -= snroffset ;
+        v -= s->floor ;
+        if (v < 0) v = 0;
+        v &= 0x1fe0 ;
+        v += s->floor ;
+
+        end1=bndtab[j] + bndsz[j];
+        if (end1 > end) end1=end;
+
+        for (k = i; k < end1; k++) {
+            address = (psd[i] - v) >> 5 ;
+            if (address < 0) address=0;
+            else if (address > 63) address=63;
+            bap[i] = baptab[address];
+            i++;
+        }
+    } while (end > bndtab[j++]) ;
+}
+
+typedef struct IComplex {
+    short re,im;
+} IComplex;
+
+static void fft_init(int ln)
+{
+    int i, j, m, n;
+    float alpha;
+
+    n = 1 << ln;
+
+    for(i=0;i<(n/2);i++) {
+        alpha = 2 * M_PI * (float)i / (float)n;
+        costab[i] = fix15(cos(alpha));
+        sintab[i] = fix15(sin(alpha));
+    }
+
+    for(i=0;i<n;i++) {
+        m=0;
+        for(j=0;j<ln;j++) {
+            m |= ((i >> j) & 1) << (ln-j-1);
+        }
+        fft_rev[i]=m;
+    }
+}
+
+/* butter fly op */
+#define BF(pre, pim, qre, qim, pre1, pim1, qre1, qim1) \
+{\
+  int ax, ay, bx, by;\
+  bx=pre1;\
+  by=pim1;\
+  ax=qre1;\
+  ay=qim1;\
+  pre = (bx + ax) >> 1;\
+  pim = (by + ay) >> 1;\
+  qre = (bx - ax) >> 1;\
+  qim = (by - ay) >> 1;\
+}
+
+#define MUL16(a,b) ((a) * (b))
+
+#define CMUL(pre, pim, are, aim, bre, bim) \
+{\
+   pre = (MUL16(are, bre) - MUL16(aim, bim)) >> 15;\
+   pim = (MUL16(are, bim) + MUL16(bre, aim)) >> 15;\
+}
+
+
+/* do a 2^n point complex fft on 2^ln points. */
+static void fft(IComplex *z, int ln)
+{
+    int	j, l, np, np2;
+    int	nblocks, nloops;
+    register IComplex *p,*q;
+    int tmp_re, tmp_im;
+
+    np = 1 << ln;
+
+    /* reverse */
+    for(j=0;j<np;j++) {
+        int k;
+        IComplex tmp;
+        k = fft_rev[j];
+        if (k < j) {
+            tmp = z[k];
+            z[k] = z[j];
+            z[j] = tmp;
+        }
+    }
+
+    /* pass 0 */
+
+    p=&z[0];
+    j=(np >> 1);
+    do {
+        BF(p[0].re, p[0].im, p[1].re, p[1].im, 
+           p[0].re, p[0].im, p[1].re, p[1].im);
+        p+=2;
+    } while (--j != 0);
+
+    /* pass 1 */
+
+    p=&z[0];
+    j=np >> 2;
+    do {
+        BF(p[0].re, p[0].im, p[2].re, p[2].im, 
+           p[0].re, p[0].im, p[2].re, p[2].im);
+        BF(p[1].re, p[1].im, p[3].re, p[3].im, 
+           p[1].re, p[1].im, p[3].im, -p[3].re);
+        p+=4;
+    } while (--j != 0);
+
+    /* pass 2 .. ln-1 */
+
+    nblocks = np >> 3;
+    nloops = 1 << 2;
+    np2 = np >> 1;
+    do {
+        p = z;
+        q = z + nloops;
+        for (j = 0; j < nblocks; ++j) {
+
+            BF(p->re, p->im, q->re, q->im,
+               p->re, p->im, q->re, q->im);
+            
+            p++;
+            q++;
+            for(l = nblocks; l < np2; l += nblocks) {
+                CMUL(tmp_re, tmp_im, costab[l], -sintab[l], q->re, q->im);
+                BF(p->re, p->im, q->re, q->im,
+                   p->re, p->im, tmp_re, tmp_im);
+                p++;
+                q++;
+            }
+            p += nloops;
+            q += nloops;
+        }
+        nblocks = nblocks >> 1;
+        nloops = nloops << 1;
+    } while (nblocks != 0);
+}
+
+/* do a 512 point mdct */
+static void mdct512(INT32 *out, INT16 *in)
+{
+    int i, re, im, re1, im1;
+    INT16 rot[N]; 
+    IComplex x[N/4];
+
+    /* shift to simplify computations */
+    for(i=0;i<N/4;i++)
+        rot[i] = -in[i + 3*N/4];
+    for(i=N/4;i<N;i++)
+        rot[i] = in[i - N/4];
+        
+    /* pre rotation */
+    for(i=0;i<N/4;i++) {
+        re = ((int)rot[2*i] - (int)rot[N-1-2*i]) >> 1;
+        im = -((int)rot[N/2+2*i] - (int)rot[N/2-1-2*i]) >> 1;
+        CMUL(x[i].re, x[i].im, re, im, -xcos1[i], xsin1[i]);
+    }
+
+    fft(x, MDCT_NBITS - 2);
+  
+    /* post rotation */
+    for(i=0;i<N/4;i++) {
+        re = x[i].re;
+        im = x[i].im;
+        CMUL(re1, im1, re, im, xsin1[i], xcos1[i]);
+        out[2*i] = im1;
+        out[N/2-1-2*i] = re1;
+    }
+}
+
+/* XXX: use another norm ? */
+static int calc_exp_diff(UINT8 *exp1, UINT8 *exp2, int n)
+{
+    int sum, i;
+    sum = 0;
+    for(i=0;i<n;i++) {
+        sum += abs(exp1[i] - exp2[i]);
+    }
+    return sum;
+}
+
+static void compute_exp_strategy(UINT8 exp_strategy[NB_BLOCKS][AC3_MAX_CHANNELS],
+                                 UINT8 exp[NB_BLOCKS][AC3_MAX_CHANNELS][N/2],
+                                 int ch)
+{
+    int i, j;
+    int exp_diff;
+    
+    /* estimate if the exponent variation & decide if they should be
+       reused in the next frame */
+    exp_strategy[0][ch] = EXP_NEW;
+    for(i=1;i<NB_BLOCKS;i++) {
+        exp_diff = calc_exp_diff(exp[i][ch], exp[i-1][ch], N/2);
+#ifdef DEBUG            
+        printf("exp_diff=%d\n", exp_diff);
+#endif
+        if (exp_diff > EXP_DIFF_THRESHOLD)
+            exp_strategy[i][ch] = EXP_NEW;
+        else
+            exp_strategy[i][ch] = EXP_REUSE;
+    }
+    /* now select the encoding strategy type : if exponents are often
+       recoded, we use a coarse encoding */
+    i = 0;
+    while (i < NB_BLOCKS) {
+        j = i + 1;
+        while (j < NB_BLOCKS && exp_strategy[j][ch] == EXP_REUSE)
+            j++;
+        switch(j - i) {
+        case 1:
+            exp_strategy[i][ch] = EXP_D45;
+            break;
+        case 2:
+        case 3:
+            exp_strategy[i][ch] = EXP_D25;
+            break;
+        default:
+            exp_strategy[i][ch] = EXP_D15;
+            break;
+        }
+        i = j;
+    }
+}
+
+/* set exp[i] to min(exp[i], exp1[i]) */
+static void exponent_min(UINT8 exp[N/2], UINT8 exp1[N/2], int n)
+{
+    int i;
+
+    for(i=0;i<n;i++) {
+        if (exp1[i] < exp[i])
+            exp[i] = exp1[i];
+    }
+}
+                                 
+/* update the exponents so that they are the ones the decoder will
+   decode. Return the number of bits used to code the exponents */
+static int encode_exp(UINT8 encoded_exp[N/2], 
+                      UINT8 exp[N/2], 
+                      int nb_exps,
+                      int exp_strategy)
+{
+    int group_size, nb_groups, i, j, k, recurse, exp_min, delta;
+    UINT8 exp1[N/2];
+
+    switch(exp_strategy) {
+    case EXP_D15:
+        group_size = 1;
+        break;
+    case EXP_D25:
+        group_size = 2;
+        break;
+    default:
+    case EXP_D45:
+        group_size = 4;
+        break;
+    }
+    nb_groups = ((nb_exps + (group_size * 3) - 4) / (3 * group_size)) * 3;
+
+    /* for each group, compute the minimum exponent */
+    exp1[0] = exp[0]; /* DC exponent is handled separately */
+    k = 1;
+    for(i=1;i<=nb_groups;i++) {
+        exp_min = exp[k];
+        assert(exp_min >= 0 && exp_min <= 24);
+        for(j=1;j<group_size;j++) {
+            if (exp[k+j] < exp_min)
+                exp_min = exp[k+j];
+        }
+        exp1[i] = exp_min;
+        k += group_size;
+    }
+
+    /* constraint for DC exponent */
+    if (exp1[0] > 15)
+        exp1[0] = 15;
+
+    /* Iterate until the delta constraints between each groups are
+       satisfyed. I'm sure it is possible to find a better algorithm,
+       but I am lazy */
+    do {
+        recurse = 0;
+        for(i=1;i<=nb_groups;i++) {
+            delta = exp1[i] - exp1[i-1];
+            if (delta > 2) {
+                /* if delta too big, we encode a smaller exponent */
+                exp1[i] = exp1[i-1] + 2;
+            } else if (delta < -2) {
+                /* if delta is too small, we must decrease the previous
+               exponent, which means we must recurse */
+                recurse = 1;
+                exp1[i-1] = exp1[i] + 2;
+            }
+        }
+    } while (recurse);
+    
+    /* now we have the exponent values the decoder will see */
+    encoded_exp[0] = exp1[0];
+    k = 1;
+    for(i=1;i<=nb_groups;i++) {
+        for(j=0;j<group_size;j++) {
+            encoded_exp[k+j] = exp1[i];
+        }
+        k += group_size;
+    }
+    
+#if defined(DEBUG)
+    printf("exponents: strategy=%d\n", exp_strategy);
+    for(i=0;i<=nb_groups * group_size;i++) {
+        printf("%d ", encoded_exp[i]);
+    }
+    printf("\n");
+#endif
+
+    return 4 + (nb_groups / 3) * 7;
+}
+
+/* return the size in bits taken by the mantissa */
+int compute_mantissa_size(AC3EncodeContext *s, UINT8 *m, int nb_coefs)
+{
+    int bits, mant, i;
+
+    bits = 0;
+    for(i=0;i<nb_coefs;i++) {
+        mant = m[i];
+        switch(mant) {
+        case 0:
+            /* nothing */
+            break;
+        case 1:
+            /* 3 mantissa in 5 bits */
+            if (s->mant1_cnt == 0) 
+                bits += 5;
+            if (++s->mant1_cnt == 3)
+                s->mant1_cnt = 0;
+            break;
+        case 2:
+            /* 3 mantissa in 7 bits */
+            if (s->mant2_cnt == 0) 
+                bits += 7;
+            if (++s->mant2_cnt == 3)
+                s->mant2_cnt = 0;
+            break;
+        case 3:
+            bits += 3;
+            break;
+        case 4:
+            /* 2 mantissa in 7 bits */
+            if (s->mant4_cnt == 0)
+                bits += 7;
+            if (++s->mant4_cnt == 2) 
+                s->mant4_cnt = 0;
+            break;
+        case 14:
+            bits += 14;
+            break;
+        case 15:
+            bits += 16;
+            break;
+        default:
+            bits += mant - 1;
+            break;
+        }
+    }
+    return bits;
+}
+
+
+static int bit_alloc(AC3EncodeContext *s,
+                     UINT8 bap[NB_BLOCKS][AC3_MAX_CHANNELS][N/2],
+                     UINT8 encoded_exp[NB_BLOCKS][AC3_MAX_CHANNELS][N/2],
+                     UINT8 exp_strategy[NB_BLOCKS][AC3_MAX_CHANNELS],
+                     int frame_bits, int csnroffst, int fsnroffst)
+{
+    int i, ch;
+
+    /* compute size */
+    for(i=0;i<NB_BLOCKS;i++) {
+        s->mant1_cnt = 0;
+        s->mant2_cnt = 0;
+        s->mant4_cnt = 0;
+        for(ch=0;ch<s->nb_channels;ch++) {
+            parametric_bit_allocation(s, bap[i][ch], encoded_exp[i][ch], 
+                                      0, s->nb_coefs[ch], 
+                                      (((csnroffst-15) << 4) + 
+                                       fsnroffst) << 2, 
+                                      fgaintab[s->fgaincod[ch]]);
+            frame_bits += compute_mantissa_size(s, bap[i][ch], 
+                                                 s->nb_coefs[ch]);
+        }
+    }
+#if 0
+    printf("csnr=%d fsnr=%d frame_bits=%d diff=%d\n", 
+           csnroffst, fsnroffst, frame_bits, 
+           16 * s->frame_size - ((frame_bits + 7) & ~7));
+#endif
+    return 16 * s->frame_size - frame_bits;
+}
+
+#define SNR_INC1 4
+
+static int compute_bit_allocation(AC3EncodeContext *s,
+                                  UINT8 bap[NB_BLOCKS][AC3_MAX_CHANNELS][N/2],
+                                  UINT8 encoded_exp[NB_BLOCKS][AC3_MAX_CHANNELS][N/2],
+                                  UINT8 exp_strategy[NB_BLOCKS][AC3_MAX_CHANNELS],
+                                  int frame_bits)
+{
+    int i, ch;
+    int csnroffst, fsnroffst;
+    UINT8 bap1[NB_BLOCKS][AC3_MAX_CHANNELS][N/2];
+
+    /* init default parameters */
+    s->sdecaycod = 2;
+    s->fdecaycod = 1;
+    s->sgaincod = 1;
+    s->dbkneecod = 2;
+    s->floorcod = 4;
+    for(ch=0;ch<s->nb_channels;ch++) 
+        s->fgaincod[ch] = 4;
+    
+    /* compute real values */
+    s->sdecay = sdecaytab[s->sdecaycod] >> s->halfratecod;
+    s->fdecay = fdecaytab[s->fdecaycod] >> s->halfratecod;
+    s->sgain = sgaintab[s->sgaincod];
+    s->dbknee = dbkneetab[s->dbkneecod];
+    s->floor = floortab[s->floorcod];
+
+    /* header size */
+    frame_bits += 65;
+    if (s->acmod == 2)
+        frame_bits += 2;
+
+    /* audio blocks */
+    for(i=0;i<NB_BLOCKS;i++) {
+        frame_bits += s->nb_channels * 2 + 2;
+        if (s->acmod == 2)
+            frame_bits++;
+        frame_bits += 2 * s->nb_channels;
+        for(ch=0;ch<s->nb_channels;ch++) {
+            if (exp_strategy[i][ch] != EXP_REUSE)
+                frame_bits += 6 + 2;
+        }
+        frame_bits++; /* baie */
+        frame_bits++; /* snr */
+        frame_bits += 2; /* delta / skip */
+    }
+    frame_bits++; /* cplinu for block 0 */
+    /* bit alloc info */
+    frame_bits += 2*4 + 3 + 6 + s->nb_channels * (4 + 3);
+
+    /* CRC */
+    frame_bits += 16;
+
+    /* now the big work begins : do the bit allocation. Modify the snr
+       offset until we can pack everything in the requested frame size */
+
+    csnroffst = s->csnroffst;
+    while (csnroffst >= 0 && 
+           bit_alloc(s, bap, encoded_exp, exp_strategy, frame_bits, csnroffst, 0) < 0)
+        csnroffst -= SNR_INC1;
+    if (csnroffst < 0) {
+        fprintf(stderr, "Error !!!\n");
+        return -1;
+    }
+    while ((csnroffst + SNR_INC1) <= 63 && 
+           bit_alloc(s, bap1, encoded_exp, exp_strategy, frame_bits, 
+                     csnroffst + SNR_INC1, 0) >= 0) {
+        csnroffst += SNR_INC1;
+        memcpy(bap, bap1, sizeof(bap1));
+    }
+    while ((csnroffst + 1) <= 63 && 
+           bit_alloc(s, bap1, encoded_exp, exp_strategy, frame_bits, csnroffst + 1, 0) >= 0) {
+        csnroffst++;
+        memcpy(bap, bap1, sizeof(bap1));
+    }
+
+    fsnroffst = 0;
+    while ((fsnroffst + SNR_INC1) <= 15 && 
+           bit_alloc(s, bap1, encoded_exp, exp_strategy, frame_bits, 
+                     csnroffst, fsnroffst + SNR_INC1) >= 0) {
+        fsnroffst += SNR_INC1;
+        memcpy(bap, bap1, sizeof(bap1));
+    }
+    while ((fsnroffst + 1) <= 15 && 
+           bit_alloc(s, bap1, encoded_exp, exp_strategy, frame_bits, 
+                     csnroffst, fsnroffst + 1) >= 0) {
+        fsnroffst++;
+        memcpy(bap, bap1, sizeof(bap1));
+    }
+    
+    s->csnroffst = csnroffst;
+    for(ch=0;ch<s->nb_channels;ch++)
+        s->fsnroffst[ch] = fsnroffst;
+#if defined(DEBUG_BITALLOC)
+    {
+        int j;
+
+        for(i=0;i<6;i++) {
+            for(ch=0;ch<s->nb_channels;ch++) {
+                printf("Block #%d Ch%d:\n", i, ch);
+                printf("bap=");
+                for(j=0;j<s->nb_coefs[ch];j++) {
+                    printf("%d ",bap[i][ch][j]);
+                }
+                printf("\n");
+            }
+        }
+    }
+#endif
+    return 0;
+}
+
+static int AC3_encode_init(AVCodecContext *avctx)
+{
+    int freq = avctx->sample_rate;
+    int bitrate = avctx->bit_rate;
+    int channels = avctx->channels;
+    AC3EncodeContext *s = avctx->priv_data;
+    int i, j, k, l, ch, v;
+    float alpha;
+    static unsigned short freqs[3] = { 48000, 44100, 32000 };
+
+    avctx->frame_size = AC3_FRAME_SIZE;
+    avctx->key_frame = 1; /* always key frame */
+    
+    /* number of channels */
+    if (channels == 1)
+        s->acmod = 1;
+    else if (channels == 2)
+        s->acmod = 2;
+    else
+        return -1;
+    s->nb_channels = channels;
+
+    /* frequency */
+    for(i=0;i<3;i++) {
+        for(j=0;j<3;j++) 
+            if ((freqs[j] >> i) == freq)
+                goto found;
+    }
+    return -1;
+ found:    
+    s->sample_rate = freq;
+    s->halfratecod = i;
+    s->fscod = j;
+    s->bsid = 8 + s->halfratecod;
+    s->bsmod = 0; /* complete main audio service */
+
+    /* bitrate & frame size */
+    bitrate /= 1000;
+    for(i=0;i<19;i++) {
+        if ((bitratetab[i] >> s->halfratecod) == bitrate)
+            break;
+    }
+    if (i == 19)
+        return -1;
+    s->bit_rate = bitrate;
+    s->frmsizecod = i << 1;
+    s->frame_size_min = (bitrate * 1000 * AC3_FRAME_SIZE) / (freq * 16);
+    /* for now we do not handle fractional sizes */
+    s->frame_size = s->frame_size_min;
+    
+    /* bit allocation init */
+    for(ch=0;ch<s->nb_channels;ch++) {
+        /* bandwidth for each channel */
+        /* XXX: should compute the bandwidth according to the frame
+           size, so that we avoid anoying high freq artefacts */
+        s->chbwcod[ch] = 50; /* sample bandwidth as mpeg audio layer 2 table 0 */
+        s->nb_coefs[ch] = ((s->chbwcod[ch] + 12) * 3) + 37;
+    }
+    /* initial snr offset */
+    s->csnroffst = 40;
+
+    /* compute bndtab and masktab from bandsz */
+    k = 0;
+    l = 0;
+    for(i=0;i<50;i++) {
+        bndtab[i] = l;
+        v = bndsz[i];
+        for(j=0;j<v;j++) masktab[k++]=i;
+        l += v;
+    }
+    bndtab[50] = 0;
+
+    /* mdct init */
+    fft_init(MDCT_NBITS - 2);
+    for(i=0;i<N/4;i++) {
+        alpha = 2 * M_PI * (i + 1.0 / 8.0) / (float)N;
+        xcos1[i] = fix15(-cos(alpha));
+        xsin1[i] = fix15(-sin(alpha));
+    }
+
+    ac3_crc_init();
+
+    return 0;
+}
+
+/* output the AC3 frame header */
+static void output_frame_header(AC3EncodeContext *s, unsigned char *frame)
+{
+    init_put_bits(&s->pb, frame, AC3_MAX_CODED_FRAME_SIZE, NULL, NULL);
+
+    put_bits(&s->pb, 16, 0x0b77); /* frame header */
+    put_bits(&s->pb, 16, 0); /* crc1: will be filled later */
+    put_bits(&s->pb, 2, s->fscod);
+    put_bits(&s->pb, 6, s->frmsizecod + (s->frame_size - s->frame_size_min));
+    put_bits(&s->pb, 5, s->bsid);
+    put_bits(&s->pb, 3, s->bsmod);
+    put_bits(&s->pb, 3, s->acmod);
+    if (s->acmod == 2) {
+        put_bits(&s->pb, 2, 0); /* surround not indicated */
+    }
+    put_bits(&s->pb, 1, 0); /* no LFE */
+    put_bits(&s->pb, 5, 31); /* dialog norm: -31 db */
+    put_bits(&s->pb, 1, 0); /* no compression control word */
+    put_bits(&s->pb, 1, 0); /* no lang code */
+    put_bits(&s->pb, 1, 0); /* no audio production info */
+    put_bits(&s->pb, 1, 0); /* no copyright */
+    put_bits(&s->pb, 1, 1); /* original bitstream */
+    put_bits(&s->pb, 1, 0); /* no time code 1 */
+    put_bits(&s->pb, 1, 0); /* no time code 2 */
+    put_bits(&s->pb, 1, 0); /* no addtional bit stream info */
+}
+
+/* symetric quantization on 'levels' levels */
+static inline int sym_quant(int c, int e, int levels)
+{
+    int v;
+
+    if (c >= 0) {
+        v = (levels * (c << e)) >> 25;
+        v = (levels >> 1) + v;
+    } else {
+        v = (levels * ((-c) << e)) >> 25;
+        v = (levels >> 1) - v;
+    }
+    assert (v >= 0 && v < levels);
+    return v;
+}
+
+/* asymetric quantization on 2^qbits levels */
+static inline int asym_quant(int c, int e, int qbits)
+{
+    int lshift, m, v;
+
+    lshift = e + qbits - 24;
+    if (lshift >= 0)
+        v = c << lshift;
+    else
+        v = c >> (-lshift);
+    /* rounding */
+    v = (v + 1) >> 1;
+    m = (1 << (qbits-1));
+    if (v >= m)
+        v = m - 1;
+    assert(v >= -m);
+    return v & ((1 << qbits)-1);
+}
+
+/* Output one audio block. There are NB_BLOCKS audio blocks in one AC3
+   frame */
+static void output_audio_block(AC3EncodeContext *s,
+                               UINT8 exp_strategy[AC3_MAX_CHANNELS],
+                               UINT8 encoded_exp[AC3_MAX_CHANNELS][N/2],
+                               UINT8 bap[AC3_MAX_CHANNELS][N/2],
+                               INT32 mdct_coefs[AC3_MAX_CHANNELS][N/2],
+                               INT8 global_exp[AC3_MAX_CHANNELS],
+                               int block_num)
+{
+    int ch, nb_groups, group_size, i, baie;
+    UINT8 *p;
+    UINT16 qmant[AC3_MAX_CHANNELS][N/2];
+    int exp0, exp1;
+    int mant1_cnt, mant2_cnt, mant4_cnt;
+    UINT16 *qmant1_ptr, *qmant2_ptr, *qmant4_ptr;
+    int delta0, delta1, delta2;
+
+    for(ch=0;ch<s->nb_channels;ch++) 
+        put_bits(&s->pb, 1, 0); /* 512 point MDCT */
+    for(ch=0;ch<s->nb_channels;ch++) 
+        put_bits(&s->pb, 1, 1); /* no dither */
+    put_bits(&s->pb, 1, 0); /* no dynamic range */
+    if (block_num == 0) {
+        /* for block 0, even if no coupling, we must say it. This is a
+           waste of bit :-) */
+        put_bits(&s->pb, 1, 1); /* coupling strategy present */
+        put_bits(&s->pb, 1, 0); /* no coupling strategy */
+    } else {
+        put_bits(&s->pb, 1, 0); /* no new coupling strategy */
+    }
+
+    if (s->acmod == 2) {
+        put_bits(&s->pb, 1, 0); /* no matrixing (but should be used in the future) */
+    }
+
+#if defined(DEBUG) 
+    {
+        static int count = 0;
+        printf("Block #%d (%d)\n", block_num, count++);
+    }
+#endif
+    /* exponent strategy */
+    for(ch=0;ch<s->nb_channels;ch++) {
+        put_bits(&s->pb, 2, exp_strategy[ch]);
+    }
+    
+    for(ch=0;ch<s->nb_channels;ch++) {
+        if (exp_strategy[ch] != EXP_REUSE)
+            put_bits(&s->pb, 6, s->chbwcod[ch]);
+    }
+    
+    /* exponents */
+    for (ch = 0; ch < s->nb_channels; ch++) {
+        switch(exp_strategy[ch]) {
+        case EXP_REUSE:
+            continue;
+        case EXP_D15:
+            group_size = 1;
+            break;
+        case EXP_D25:
+            group_size = 2;
+            break;
+        default:
+        case EXP_D45:
+            group_size = 4;
+            break;
+        }
+        nb_groups = (s->nb_coefs[ch] + (group_size * 3) - 4) / (3 * group_size);
+        p = encoded_exp[ch];
+
+        /* first exponent */
+        exp1 = *p++;
+        put_bits(&s->pb, 4, exp1);
+
+        /* next ones are delta encoded */
+        for(i=0;i<nb_groups;i++) {
+            /* merge three delta in one code */
+            exp0 = exp1;
+            exp1 = p[0];
+            p += group_size;
+            delta0 = exp1 - exp0 + 2;
+
+            exp0 = exp1;
+            exp1 = p[0];
+            p += group_size;
+            delta1 = exp1 - exp0 + 2;
+
+            exp0 = exp1;
+            exp1 = p[0];
+            p += group_size;
+            delta2 = exp1 - exp0 + 2;
+
+            put_bits(&s->pb, 7, ((delta0 * 5 + delta1) * 5) + delta2);
+        }
+
+        put_bits(&s->pb, 2, 0); /* no gain range info */
+    }
+
+    /* bit allocation info */
+    baie = (block_num == 0);
+    put_bits(&s->pb, 1, baie);
+    if (baie) {
+        put_bits(&s->pb, 2, s->sdecaycod);
+        put_bits(&s->pb, 2, s->fdecaycod);
+        put_bits(&s->pb, 2, s->sgaincod);
+        put_bits(&s->pb, 2, s->dbkneecod);
+        put_bits(&s->pb, 3, s->floorcod);
+    }
+
+    /* snr offset */
+    put_bits(&s->pb, 1, baie); /* always present with bai */
+    if (baie) {
+        put_bits(&s->pb, 6, s->csnroffst);
+        for(ch=0;ch<s->nb_channels;ch++) {
+            put_bits(&s->pb, 4, s->fsnroffst[ch]);
+            put_bits(&s->pb, 3, s->fgaincod[ch]);
+        }
+    }
+    
+    put_bits(&s->pb, 1, 0); /* no delta bit allocation */
+    put_bits(&s->pb, 1, 0); /* no data to skip */
+
+    /* mantissa encoding : we use two passes to handle the grouping. A
+       one pass method may be faster, but it would necessitate to
+       modify the output stream. */
+
+    /* first pass: quantize */
+    mant1_cnt = mant2_cnt = mant4_cnt = 0;
+    qmant1_ptr = qmant2_ptr = qmant4_ptr = NULL;
+
+    for (ch = 0; ch < s->nb_channels; ch++) {
+        int b, c, e, v;
+
+        for(i=0;i<s->nb_coefs[ch];i++) {
+            c = mdct_coefs[ch][i];
+            e = encoded_exp[ch][i] - global_exp[ch];
+            b = bap[ch][i];
+            switch(b) {
+            case 0:
+                v = 0;
+                break;
+            case 1:
+                v = sym_quant(c, e, 3);
+                switch(mant1_cnt) {
+                case 0:
+                    qmant1_ptr = &qmant[ch][i];
+                    v = 9 * v;
+                    mant1_cnt = 1;
+                    break;
+                case 1:
+                    *qmant1_ptr += 3 * v;
+                    mant1_cnt = 2;
+                    v = 128;
+                    break;
+                default:
+                    *qmant1_ptr += v;
+                    mant1_cnt = 0;
+                    v = 128;
+                    break;
+                }
+                break;
+            case 2:
+                v = sym_quant(c, e, 5);
+                switch(mant2_cnt) {
+                case 0:
+                    qmant2_ptr = &qmant[ch][i];
+                    v = 25 * v;
+                    mant2_cnt = 1;
+                    break;
+                case 1:
+                    *qmant2_ptr += 5 * v;
+                    mant2_cnt = 2;
+                    v = 128;
+                    break;
+                default:
+                    *qmant2_ptr += v;
+                    mant2_cnt = 0;
+                    v = 128;
+                    break;
+                }
+                break;
+            case 3:
+                v = sym_quant(c, e, 7);
+                break;
+            case 4:
+                v = sym_quant(c, e, 11);
+                switch(mant4_cnt) {
+                case 0:
+                    qmant4_ptr = &qmant[ch][i];
+                    v = 11 * v;
+                    mant4_cnt = 1;
+                    break;
+                default:
+                    *qmant4_ptr += v;
+                    mant4_cnt = 0;
+                    v = 128;
+                    break;
+                }
+                break;
+            case 5:
+                v = sym_quant(c, e, 15);
+                break;
+            case 14:
+                v = asym_quant(c, e, 14);
+                break;
+            case 15:
+                v = asym_quant(c, e, 16);
+                break;
+            default:
+                v = asym_quant(c, e, b - 1);
+                break;
+            }
+            qmant[ch][i] = v;
+        }
+    }
+
+    /* second pass : output the values */
+    for (ch = 0; ch < s->nb_channels; ch++) {
+        int b, q;
+        
+        for(i=0;i<s->nb_coefs[ch];i++) {
+            q = qmant[ch][i];
+            b = bap[ch][i];
+            switch(b) {
+            case 0:
+                break;
+            case 1:
+                if (q != 128) 
+                    put_bits(&s->pb, 5, q);
+                break;
+            case 2:
+                if (q != 128) 
+                    put_bits(&s->pb, 7, q);
+                break;
+            case 3:
+                put_bits(&s->pb, 3, q);
+                break;
+            case 4:
+                if (q != 128)
+                    put_bits(&s->pb, 7, q);
+                break;
+            case 14:
+                put_bits(&s->pb, 14, q);
+                break;
+            case 15:
+                put_bits(&s->pb, 16, q);
+                break;
+            default:
+                put_bits(&s->pb, b - 1, q);
+                break;
+            }
+        }
+    }
+}
+
+/* compute the ac3 crc */
+
+#define CRC16_POLY ((1 << 0) | (1 << 2) | (1 << 15) | (1 << 16))
+
+static void ac3_crc_init(void)
+{
+    unsigned int c, n, k;
+
+    for(n=0;n<256;n++) {
+        c = n << 8;
+        for (k = 0; k < 8; k++) {
+            if (c & (1 << 15)) 
+                c = ((c << 1) & 0xffff) ^ (CRC16_POLY & 0xffff);
+            else
+                c = c << 1;
+        }
+        crc_table[n] = c;
+    }
+}
+
+static unsigned int ac3_crc(UINT8 *data, int n, unsigned int crc)
+{
+    int i;
+    for(i=0;i<n;i++) {
+        crc = (crc_table[data[i] ^ (crc >> 8)] ^ (crc << 8)) & 0xffff;
+    }
+    return crc;
+}
+
+static unsigned int mul_poly(unsigned int a, unsigned int b, unsigned int poly)
+{
+    unsigned int c;
+
+    c = 0;
+    while (a) {
+        if (a & 1)
+            c ^= b;
+        a = a >> 1;
+        b = b << 1;
+        if (b & (1 << 16))
+            b ^= poly;
+    }
+    return c;
+}
+
+static unsigned int pow_poly(unsigned int a, unsigned int n, unsigned int poly)
+{
+    unsigned int r;
+    r = 1;
+    while (n) {
+        if (n & 1)
+            r = mul_poly(r, a, poly);
+        a = mul_poly(a, a, poly);
+        n >>= 1;
+    }
+    return r;
+}
+
+
+/* compute log2(max(abs(tab[]))) */
+static int log2_tab(INT16 *tab, int n)
+{
+    int i, v;
+
+    v = 0;
+    for(i=0;i<n;i++) {
+        v |= abs(tab[i]);
+    }
+    return log2(v);
+}
+
+static void lshift_tab(INT16 *tab, int n, int lshift)
+{
+    int i;
+
+    if (lshift > 0) {
+        for(i=0;i<n;i++) {
+            tab[i] <<= lshift;
+        }
+    } else if (lshift < 0) {
+        lshift = -lshift;
+        for(i=0;i<n;i++) {
+            tab[i] >>= lshift;
+        }
+    }
+}
+
+/* fill the end of the frame and compute the two crcs */
+static int output_frame_end(AC3EncodeContext *s)
+{
+    int frame_size, frame_size_58, n, crc1, crc2, crc_inv;
+    UINT8 *frame;
+
+    frame_size = s->frame_size; /* frame size in words */
+    /* align to 8 bits */
+    flush_put_bits(&s->pb);
+    /* add zero bytes to reach the frame size */
+    frame = s->pb.buf;
+    n = 2 * s->frame_size - (s->pb.buf_ptr - frame) - 2;
+    assert(n >= 0);
+    memset(s->pb.buf_ptr, 0, n);
+    
+    /* Now we must compute both crcs : this is not so easy for crc1
+       because it is at the beginning of the data... */
+    frame_size_58 = (frame_size >> 1) + (frame_size >> 3);
+    crc1 = ac3_crc(frame + 4, (2 * frame_size_58) - 4, 0);
+    /* XXX: could precompute crc_inv */
+    crc_inv = pow_poly((CRC16_POLY >> 1), (16 * frame_size_58) - 16, CRC16_POLY);
+    crc1 = mul_poly(crc_inv, crc1, CRC16_POLY);
+    frame[2] = crc1 >> 8;
+    frame[3] = crc1;
+    
+    crc2 = ac3_crc(frame + 2 * frame_size_58, (frame_size - frame_size_58) * 2 - 2, 0);
+    frame[2*frame_size - 2] = crc2 >> 8;
+    frame[2*frame_size - 1] = crc2;
+
+    //    printf("n=%d frame_size=%d\n", n, frame_size);
+    return frame_size * 2;
+}
+
+int AC3_encode_frame(AVCodecContext *avctx,
+                     unsigned char *frame, int buf_size, void *data)
+{
+    AC3EncodeContext *s = avctx->priv_data;
+    short *samples = data;
+    int i, j, k, v, ch;
+    INT16 input_samples[N];
+    INT32 mdct_coef[NB_BLOCKS][AC3_MAX_CHANNELS][N/2];
+    UINT8 exp[NB_BLOCKS][AC3_MAX_CHANNELS][N/2];
+    UINT8 exp_strategy[NB_BLOCKS][AC3_MAX_CHANNELS];
+    UINT8 encoded_exp[NB_BLOCKS][AC3_MAX_CHANNELS][N/2];
+    UINT8 bap[NB_BLOCKS][AC3_MAX_CHANNELS][N/2];
+    INT8 exp_samples[NB_BLOCKS][AC3_MAX_CHANNELS];
+    int frame_bits;
+
+    frame_bits = 0;
+    for(ch=0;ch<s->nb_channels;ch++) {
+        /* fixed mdct to the six sub blocks & exponent computation */
+        for(i=0;i<NB_BLOCKS;i++) {
+            INT16 *sptr;
+            int sinc;
+
+            /* compute input samples */
+            memcpy(input_samples, s->last_samples[ch], N/2 * sizeof(INT16));
+            sinc = s->nb_channels;
+            sptr = samples + (sinc * (N/2) * i) + ch;
+            for(j=0;j<N/2;j++) {
+                v = *sptr;
+                input_samples[j + N/2] = v;
+                s->last_samples[ch][j] = v; 
+                sptr += sinc;
+            }
+
+            /* apply the MDCT window */
+            for(j=0;j<N/2;j++) {
+                input_samples[j] = MUL16(input_samples[j], 
+                                         ac3_window[j]) >> 15;
+                input_samples[N-j-1] = MUL16(input_samples[N-j-1], 
+                                             ac3_window[j]) >> 15;
+            }
+        
+            /* Normalize the samples to use the maximum available
+               precision */
+            v = 14 - log2_tab(input_samples, N);
+            if (v < 0)
+                v = 0;
+            exp_samples[i][ch] = v - 8;
+            lshift_tab(input_samples, N, v);
+
+            /* do the MDCT */
+            mdct512(mdct_coef[i][ch], input_samples);
+            
+            /* compute "exponents". We take into account the
+               normalization there */
+            for(j=0;j<N/2;j++) {
+                int e;
+                v = abs(mdct_coef[i][ch][j]);
+                if (v == 0)
+                    e = 24;
+                else {
+                    e = 23 - log2(v) + exp_samples[i][ch];
+                    if (e >= 24) {
+                        e = 24;
+                        mdct_coef[i][ch][j] = 0;
+                    }
+                }
+                exp[i][ch][j] = e;
+            }
+        }
+        
+        compute_exp_strategy(exp_strategy, exp, ch);
+
+        /* compute the exponents as the decoder will see them. The
+           EXP_REUSE case must be handled carefully : we select the
+           min of the exponents */
+        i = 0;
+        while (i < NB_BLOCKS) {
+            j = i + 1;
+            while (j < NB_BLOCKS && exp_strategy[j][ch] == EXP_REUSE) {
+                exponent_min(exp[i][ch], exp[j][ch], s->nb_coefs[ch]);
+                j++;
+            }
+            frame_bits += encode_exp(encoded_exp[i][ch],
+                                     exp[i][ch], s->nb_coefs[ch], 
+                                     exp_strategy[i][ch]);
+            /* copy encoded exponents for reuse case */
+            for(k=i+1;k<j;k++) {
+                memcpy(encoded_exp[k][ch], encoded_exp[i][ch], 
+                       s->nb_coefs[ch] * sizeof(UINT8));
+            }
+            i = j;
+        }
+    }
+
+    compute_bit_allocation(s, bap, encoded_exp, exp_strategy, frame_bits);
+    /* everything is known... let's output the frame */
+    output_frame_header(s, frame);
+        
+    for(i=0;i<NB_BLOCKS;i++) {
+        output_audio_block(s, exp_strategy[i], encoded_exp[i], 
+                           bap[i], mdct_coef[i], exp_samples[i], i);
+    }
+    return output_frame_end(s);
+}
+
+#if 0
+/*************************************************************************/
+/* TEST */
+
+#define FN (N/4)
+
+void fft_test(void)
+{
+    IComplex in[FN], in1[FN];
+    int k, n, i;
+    float sum_re, sum_im, a;
+
+    /* FFT test */
+
+    for(i=0;i<FN;i++) {
+        in[i].re = random() % 65535 - 32767;
+        in[i].im = random() % 65535 - 32767;
+        in1[i] = in[i];
+    }
+    fft(in, 7);
+
+    /* do it by hand */
+    for(k=0;k<FN;k++) {
+        sum_re = 0;
+        sum_im = 0;
+        for(n=0;n<FN;n++) {
+            a = -2 * M_PI * (n * k) / FN;
+            sum_re += in1[n].re * cos(a) - in1[n].im * sin(a);
+            sum_im += in1[n].re * sin(a) + in1[n].im * cos(a);
+        }
+        printf("%3d: %6d,%6d %6.0f,%6.0f\n", 
+               k, in[k].re, in[k].im, sum_re / FN, sum_im / FN); 
+    }
+}
+
+void mdct_test(void)
+{
+    INT16 input[N];
+    INT32 output[N/2];
+    float input1[N];
+    float output1[N/2];
+    float s, a, err, e, emax;
+    int i, k, n;
+
+    for(i=0;i<N;i++) {
+        input[i] = (random() % 65535 - 32767) * 9 / 10;
+        input1[i] = input[i];
+    }
+
+    mdct512(output, input);
+    
+    /* do it by hand */
+    for(k=0;k<N/2;k++) {
+        s = 0;
+        for(n=0;n<N;n++) {
+            a = (2*M_PI*(2*n+1+N/2)*(2*k+1) / (4 * N));
+            s += input1[n] * cos(a);
+        }
+        output1[k] = -2 * s / N;
+    }
+    
+    err = 0;
+    emax = 0;
+    for(i=0;i<N/2;i++) {
+        printf("%3d: %7d %7.0f\n", i, output[i], output1[i]);
+        e = output[i] - output1[i];
+        if (e > emax)
+            emax = e;
+        err += e * e;
+    }
+    printf("err2=%f emax=%f\n", err / (N/2), emax);
+}
+
+void test_ac3(void)
+{
+    AC3EncodeContext ctx;
+    unsigned char frame[AC3_MAX_CODED_FRAME_SIZE];
+    short samples[AC3_FRAME_SIZE];
+    int ret, i;
+    
+    AC3_encode_init(&ctx, 44100, 64000, 1);
+
+    fft_test();
+    mdct_test();
+
+    for(i=0;i<AC3_FRAME_SIZE;i++)
+        samples[i] = (int)(sin(2*M_PI*i*1000.0/44100) * 10000);
+    ret = AC3_encode_frame(&ctx, frame, samples);
+    printf("ret=%d\n", ret);
+}
+#endif
+
+AVCodec ac3_encoder = {
+    "ac3",
+    CODEC_TYPE_AUDIO,
+    CODEC_ID_AC3,
+    sizeof(AC3EncodeContext),
+    AC3_encode_init,
+    AC3_encode_frame,
+    NULL,
+};
diff --git a/libavcodec/ac3enc.h b/libavcodec/ac3enc.h
new file mode 100644
index 0000000000..40cc53aced
--- /dev/null
+++ b/libavcodec/ac3enc.h
@@ -0,0 +1,32 @@
+
+#define AC3_FRAME_SIZE (6*256)
+#define AC3_MAX_CODED_FRAME_SIZE 3840 /* in bytes */
+#define AC3_MAX_CHANNELS 2 /* we handle at most two channels, although
+                              AC3 allows 6 channels */
+
+typedef struct AC3EncodeContext {
+    PutBitContext pb;
+    int nb_channels;
+    int bit_rate;
+    int sample_rate;
+    int bsid;
+    int frame_size_min; /* minimum frame size in case rounding is necessary */
+    int frame_size; /* current frame size in words */
+    int halfratecod;
+    int frmsizecod;
+    int fscod; /* frequency */
+    int acmod;
+    int bsmod;
+    short last_samples[AC3_MAX_CHANNELS][256];
+    int chbwcod[AC3_MAX_CHANNELS];
+    int nb_coefs[AC3_MAX_CHANNELS];
+    
+    /* bitrate allocation control */
+    int sgaincod, sdecaycod, fdecaycod, dbkneecod, floorcod; 
+    int sgain, sdecay, fdecay, dbknee, floor;
+    int csnroffst;
+    int fgaincod[AC3_MAX_CHANNELS];
+    int fsnroffst[AC3_MAX_CHANNELS];
+    /* mantissa encoding */
+    int mant1_cnt, mant2_cnt, mant4_cnt;
+} AC3EncodeContext;
diff --git a/libavcodec/ac3tab.h b/libavcodec/ac3tab.h
new file mode 100644
index 0000000000..2d379f0404
--- /dev/null
+++ b/libavcodec/ac3tab.h
@@ -0,0 +1,180 @@
+/* tables taken directly from AC3 spec */
+
+/* possible bitrates */
+static const UINT16 bitratetab[19] = {
+    32, 40, 48, 56, 64, 80, 96, 112, 128, 
+    160, 192, 224, 256, 320, 384, 448, 512, 576, 640 
+};
+
+/* AC3 MDCT window */
+
+/* MDCT window */
+static const INT16 ac3_window[256]= {
+    4,    7,   12,   16,   21,   28,   34,   42,
+   51,   61,   72,   84,   97,  111,  127,  145,
+  164,  184,  207,  231,  257,  285,  315,  347,
+  382,  419,  458,  500,  544,  591,  641,  694,
+  750,  810,  872,  937, 1007, 1079, 1155, 1235,
+ 1318, 1406, 1497, 1593, 1692, 1796, 1903, 2016,
+ 2132, 2253, 2379, 2509, 2644, 2783, 2927, 3076,
+ 3230, 3389, 3552, 3721, 3894, 4072, 4255, 4444,
+ 4637, 4835, 5038, 5246, 5459, 5677, 5899, 6127,
+ 6359, 6596, 6837, 7083, 7334, 7589, 7848, 8112,
+ 8380, 8652, 8927, 9207, 9491, 9778,10069,10363,
+10660,10960,11264,11570,11879,12190,12504,12820,
+13138,13458,13780,14103,14427,14753,15079,15407,
+15735,16063,16392,16720,17049,17377,17705,18032,
+18358,18683,19007,19330,19651,19970,20287,20602,
+20914,21225,21532,21837,22139,22438,22733,23025,
+23314,23599,23880,24157,24430,24699,24964,25225,
+25481,25732,25979,26221,26459,26691,26919,27142,
+27359,27572,27780,27983,28180,28373,28560,28742,
+28919,29091,29258,29420,29577,29729,29876,30018,
+30155,30288,30415,30538,30657,30771,30880,30985,
+31086,31182,31274,31363,31447,31528,31605,31678,
+31747,31814,31877,31936,31993,32046,32097,32145,
+32190,32232,32272,32310,32345,32378,32409,32438,
+32465,32490,32513,32535,32556,32574,32592,32608,
+32623,32636,32649,32661,32671,32681,32690,32698,
+32705,32712,32718,32724,32729,32733,32737,32741,
+32744,32747,32750,32752,32754,32756,32757,32759,
+32760,32761,32762,32763,32764,32764,32765,32765,
+32766,32766,32766,32766,32767,32767,32767,32767,
+32767,32767,32767,32767,32767,32767,32767,32767,
+32767,32767,32767,32767,32767,32767,32767,32767,  
+};
+
+static UINT8 masktab[253];
+
+static const UINT8 latab[260]= {
+0x0040,0x003f,0x003e,0x003d,0x003c,0x003b,0x003a,0x0039,0x0038,0x0037,
+0x0036,0x0035,0x0034,0x0034,0x0033,0x0032,0x0031,0x0030,0x002f,0x002f,
+0x002e,0x002d,0x002c,0x002c,0x002b,0x002a,0x0029,0x0029,0x0028,0x0027,
+0x0026,0x0026,0x0025,0x0024,0x0024,0x0023,0x0023,0x0022,0x0021,0x0021,
+0x0020,0x0020,0x001f,0x001e,0x001e,0x001d,0x001d,0x001c,0x001c,0x001b,
+0x001b,0x001a,0x001a,0x0019,0x0019,0x0018,0x0018,0x0017,0x0017,0x0016,
+0x0016,0x0015,0x0015,0x0015,0x0014,0x0014,0x0013,0x0013,0x0013,0x0012,
+0x0012,0x0012,0x0011,0x0011,0x0011,0x0010,0x0010,0x0010,0x000f,0x000f,
+0x000f,0x000e,0x000e,0x000e,0x000d,0x000d,0x000d,0x000d,0x000c,0x000c,
+0x000c,0x000c,0x000b,0x000b,0x000b,0x000b,0x000a,0x000a,0x000a,0x000a,
+0x000a,0x0009,0x0009,0x0009,0x0009,0x0009,0x0008,0x0008,0x0008,0x0008,
+0x0008,0x0008,0x0007,0x0007,0x0007,0x0007,0x0007,0x0007,0x0006,0x0006,
+0x0006,0x0006,0x0006,0x0006,0x0006,0x0006,0x0005,0x0005,0x0005,0x0005,
+0x0005,0x0005,0x0005,0x0005,0x0004,0x0004,0x0004,0x0004,0x0004,0x0004,
+0x0004,0x0004,0x0004,0x0004,0x0004,0x0003,0x0003,0x0003,0x0003,0x0003,
+0x0003,0x0003,0x0003,0x0003,0x0003,0x0003,0x0003,0x0003,0x0003,0x0002,
+0x0002,0x0002,0x0002,0x0002,0x0002,0x0002,0x0002,0x0002,0x0002,0x0002,
+0x0002,0x0002,0x0002,0x0002,0x0002,0x0002,0x0002,0x0002,0x0001,0x0001,
+0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,
+0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,
+0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,
+0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
+0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
+0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
+0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
+0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
+};
+
+static const UINT16 hth[50][3]= {
+{ 0x04d0,0x04f0,0x0580 },
+{ 0x04d0,0x04f0,0x0580 },
+{ 0x0440,0x0460,0x04b0 },
+{ 0x0400,0x0410,0x0450 },
+{ 0x03e0,0x03e0,0x0420 },
+{ 0x03c0,0x03d0,0x03f0 },
+{ 0x03b0,0x03c0,0x03e0 },
+{ 0x03b0,0x03b0,0x03d0 },
+{ 0x03a0,0x03b0,0x03c0 },
+{ 0x03a0,0x03a0,0x03b0 },
+{ 0x03a0,0x03a0,0x03b0 },
+{ 0x03a0,0x03a0,0x03b0 },
+{ 0x03a0,0x03a0,0x03a0 },
+{ 0x0390,0x03a0,0x03a0 },
+{ 0x0390,0x0390,0x03a0 },
+{ 0x0390,0x0390,0x03a0 },
+{ 0x0380,0x0390,0x03a0 },
+{ 0x0380,0x0380,0x03a0 },
+{ 0x0370,0x0380,0x03a0 },
+{ 0x0370,0x0380,0x03a0 },
+{ 0x0360,0x0370,0x0390 },
+{ 0x0360,0x0370,0x0390 },
+{ 0x0350,0x0360,0x0390 },
+{ 0x0350,0x0360,0x0390 },
+{ 0x0340,0x0350,0x0380 },
+{ 0x0340,0x0350,0x0380 },
+{ 0x0330,0x0340,0x0380 },
+{ 0x0320,0x0340,0x0370 },
+{ 0x0310,0x0320,0x0360 },
+{ 0x0300,0x0310,0x0350 },
+{ 0x02f0,0x0300,0x0340 },
+{ 0x02f0,0x02f0,0x0330 },
+{ 0x02f0,0x02f0,0x0320 },
+{ 0x02f0,0x02f0,0x0310 },
+{ 0x0300,0x02f0,0x0300 },
+{ 0x0310,0x0300,0x02f0 },
+{ 0x0340,0x0320,0x02f0 },
+{ 0x0390,0x0350,0x02f0 },
+{ 0x03e0,0x0390,0x0300 },
+{ 0x0420,0x03e0,0x0310 },
+{ 0x0460,0x0420,0x0330 },
+{ 0x0490,0x0450,0x0350 },
+{ 0x04a0,0x04a0,0x03c0 },
+{ 0x0460,0x0490,0x0410 },
+{ 0x0440,0x0460,0x0470 },
+{ 0x0440,0x0440,0x04a0 },
+{ 0x0520,0x0480,0x0460 },
+{ 0x0800,0x0630,0x0440 },
+{ 0x0840,0x0840,0x0450 },
+{ 0x0840,0x0840,0x04e0 },
+};
+
+static const UINT8 baptab[64]= {
+    0, 1, 1, 1, 1, 1, 2, 2, 3, 3, 
+    3, 4, 4, 5, 5, 6, 6, 6, 6, 7, 
+    7, 7, 7, 8, 8, 8, 8, 9, 9, 9, 
+    9, 10, 10, 10, 10, 11, 11, 11, 11, 12, 
+    12, 12, 12, 13, 13, 13, 13, 14, 14, 14, 
+    14, 14, 14, 14, 14, 15, 15, 15, 15, 15, 
+    15, 15, 15, 15,
+};
+
+static const UINT8 sdecaytab[4]={ 
+    0x0f, 0x11, 0x13, 0x15,
+};
+
+static const UINT8 fdecaytab[4]={ 
+    0x3f, 0x53, 0x67, 0x7b, 
+};
+
+static const UINT16 sgaintab[4]= { 
+    0x540, 0x4d8, 0x478, 0x410,
+};
+
+static const UINT16 dbkneetab[4]= { 
+    0x000, 0x700, 0x900, 0xb00,
+};
+
+static const UINT16 floortab[8]= { 
+    0x2f0, 0x2b0, 0x270, 0x230, 0x1f0, 0x170, 0x0f0, 0xf800,
+};
+
+static const UINT16 fgaintab[8]= {
+    0x080, 0x100, 0x180, 0x200, 0x280, 0x300, 0x380, 0x400,
+};
+
+static const UINT8 bndsz[50]={
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 3, 3, 3, 3, 3, 3, 
+    3, 6, 6, 6, 6, 6, 6, 12, 12, 12, 12, 24, 24, 24, 24, 24 
+};
+
+static UINT8 bndtab[51]; 
+
+/* fft & mdct sin cos tables */
+static INT16 costab[64];
+static INT16 sintab[64];
+static INT16 fft_rev[512];
+static INT16 xcos1[128];
+static INT16 xsin1[128];
+
+static UINT16 crc_table[256];
diff --git a/libavcodec/apiexample.c b/libavcodec/apiexample.c
new file mode 100644
index 0000000000..e7ae6e80be
--- /dev/null
+++ b/libavcodec/apiexample.c
@@ -0,0 +1,407 @@
+/* avcodec API use example.
+ *
+ * Note that this library only handles codecs (mpeg, mpeg4, etc...),
+ * not file formats (avi, vob, etc...). See library 'libav' for the
+ * format handling 
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <math.h>
+
+#include "avcodec.h"
+
+#define INBUF_SIZE 4096
+
+/*
+ * Audio encoding example 
+ */
+void audio_encode_example(const char *filename)
+{
+    AVCodec *codec;
+    AVCodecContext codec_context, *c = &codec_context;
+    int frame_size, i, j, out_size, outbuf_size;
+    FILE *f;
+    short *samples;
+    float t, tincr;
+    UINT8 *outbuf;
+
+    printf("Audio encoding\n");
+
+    /* find the MP2 encoder */
+    codec = avcodec_find_encoder(CODEC_ID_MP2);
+    if (!codec) {
+        fprintf(stderr, "codec not found\n");
+        exit(1);
+    }
+
+    /* put default values */
+    memset(c, 0, sizeof(*c));
+
+    /* put sample parameters */
+    c->bit_rate = 64000;
+    c->sample_rate = 44100;
+    c->channels = 2;
+
+    /* open it */
+    if (avcodec_open(c, codec) < 0) {
+        fprintf(stderr, "could not open codec\n");
+        exit(1);
+    }
+    
+    /* the codec gives us the frame size, in samples */
+    frame_size = c->frame_size;
+    samples = malloc(frame_size * 2 * c->channels);
+    outbuf_size = 10000;
+    outbuf = malloc(outbuf_size);
+
+    f = fopen(filename, "w");
+    if (!f) {
+        fprintf(stderr, "could not open %s\n", filename);
+        exit(1);
+    }
+        
+    /* encode a single tone sound */
+    t = 0;
+    tincr = 2 * M_PI * 440.0 / c->sample_rate;
+    for(i=0;i<200;i++) {
+        for(j=0;j<frame_size;j++) {
+            samples[2*j] = (int)(sin(t) * 10000);
+            samples[2*j+1] = samples[2*j];
+            t += tincr;
+        }
+        /* encode the samples */
+        out_size = avcodec_encode_audio(c, outbuf, outbuf_size, samples);
+        fwrite(outbuf, 1, out_size, f);
+    }
+    fclose(f);
+    free(outbuf);
+    free(samples);
+
+    avcodec_close(c);
+}
+
+/*
+ * Audio decoding. 
+ */
+void audio_decode_example(const char *outfilename, const char *filename)
+{
+    AVCodec *codec;
+    AVCodecContext codec_context, *c = &codec_context;
+    int out_size, size, len;
+    FILE *f, *outfile;
+    UINT8 *outbuf;
+    UINT8 inbuf[INBUF_SIZE], *inbuf_ptr;
+
+    printf("Audio decoding\n");
+
+    /* find the mpeg audio decoder */
+    codec = avcodec_find_decoder(CODEC_ID_MP2);
+    if (!codec) {
+        fprintf(stderr, "codec not found\n");
+        exit(1);
+    }
+
+    /* put default values */
+    memset(c, 0, sizeof(*c));
+
+    /* open it */
+    if (avcodec_open(c, codec) < 0) {
+        fprintf(stderr, "could not open codec\n");
+        exit(1);
+    }
+    
+    outbuf = malloc(AVCODEC_MAX_AUDIO_FRAME_SIZE);
+
+    f = fopen(filename, "r");
+    if (!f) {
+        fprintf(stderr, "could not open %s\n", filename);
+        exit(1);
+    }
+    outfile = fopen(outfilename, "w");
+    if (!outfile) {
+        fprintf(stderr, "could not open %s\n", outfilename);
+        exit(1);
+    }
+        
+    /* decode until eof */
+    inbuf_ptr = inbuf;
+    for(;;) {
+        size = fread(inbuf, 1, INBUF_SIZE, f);
+        if (size == 0)
+            break;
+
+        inbuf_ptr = inbuf;
+        while (size > 0) {
+            len = avcodec_decode_audio(c, (short *)outbuf, &out_size, 
+                                       inbuf_ptr, size);
+            if (len < 0) {
+                fprintf(stderr, "Error while decoding\n");
+                exit(1);
+            }
+            if (out_size > 0) {
+                /* if a frame has been decoded, output it */
+                fwrite(outbuf, 1, out_size, outfile);
+            }
+            size -= len;
+            inbuf_ptr += len;
+        }
+    }
+
+    fclose(outfile);
+    fclose(f);
+    free(outbuf);
+
+    avcodec_close(c);
+}
+
+/*
+ * Video encoding example 
+ */
+void video_encode_example(const char *filename)
+{
+    AVCodec *codec;
+    AVCodecContext codec_context, *c = &codec_context;
+    int i, out_size, size, x, y, outbuf_size;
+    FILE *f;
+    AVPicture picture;
+    UINT8 *outbuf, *picture_buf;
+
+    printf("Video encoding\n");
+
+    /* find the mpeg1 video encoder */
+    codec = avcodec_find_encoder(CODEC_ID_MPEG1VIDEO);
+    if (!codec) {
+        fprintf(stderr, "codec not found\n");
+        exit(1);
+    }
+
+    /* put default values */
+    memset(c, 0, sizeof(*c));
+
+    /* put sample parameters */
+    c->bit_rate = 400000;
+    /* resolution must be a multiple of two */
+    c->width = 352;  
+    c->height = 288;
+    /* frames per second */
+    c->frame_rate = 25 * FRAME_RATE_BASE;  
+    c->gop_size = 10; /* emit one intra frame every ten frames */
+
+    /* open it */
+    if (avcodec_open(c, codec) < 0) {
+        fprintf(stderr, "could not open codec\n");
+        exit(1);
+    }
+    
+    /* the codec gives us the frame size, in samples */
+
+    f = fopen(filename, "w");
+    if (!f) {
+        fprintf(stderr, "could not open %s\n", filename);
+        exit(1);
+    }
+    
+    /* alloc image and output buffer */
+    outbuf_size = 100000;
+    outbuf = malloc(outbuf_size);
+    size = c->width * c->height;
+    picture_buf = malloc((size * 3) / 2); /* size for YUV 420 */
+    
+    picture.data[0] = picture_buf;
+    picture.data[1] = picture.data[0] + size;
+    picture.data[2] = picture.data[1] + size / 4;
+    picture.linesize[0] = c->width;
+    picture.linesize[1] = c->width / 2;
+    picture.linesize[2] = c->width / 2;
+
+    /* encode 1 second of video */
+    for(i=0;i<25;i++) {
+        printf("encoding frame %3d\r", i);
+        fflush(stdout);
+        /* prepare a dummy image */
+        /* Y */
+        for(y=0;y<c->height;y++) {
+            for(x=0;x<c->width;x++) {
+                picture.data[0][y * picture.linesize[0] + x] = x + y + i * 3;
+            }
+        }
+
+        /* Cb and Cr */
+        for(y=0;y<c->height/2;y++) {
+            for(x=0;x<c->width/2;x++) {
+                picture.data[1][y * picture.linesize[1] + x] = 128 + y + i * 2;
+                picture.data[2][y * picture.linesize[2] + x] = 64 + x + i * 5;
+            }
+        }
+
+        /* encode the image */
+        out_size = avcodec_encode_video(c, outbuf, outbuf_size, &picture);
+        fwrite(outbuf, 1, out_size, f);
+    }
+
+    /* add sequence end code to have a real mpeg file */
+    outbuf[0] = 0x00;
+    outbuf[1] = 0x00;
+    outbuf[2] = 0x01;
+    outbuf[3] = 0xb7;
+    fwrite(outbuf, 1, 4, f);
+    fclose(f);
+    free(picture_buf);
+    free(outbuf);
+
+    avcodec_close(c);
+    printf("\n");
+}
+
+/*
+ * Video decoding example 
+ */
+
+void pgm_save(unsigned char *buf,int wrap, int xsize,int ysize,char *filename) 
+{
+    FILE *f;
+    int i;
+
+    f=fopen(filename,"w");
+    fprintf(f,"P5\n%d %d\n%d\n",xsize,ysize,255);
+    for(i=0;i<ysize;i++)
+        fwrite(buf + i * wrap,1,xsize,f);
+    fclose(f);
+}
+
+void video_decode_example(const char *outfilename, const char *filename)
+{
+    AVCodec *codec;
+    AVCodecContext codec_context, *c = &codec_context;
+    int frame, size, got_picture, len;
+    FILE *f;
+    AVPicture picture;
+    UINT8 inbuf[INBUF_SIZE], *inbuf_ptr;
+    char buf[1024];
+
+    printf("Video decoding\n");
+
+    /* find the mpeg1 video decoder */
+    codec = avcodec_find_decoder(CODEC_ID_MPEG1VIDEO);
+    if (!codec) {
+        fprintf(stderr, "codec not found\n");
+        exit(1);
+    }
+
+    /* put default values */
+    memset(c, 0, sizeof(*c));
+
+    /* for some codecs, such as msmpeg4 and opendivx, width and height
+       MUST be initialized there because these info are not available
+       in the bitstream */
+
+    /* open it */
+    if (avcodec_open(c, codec) < 0) {
+        fprintf(stderr, "could not open codec\n");
+        exit(1);
+    }
+    
+    /* the codec gives us the frame size, in samples */
+
+    f = fopen(filename, "r");
+    if (!f) {
+        fprintf(stderr, "could not open %s\n", filename);
+        exit(1);
+    }
+    
+    frame = 0;
+    for(;;) {
+        size = fread(inbuf, 1, INBUF_SIZE, f);
+        if (size == 0)
+            break;
+
+        /* NOTE1: some codecs are stream based (mpegvideo, mpegaudio)
+           and this is the only method to use them because you cannot
+           know the compressed data size before analysing it. 
+
+           BUT some other codecs (msmpeg4, opendivx) are inherently
+           frame based, so you must call them with all the data for
+           one frame exactly. You must also initialize 'width' and
+           'height' before initializing them. */
+
+        /* NOTE2: some codecs allow the raw parameters (frame size,
+           sample rate) to be changed at any frame. We handle this, so
+           you should also take care of it */
+
+        /* here, we use a stream based decoder (mpeg1video), so we
+           feed decoder and see if it could decode a frame */
+        inbuf_ptr = inbuf;
+        while (size > 0) {
+            len = avcodec_decode_video(c, &picture, &got_picture, 
+                                       inbuf_ptr, size);
+            if (len < 0) {
+                fprintf(stderr, "Error while decoding frame %d\n", frame);
+                exit(1);
+            }
+            if (got_picture) {
+                printf("saving frame %3d\r", frame);
+                fflush(stdout);
+
+                /* the picture is allocated by the decoder. no need to
+                   free it */
+                snprintf(buf, sizeof(buf), outfilename, frame);
+                pgm_save(picture.data[0], picture.linesize[0], 
+                         c->width, c->height, buf);
+                frame++;
+            }
+            size -= len;
+            inbuf_ptr += len;
+        }
+    }
+
+    /* some codecs, such as MPEG, transmit the I and P frame with a
+       latency of one frame. You must do the following to have a
+       chance to get the last frame of the video */
+    len = avcodec_decode_video(c, &picture, &got_picture, 
+                               NULL, 0);
+    if (got_picture) {
+        printf("saving frame %3d\r", frame);
+        fflush(stdout);
+        
+        /* the picture is allocated by the decoder. no need to
+           free it */
+        snprintf(buf, sizeof(buf), outfilename, frame);
+        pgm_save(picture.data[0], picture.linesize[0], 
+                 c->width, c->height, buf);
+        frame++;
+    }
+        
+    fclose(f);
+
+    avcodec_close(c);
+    printf("\n");
+}
+
+
+int main(int argc, char **argv)
+{
+    const char *filename;
+
+    /* must be called before using avcodec lib */
+    avcodec_init();
+
+    /* register all the codecs (you can also register only the codec
+       you wish to have smaller code */
+    avcodec_register_all();
+    
+    if (argc <= 1) {
+        audio_encode_example("/tmp/test.mp2");
+        audio_decode_example("/tmp/test.sw", "/tmp/test.mp2");
+
+        video_encode_example("/tmp/test.mpg");
+        filename = "/tmp/test.mpg";
+    } else {
+        filename = argv[1];
+    }
+
+    //    audio_decode_example("/tmp/test.sw", filename);
+    video_decode_example("/tmp/test%d.pgm", filename);
+
+    return 0;
+}
diff --git a/libavcodec/avcodec.h b/libavcodec/avcodec.h
new file mode 100644
index 0000000000..8516d2a176
--- /dev/null
+++ b/libavcodec/avcodec.h
@@ -0,0 +1,177 @@
+#include "common.h"
+
+enum CodecID {
+    CODEC_ID_NONE, 
+    CODEC_ID_MPEG1VIDEO,
+    CODEC_ID_H263,
+    CODEC_ID_RV10,
+    CODEC_ID_MP2,
+    CODEC_ID_AC3,
+    CODEC_ID_MJPEG,
+    CODEC_ID_OPENDIVX,
+    CODEC_ID_PCM,
+    CODEC_ID_RAWVIDEO,
+    CODEC_ID_MSMPEG4,
+    CODEC_ID_H263P,
+    CODEC_ID_H263I,
+};
+
+enum CodecType {
+    CODEC_TYPE_VIDEO,
+    CODEC_TYPE_AUDIO,
+};
+
+enum PixelFormat {
+    PIX_FMT_YUV420P,
+    PIX_FMT_YUV422,
+    PIX_FMT_RGB24,
+    PIX_FMT_BGR24,
+};
+
+/* in bytes */
+#define AVCODEC_MAX_AUDIO_FRAME_SIZE 18432
+
+/* motion estimation type */
+extern int motion_estimation_method;
+#define ME_ZERO   0
+#define ME_FULL   1
+#define ME_LOG    2
+#define ME_PHODS  3
+
+/* encoding support */
+
+#define CODEC_FLAG_HQ     0x0001 /* high quality (non real time) encoding */
+#define CODEC_FLAG_QSCALE 0x0002 /* use fixed qscale */
+
+#define FRAME_RATE_BASE 10000
+
+typedef struct AVCodecContext {
+    int bit_rate;
+    int flags;
+    int sub_id;    /* some codecs needs additionnal format info. It is
+                      stored there */
+    /* video only */
+    int frame_rate; /* frames per sec multiplied by FRAME_RATE_BASE */
+    int width, height;
+    int gop_size; /* 0 = intra only */
+    int pix_fmt;  /* pixel format, see PIX_FMT_xxx */
+    /* audio only */
+    int sample_rate; /* samples per sec */
+    int channels;
+
+    /* the following data should not be initialized */
+    int frame_size; /* in samples, initialized when calling 'init' */
+    int frame_number; /* audio or video frame number */
+    int key_frame;    /* true if the previous compressed frame was 
+                         a key frame (intra, or seekable) */
+    int quality;      /* quality of the previous encoded frame 
+                         (between 1 (good) and 31 (bad)) */
+    struct AVCodec *codec;
+    void *priv_data;
+
+    /* the following fields are ignored */
+    char codec_name[32];
+    int codec_type; /* see CODEC_TYPE_xxx */
+    int codec_id; /* see CODEC_ID_xxx */
+    unsigned int codec_tag;  /* codec tag, only used if unknown codec */
+} AVCodecContext;
+
+typedef struct AVCodec {
+    char *name;
+    int type;
+    int id;
+    int priv_data_size;
+    int (*init)(AVCodecContext *);
+    int (*encode)(AVCodecContext *, UINT8 *buf, int buf_size, void *data);
+    int (*close)(AVCodecContext *);
+    int (*decode)(AVCodecContext *, void *outdata, int *outdata_size, 
+                  UINT8 *buf, int buf_size);
+    struct AVCodec *next;
+} AVCodec;
+
+/* three components are given, that's all */
+typedef struct AVPicture {
+    UINT8 *data[3];
+    int linesize[3];
+} AVPicture;
+
+extern AVCodec ac3_encoder;
+extern AVCodec mp2_encoder;
+extern AVCodec mpeg1video_encoder;
+extern AVCodec h263_encoder;
+extern AVCodec h263p_encoder;
+extern AVCodec rv10_encoder;
+extern AVCodec mjpeg_encoder;
+extern AVCodec opendivx_encoder;
+extern AVCodec msmpeg4_encoder;
+
+extern AVCodec h263_decoder;
+extern AVCodec opendivx_decoder;
+extern AVCodec msmpeg4_decoder;
+extern AVCodec mpeg_decoder;
+extern AVCodec h263i_decoder;
+extern AVCodec rv10_decoder;
+
+/* dummy raw codecs */
+extern AVCodec pcm_codec;
+extern AVCodec rawvideo_codec;
+
+/* the following codecs use external GPL libs */
+extern AVCodec mp3_decoder;
+extern AVCodec ac3_decoder;
+
+/* resample.c */
+
+struct ReSampleContext;
+
+typedef struct ReSampleContext ReSampleContext;
+
+ReSampleContext *audio_resample_init(int output_channels, int input_channels, 
+                                     int output_rate, int input_rate);
+int audio_resample(ReSampleContext *s, short *output, short *input, int nb_samples);
+void audio_resample_close(ReSampleContext *s);
+
+/* YUV420 format is assumed ! */
+
+struct ImgReSampleContext;
+
+typedef struct ImgReSampleContext ImgReSampleContext;
+
+ImgReSampleContext *img_resample_init(int output_width, int output_height,
+                                      int input_width, int input_height);
+void img_resample(ImgReSampleContext *s, 
+                  AVPicture *output, AVPicture *input);
+
+void img_resample_close(ImgReSampleContext *s);
+
+int img_convert_to_yuv420(UINT8 *img_out, UINT8 *img, 
+                          int pix_fmt, int width, int height);
+
+/* external high level API */
+
+extern AVCodec *first_avcodec;
+
+void avcodec_init(void);
+
+void register_avcodec(AVCodec *format);
+AVCodec *avcodec_find_encoder(enum CodecID id);
+AVCodec *avcodec_find_decoder(enum CodecID id);
+AVCodec *avcodec_find_decoder_by_name(const char *name);
+void avcodec_string(char *buf, int buf_size, AVCodecContext *enc, int encode);
+
+int avcodec_open(AVCodecContext *avctx, AVCodec *codec);
+int avcodec_decode_audio(AVCodecContext *avctx, INT16 *samples, 
+                         int *frame_size_ptr,
+                         UINT8 *buf, int buf_size);
+int avcodec_decode_video(AVCodecContext *avctx, AVPicture *picture, 
+                         int *got_picture_ptr,
+                         UINT8 *buf, int buf_size);
+int avcodec_encode_audio(AVCodecContext *avctx, UINT8 *buf, int buf_size, 
+                         const short *samples);
+int avcodec_encode_video(AVCodecContext *avctx, UINT8 *buf, int buf_size, 
+                         const AVPicture *pict);
+
+int avcodec_close(AVCodecContext *avctx);
+
+void avcodec_register_all(void);
+void avcodec_register_more(void);
diff --git a/libavcodec/common.c b/libavcodec/common.c
new file mode 100644
index 0000000000..b9365c9d8d
--- /dev/null
+++ b/libavcodec/common.c
@@ -0,0 +1,469 @@
+/*
+ * Common bit i/o utils
+ * Copyright (c) 2000, 2001 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#ifdef __FreeBSD__
+#include <sys/param.h>
+#endif
+#include <netinet/in.h>
+#include <math.h>
+#include "common.h"
+
+#define NDEBUG
+#include <assert.h>
+
+void init_put_bits(PutBitContext *s, 
+                   UINT8 *buffer, int buffer_size,
+                   void *opaque,
+                   void (*write_data)(void *, UINT8 *, int))
+{
+    s->buf = buffer;
+    s->buf_ptr = s->buf;
+    s->buf_end = s->buf + buffer_size;
+    s->bit_cnt=0;
+    s->bit_buf=0;
+    s->data_out_size = 0;
+    s->write_data = write_data;
+    s->opaque = opaque;
+}
+
+static void flush_buffer(PutBitContext *s)
+{
+    int size;
+    if (s->write_data) {
+        size = s->buf_ptr - s->buf;
+        if (size > 0)
+            s->write_data(s->opaque, s->buf, size);
+        s->buf_ptr = s->buf;
+        s->data_out_size += size;
+    }
+}
+
+void put_bits(PutBitContext *s, int n, unsigned int value)
+{
+    unsigned int bit_buf;
+    int bit_cnt;
+
+#ifdef STATS
+    st_out_bit_counts[st_current_index] += n;
+#endif
+    //    printf("put_bits=%d %x\n", n, value);
+    assert(n == 32 || value < (1U << n));
+
+    bit_buf = s->bit_buf;
+    bit_cnt = s->bit_cnt;
+
+    //    printf("n=%d value=%x cnt=%d buf=%x\n", n, value, bit_cnt, bit_buf);
+    /* XXX: optimize */
+    if (n < (32-bit_cnt)) {
+        bit_buf |= value << (32 - n - bit_cnt);
+        bit_cnt+=n;
+    } else {
+        bit_buf |= value >> (n + bit_cnt - 32);
+        *(UINT32 *)s->buf_ptr = htonl(bit_buf);
+        //printf("bitbuf = %08x\n", bit_buf);
+        s->buf_ptr+=4;
+        if (s->buf_ptr >= s->buf_end)
+            flush_buffer(s);
+        bit_cnt=bit_cnt + n - 32;
+        if (bit_cnt == 0) {
+            bit_buf = 0;
+        } else {
+            bit_buf = value << (32 - bit_cnt);
+        }
+    }
+    
+    s->bit_buf = bit_buf;
+    s->bit_cnt = bit_cnt;
+}
+
+/* return the number of bits output */
+long long get_bit_count(PutBitContext *s)
+{
+    return (s->buf_ptr - s->buf + s->data_out_size) * 8 + (long long)s->bit_cnt;
+}
+
+void align_put_bits(PutBitContext *s)
+{
+    put_bits(s,(8 - s->bit_cnt) & 7,0);
+}
+
+/* pad the end of the output stream with zeros */
+void flush_put_bits(PutBitContext *s)
+{
+    while (s->bit_cnt > 0) {
+        /* XXX: should test end of buffer */
+        *s->buf_ptr++=s->bit_buf >> 24;
+        s->bit_buf<<=8;
+        s->bit_cnt-=8;
+    }
+    flush_buffer(s);
+    s->bit_cnt=0;
+    s->bit_buf=0;
+}
+
+/* for jpeg : espace 0xff with 0x00 after it */
+void jput_bits(PutBitContext *s, int n, unsigned int value)
+{
+    unsigned int bit_buf, b;
+    int bit_cnt, i;
+    
+    assert(n == 32 || value < (1U << n));
+
+    bit_buf = s->bit_buf;
+    bit_cnt = s->bit_cnt;
+
+    //printf("n=%d value=%x cnt=%d buf=%x\n", n, value, bit_cnt, bit_buf);
+    /* XXX: optimize */
+    if (n < (32-bit_cnt)) {
+        bit_buf |= value << (32 - n - bit_cnt);
+        bit_cnt+=n;
+    } else {
+        bit_buf |= value >> (n + bit_cnt - 32);
+        /* handle escape */
+        for(i=0;i<4;i++) {
+            b = (bit_buf >> 24);
+            *(s->buf_ptr++) = b;
+            if (b == 0xff)
+                *(s->buf_ptr++) = 0;
+            bit_buf <<= 8;
+        }
+        /* we flush the buffer sooner to handle worst case */
+        if (s->buf_ptr >= (s->buf_end - 8))
+            flush_buffer(s);
+
+        bit_cnt=bit_cnt + n - 32;
+        if (bit_cnt == 0) {
+            bit_buf = 0;
+        } else {
+            bit_buf = value << (32 - bit_cnt);
+        }
+    }
+    
+    s->bit_buf = bit_buf;
+    s->bit_cnt = bit_cnt;
+}
+
+/* pad the end of the output stream with zeros */
+void jflush_put_bits(PutBitContext *s)
+{
+    unsigned int b;
+
+    while (s->bit_cnt > 0) {
+        b = s->bit_buf >> 24;
+        *s->buf_ptr++ = b;
+        if (b == 0xff)
+            *s->buf_ptr++ = 0;
+        s->bit_buf<<=8;
+        s->bit_cnt-=8;
+    }
+    flush_buffer(s);
+    s->bit_cnt=0;
+    s->bit_buf=0;
+}
+
+/* bit input functions */
+
+void init_get_bits(GetBitContext *s, 
+                   UINT8 *buffer, int buffer_size)
+{
+    s->buf = buffer;
+    s->buf_ptr = buffer;
+    s->buf_end = buffer + buffer_size;
+    s->bit_cnt = 0;
+    s->bit_buf = 0;
+    while (s->buf_ptr < s->buf_end && 
+           s->bit_cnt < 32) {
+        s->bit_buf |= (*s->buf_ptr++ << (24 - s->bit_cnt));
+        s->bit_cnt += 8;
+    }
+}
+
+/* n must be >= 1 and <= 32 */
+unsigned int get_bits(GetBitContext *s, int n)
+{
+    unsigned int val;
+    int bit_cnt;
+    unsigned int bit_buf;
+    UINT8 *buf_ptr;
+
+#ifdef STATS
+    st_bit_counts[st_current_index] += n;
+#endif
+
+    bit_cnt = s->bit_cnt;
+    bit_buf = s->bit_buf;
+    
+    bit_cnt -= n;
+    if (bit_cnt >= 0) {
+        /* most common case here */
+        val = bit_buf >> (32 - n);
+        bit_buf <<= n; 
+    } else {
+        val = bit_buf >> (32 - n);
+        buf_ptr = s->buf_ptr;
+        buf_ptr += 4;
+        /* handle common case: we can read everything */
+        if (buf_ptr <= s->buf_end) {
+            bit_buf = (buf_ptr[-4] << 24) |
+                (buf_ptr[-3] << 16) |
+                (buf_ptr[-2] << 8) |
+                (buf_ptr[-1]);
+        } else {
+            buf_ptr -= 4;
+            bit_buf = 0;
+            if (buf_ptr < s->buf_end)
+                bit_buf |= *buf_ptr++ << 24;
+            if (buf_ptr < s->buf_end)
+                bit_buf |= *buf_ptr++ << 16;
+            if (buf_ptr < s->buf_end)
+                bit_buf |= *buf_ptr++ << 8;
+            if (buf_ptr < s->buf_end)
+                bit_buf |= *buf_ptr++;
+        }
+        s->buf_ptr = buf_ptr;
+        val |= bit_buf >> (32 + bit_cnt);
+        bit_buf <<= - bit_cnt;
+        bit_cnt += 32;
+    }
+    s->bit_buf = bit_buf;
+    s->bit_cnt = bit_cnt;
+    return val;
+}
+
+void align_get_bits(GetBitContext *s)
+{
+    int n;
+    n = s->bit_cnt & 7;
+    if (n > 0) {
+        get_bits(s, n);
+    }
+}
+
+/* VLC decoding */
+
+//#define DEBUG_VLC
+
+#define GET_DATA(v, table, i, wrap, size) \
+{\
+    UINT8 *ptr = (UINT8 *)table + i * wrap;\
+    switch(size) {\
+    case 1:\
+        v = *(UINT8 *)ptr;\
+        break;\
+    case 2:\
+        v = *(UINT16 *)ptr;\
+        break;\
+    default:\
+        v = *(UINT32 *)ptr;\
+        break;\
+    }\
+}
+
+
+static int alloc_table(VLC *vlc, int size)
+{
+    int index;
+    index = vlc->table_size;
+    vlc->table_size += size;
+    if (vlc->table_size > vlc->table_allocated) {
+        vlc->table_allocated += (1 << vlc->bits);
+        vlc->table_bits = realloc(vlc->table_bits, 
+                                  sizeof(INT8) * vlc->table_allocated);
+        vlc->table_codes = realloc(vlc->table_codes,
+                                   sizeof(INT16) * vlc->table_allocated);
+        if (!vlc->table_bits ||
+            !vlc->table_codes)
+            return -1;
+    }
+    return index;
+}
+
+static int build_table(VLC *vlc, int table_nb_bits, 
+                       int nb_codes,
+                       const void *bits, int bits_wrap, int bits_size,
+                       const void *codes, int codes_wrap, int codes_size,
+                       UINT32 code_prefix, int n_prefix)
+{
+    int i, j, k, n, table_size, table_index, nb, n1, index;
+    UINT32 code;
+    INT8 *table_bits;
+    INT16 *table_codes;
+
+    table_size = 1 << table_nb_bits;
+    table_index = alloc_table(vlc, table_size);
+#ifdef DEBUG_VLC
+    printf("new table index=%d size=%d code_prefix=%x n=%d\n", 
+           table_index, table_size, code_prefix, n_prefix);
+#endif
+    if (table_index < 0)
+        return -1;
+    table_bits = &vlc->table_bits[table_index];
+    table_codes = &vlc->table_codes[table_index];
+
+    for(i=0;i<table_size;i++) {
+        table_bits[i] = 0;
+        table_codes[i] = -1;
+    }
+
+    /* first pass: map codes and compute auxillary table sizes */
+    for(i=0;i<nb_codes;i++) {
+        GET_DATA(n, bits, i, bits_wrap, bits_size);
+        GET_DATA(code, codes, i, codes_wrap, codes_size);
+        /* we accept tables with holes */
+        if (n <= 0)
+            continue;
+#if defined(DEBUG_VLC) && 0
+        printf("i=%d n=%d code=0x%x\n", i, n, code);
+#endif
+        /* if code matches the prefix, it is in the table */
+        n -= n_prefix;
+        if (n > 0 && (code >> n) == code_prefix) {
+            if (n <= table_nb_bits) {
+                /* no need to add another table */
+                j = (code << (table_nb_bits - n)) & (table_size - 1);
+                nb = 1 << (table_nb_bits - n);
+                for(k=0;k<nb;k++) {
+#ifdef DEBUG_VLC
+                    printf("%4x: code=%d n=%d\n",
+                           j, i, n);
+#endif
+                    if (table_bits[j] != 0) {
+                        fprintf(stderr, "incorrect codes\n");
+                        exit(1);
+                    }
+                    table_bits[j] = n;
+                    table_codes[j] = i;
+                    j++;
+                }
+            } else {
+                n -= table_nb_bits;
+                j = (code >> n) & ((1 << table_nb_bits) - 1);
+#ifdef DEBUG_VLC
+                printf("%4x: n=%d (subtable)\n",
+                       j, n);
+#endif
+                /* compute table size */
+                n1 = -table_bits[j];
+                if (n > n1)
+                    n1 = n;
+                table_bits[j] = -n1;
+            }
+        }
+    }
+
+    /* second pass : fill auxillary tables recursively */
+    for(i=0;i<table_size;i++) {
+        n = table_bits[i];
+        if (n < 0) {
+            n = -n;
+            if (n > table_nb_bits) {
+                n = table_nb_bits;
+                table_bits[i] = -n;
+            }
+            index = build_table(vlc, n, nb_codes,
+                                bits, bits_wrap, bits_size,
+                                codes, codes_wrap, codes_size,
+                                (code_prefix << table_nb_bits) | i,
+                                n_prefix + table_nb_bits);
+            if (index < 0)
+                return -1;
+            /* note: realloc has been done, so reload tables */
+            table_bits = &vlc->table_bits[table_index];
+            table_codes = &vlc->table_codes[table_index];
+            table_codes[i] = index;
+        }
+    }
+    return table_index;
+}
+
+
+/* wrap and size allow to handle most types of storage.  */
+int init_vlc(VLC *vlc, int nb_bits, int nb_codes,
+             const void *bits, int bits_wrap, int bits_size,
+             const void *codes, int codes_wrap, int codes_size)
+{
+    vlc->bits = nb_bits;
+    vlc->table_bits = NULL;
+    vlc->table_codes = NULL;
+    vlc->table_allocated = 0;
+    vlc->table_size = 0;
+#ifdef DEBUG_VLC
+    printf("build table nb_codes=%d\n", nb_codes);
+#endif
+
+    if (build_table(vlc, nb_bits, nb_codes,
+                    bits, bits_wrap, bits_size,
+                    codes, codes_wrap, codes_size,
+                    0, 0) < 0) {
+        if (vlc->table_bits)
+            free(vlc->table_bits);
+        if (vlc->table_codes)
+            free(vlc->table_codes);
+        return -1;
+    }
+    return 0;
+}
+
+
+void free_vlc(VLC *vlc)
+{
+    free(vlc->table_bits);
+    free(vlc->table_codes);
+}
+
+int get_vlc(GetBitContext *s, VLC *vlc)
+{
+    int bit_cnt, code, n, nb_bits, index;
+    UINT32 bit_buf;
+    INT16 *table_codes;
+    INT8 *table_bits;
+    UINT8 *buf_ptr;
+
+    SAVE_BITS(s);
+    nb_bits = vlc->bits;
+    table_codes = vlc->table_codes;
+    table_bits = vlc->table_bits;
+    for(;;) {
+        SHOW_BITS(s, index, nb_bits);
+        code = table_codes[index];
+        n = table_bits[index];
+        if (n > 0) {
+            /* most common case */
+            FLUSH_BITS(n);
+#ifdef STATS
+            st_bit_counts[st_current_index] += n;
+#endif
+            break;
+        } else if (n == 0) {
+            return -1;
+        } else {
+            FLUSH_BITS(nb_bits);
+#ifdef STATS
+            st_bit_counts[st_current_index] += nb_bits;
+#endif
+            nb_bits = -n;
+            table_codes = vlc->table_codes + code;
+            table_bits = vlc->table_bits + code;
+        }
+    }
+    RESTORE_BITS(s);
+    return code;
+}
diff --git a/libavcodec/common.h b/libavcodec/common.h
new file mode 100644
index 0000000000..bba7d1fb9d
--- /dev/null
+++ b/libavcodec/common.h
@@ -0,0 +1,170 @@
+#ifndef COMMON_H
+#define COMMON_H
+
+#include "../config.h"
+
+#ifndef USE_LIBAVCODEC
+// workaround for typedef conflict in MPlayer
+typedef unsigned short UINT16;
+typedef signed short INT16;
+#endif
+
+typedef unsigned char UINT8;
+typedef unsigned int UINT32;
+typedef unsigned long long UINT64;
+typedef signed char INT8;
+typedef signed int INT32;
+typedef signed long long INT64;
+
+/* bit output */
+
+struct PutBitContext;
+
+typedef void (*WriteDataFunc)(void *, UINT8 *, int);
+
+typedef struct PutBitContext {
+    UINT8 *buf, *buf_ptr, *buf_end;
+    int bit_cnt;
+    UINT32 bit_buf;
+    long long data_out_size; /* in bytes */
+    void *opaque;
+    WriteDataFunc write_data;
+} PutBitContext;
+
+void init_put_bits(PutBitContext *s, 
+                   UINT8 *buffer, int buffer_size,
+                   void *opaque,
+                   void (*write_data)(void *, UINT8 *, int));
+void put_bits(PutBitContext *s, int n, unsigned int value);
+long long get_bit_count(PutBitContext *s);
+void align_put_bits(PutBitContext *s);
+void flush_put_bits(PutBitContext *s);
+
+/* jpeg specific put_bits */
+void jput_bits(PutBitContext *s, int n, unsigned int value);
+void jflush_put_bits(PutBitContext *s);
+
+/* bit input */
+
+typedef struct GetBitContext {
+    UINT8 *buf, *buf_ptr, *buf_end;
+    int bit_cnt;
+    UINT32 bit_buf;
+} GetBitContext;
+
+typedef struct VLC {
+    int bits;
+    INT16 *table_codes;
+    INT8 *table_bits;
+    int table_size, table_allocated;
+} VLC;
+
+void init_get_bits(GetBitContext *s, 
+                   UINT8 *buffer, int buffer_size);
+
+unsigned int get_bits(GetBitContext *s, int n);
+void align_get_bits(GetBitContext *s);
+int init_vlc(VLC *vlc, int nb_bits, int nb_codes,
+             const void *bits, int bits_wrap, int bits_size,
+             const void *codes, int codes_wrap, int codes_size);
+void free_vlc(VLC *vlc);
+int get_vlc(GetBitContext *s, VLC *vlc);
+
+/* macro to go faster */
+/* n must be <= 24 */
+/* XXX: optimize buffer end test */
+#define SHOW_BITS(s, val, n)\
+{\
+    if (bit_cnt < n && buf_ptr < (s)->buf_end) {\
+        bit_buf |= *buf_ptr++ << (24 - bit_cnt);\
+        bit_cnt += 8;\
+        if (bit_cnt < n && buf_ptr < (s)->buf_end) {\
+            bit_buf |= *buf_ptr++ << (24 - bit_cnt);\
+            bit_cnt += 8;\
+            if (bit_cnt < n && buf_ptr < (s)->buf_end) {\
+                bit_buf |= *buf_ptr++ << (24 - bit_cnt);\
+                bit_cnt += 8;\
+            }\
+        }\
+    }\
+    val = bit_buf >> (32 - n);\
+}
+
+/* SHOW_BITS with n1 >= n must be been done before */
+#define FLUSH_BITS(n)\
+{\
+    bit_buf <<= n;\
+    bit_cnt -= n;\
+}
+
+#define SAVE_BITS(s) \
+{\
+    bit_cnt = (s)->bit_cnt;\
+    bit_buf = (s)->bit_buf;\
+    buf_ptr = (s)->buf_ptr;\
+}
+
+#define RESTORE_BITS(s) \
+{\
+    (s)->buf_ptr = buf_ptr;\
+    (s)->bit_buf = bit_buf;\
+    (s)->bit_cnt = bit_cnt;\
+}
+
+/* define it to include statistics code (useful only for optimizing
+   codec efficiency */
+//#define STATS
+
+#ifdef STATS
+
+enum {
+    ST_UNKNOWN,
+    ST_DC,
+    ST_INTRA_AC,
+    ST_INTER_AC,
+    ST_INTRA_MB,
+    ST_INTER_MB,
+    ST_MV,
+    ST_NB,
+};
+
+extern int st_current_index;
+extern unsigned int st_bit_counts[ST_NB];
+extern unsigned int st_out_bit_counts[ST_NB];
+
+void print_stats(void);
+#endif
+
+/* misc math functions */
+
+extern inline int log2(unsigned int v)
+{
+    int n;
+
+    n = 0;
+    if (v & 0xffff0000) {
+        v >>= 16;
+        n += 16;
+    }
+    if (v & 0xff00) {
+        v >>= 8;
+        n += 8;
+    }
+    if (v & 0xf0) {
+        v >>= 4;
+        n += 4;
+    }
+    if (v & 0xc) {
+        v >>= 2;
+        n += 2;
+    }
+    if (v & 0x2) {
+        n++;
+    }
+    return n;
+}
+
+/* memory */
+void *av_mallocz(int size);
+
+#endif
diff --git a/libavcodec/dct-test.c b/libavcodec/dct-test.c
new file mode 100644
index 0000000000..1da23a1d8e
--- /dev/null
+++ b/libavcodec/dct-test.c
@@ -0,0 +1,105 @@
+/* DCT test. (c) 2001 Gerard Lantau. 
+   Started from sample code by Juan J. Sierralta P.
+*/
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <sys/time.h>
+#include <unistd.h>
+
+#include "dsputil.h"
+
+extern void fdct(DCTELEM *block);
+extern void init_fdct();
+
+#define AANSCALE_BITS 12
+static const unsigned short aanscales[64] = {
+    /* precomputed values scaled up by 14 bits */
+    16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+    22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
+    21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
+    19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
+    16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+    12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
+    8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
+    4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
+};
+
+INT64 gettime(void)
+{
+    struct timeval tv;
+    gettimeofday(&tv,NULL);
+    return (INT64)tv.tv_sec * 1000000 + tv.tv_usec;
+}
+
+#define NB_ITS 20000
+#define NB_ITS_SPEED 50000
+
+void dct_error(const char *name,
+               void (*fdct_func)(DCTELEM *block))
+{
+    int it, i, scale;
+    DCTELEM block[64], block1[64];
+    int err_inf, v;
+    INT64 err2, ti, ti1, it1;
+
+    srandom(0);
+
+    err_inf = 0;
+    err2 = 0;
+    for(it=0;it<NB_ITS;it++) {
+        for(i=0;i<64;i++) 
+            block1[i] = random() % 256;
+        memcpy(block, block1, sizeof(DCTELEM) * 64);
+        
+        fdct_func(block);
+        if (fdct_func == jpeg_fdct_ifast) {
+            for(i=0; i<64; i++) {
+                scale = (1 << (AANSCALE_BITS + 11)) / aanscales[i];
+                block[i] = (block[i] * scale) >> AANSCALE_BITS;
+            }
+        }
+
+        fdct(block1);
+
+        for(i=0;i<64;i++) {
+            v = abs(block[i] - block1[i]);
+            if (v > err_inf)
+                err_inf = v;
+            err2 += v * v;
+        }
+    }
+    printf("DCT %s: err_inf=%d err2=%0.2f\n", 
+           name, err_inf, (double)err2 / NB_ITS / 64.0);
+
+    /* speed test */
+    for(i=0;i<64;i++) 
+        block1[i] = 255 - 63 + i;
+
+    ti = gettime();
+    it1 = 0;
+    do {
+        for(it=0;it<NB_ITS_SPEED;it++) {
+            memcpy(block, block1, sizeof(DCTELEM) * 64);
+            fdct_func(block);
+        }
+        it1 += NB_ITS_SPEED;
+        ti1 = gettime() - ti;
+    } while (ti1 < 1000000);
+
+    printf("DCT %s: %0.1f kdct/s\n", 
+           name, (double)it1 * 1000.0 / (double)ti1);
+}
+
+int main(int argc, char **argv)
+{
+    init_fdct();
+    
+    printf("ffmpeg DCT test\n");
+
+    dct_error("REF", fdct); /* only to verify code ! */
+    dct_error("AAN", jpeg_fdct_ifast);
+    dct_error("MMX", fdct_mmx);
+    return 0;
+}
+	
diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c
new file mode 100644
index 0000000000..7e544a4524
--- /dev/null
+++ b/libavcodec/dsputil.c
@@ -0,0 +1,383 @@
+/*
+ * DSP utils
+ * Copyright (c) 2000, 2001 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include "avcodec.h"
+#include "dsputil.h"
+
+#ifdef CONFIG_MMX
+int mm_flags; /* multimedia extension flags */
+#endif
+
+void (*get_pixels)(DCTELEM *block, const UINT8 *pixels, int line_size);
+void (*put_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size);
+void (*add_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size);
+
+op_pixels_abs_func pix_abs16x16;
+op_pixels_abs_func pix_abs16x16_x2;
+op_pixels_abs_func pix_abs16x16_y2;
+op_pixels_abs_func pix_abs16x16_xy2;
+
+static UINT8 cropTbl[256 + 2 * MAX_NEG_CROP];
+UINT32 squareTbl[512];
+
+void get_pixels_c(DCTELEM *block, const UINT8 *pixels, int line_size)
+{
+    DCTELEM *p;
+    const UINT8 *pix;
+    int i;
+
+    /* read the pixels */
+    p = block;
+    pix = pixels;
+    for(i=0;i<8;i++) {
+        p[0] = pix[0];
+        p[1] = pix[1];
+        p[2] = pix[2];
+        p[3] = pix[3];
+        p[4] = pix[4];
+        p[5] = pix[5];
+        p[6] = pix[6];
+        p[7] = pix[7];
+        pix += line_size;
+        p += 8;
+    }
+}
+
+void put_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size)
+{
+    const DCTELEM *p;
+    UINT8 *pix;
+    int i;
+    UINT8 *cm = cropTbl + MAX_NEG_CROP;
+    
+    /* read the pixels */
+    p = block;
+    pix = pixels;
+    for(i=0;i<8;i++) {
+        pix[0] = cm[p[0]];
+        pix[1] = cm[p[1]];
+        pix[2] = cm[p[2]];
+        pix[3] = cm[p[3]];
+        pix[4] = cm[p[4]];
+        pix[5] = cm[p[5]];
+        pix[6] = cm[p[6]];
+        pix[7] = cm[p[7]];
+        pix += line_size;
+        p += 8;
+    }
+}
+
+void add_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size)
+{
+    const DCTELEM *p;
+    UINT8 *pix;
+    int i;
+    UINT8 *cm = cropTbl + MAX_NEG_CROP;
+    
+    /* read the pixels */
+    p = block;
+    pix = pixels;
+    for(i=0;i<8;i++) {
+        pix[0] = cm[pix[0] + p[0]];
+        pix[1] = cm[pix[1] + p[1]];
+        pix[2] = cm[pix[2] + p[2]];
+        pix[3] = cm[pix[3] + p[3]];
+        pix[4] = cm[pix[4] + p[4]];
+        pix[5] = cm[pix[5] + p[5]];
+        pix[6] = cm[pix[6] + p[6]];
+        pix[7] = cm[pix[7] + p[7]];
+        pix += line_size;
+        p += 8;
+    }
+}
+
+#define PIXOP(BTYPE, OPNAME, OP, INCR)                                                   \
+                                                                                         \
+static void OPNAME ## _pixels(BTYPE *block, const UINT8 *pixels, int line_size, int h)    \
+{                                                                                        \
+    BTYPE *p;                                                                            \
+    const UINT8 *pix;                                                                    \
+                                                                                         \
+    p = block;                                                                           \
+    pix = pixels;                                                                        \
+    do {                                                                                 \
+        OP(p[0], pix[0]);                                                                  \
+        OP(p[1], pix[1]);                                                                  \
+        OP(p[2], pix[2]);                                                                  \
+        OP(p[3], pix[3]);                                                                  \
+        OP(p[4], pix[4]);                                                                  \
+        OP(p[5], pix[5]);                                                                  \
+        OP(p[6], pix[6]);                                                                  \
+        OP(p[7], pix[7]);                                                                  \
+        pix += line_size;                                                                \
+        p += INCR;                                                                       \
+    } while (--h);;                                                                       \
+}                                                                                        \
+                                                                                         \
+static void OPNAME ## _pixels_x2(BTYPE *block, const UINT8 *pixels, int line_size, int h)     \
+{                                                                                        \
+    BTYPE *p;                                                                          \
+    const UINT8 *pix;                                                                    \
+                                                                                         \
+    p = block;                                                                           \
+    pix = pixels;                                                                        \
+    do {                                                                   \
+        OP(p[0], avg2(pix[0], pix[1]));                                                    \
+        OP(p[1], avg2(pix[1], pix[2]));                                                    \
+        OP(p[2], avg2(pix[2], pix[3]));                                                    \
+        OP(p[3], avg2(pix[3], pix[4]));                                                    \
+        OP(p[4], avg2(pix[4], pix[5]));                                                    \
+        OP(p[5], avg2(pix[5], pix[6]));                                                    \
+        OP(p[6], avg2(pix[6], pix[7]));                                                    \
+        OP(p[7], avg2(pix[7], pix[8]));                                                    \
+        pix += line_size;                                                                \
+        p += INCR;                                                                       \
+    } while (--h);                                                                        \
+}                                                                                        \
+                                                                                         \
+static void OPNAME ## _pixels_y2(BTYPE *block, const UINT8 *pixels, int line_size, int h)     \
+{                                                                                        \
+    BTYPE *p;                                                                          \
+    const UINT8 *pix;                                                                    \
+    const UINT8 *pix1;                                                                   \
+                                                                                         \
+    p = block;                                                                           \
+    pix = pixels;                                                                        \
+    pix1 = pixels + line_size;                                                           \
+    do {                                                                                 \
+        OP(p[0], avg2(pix[0], pix1[0]));                                                   \
+        OP(p[1], avg2(pix[1], pix1[1]));                                                   \
+        OP(p[2], avg2(pix[2], pix1[2]));                                                   \
+        OP(p[3], avg2(pix[3], pix1[3]));                                                   \
+        OP(p[4], avg2(pix[4], pix1[4]));                                                   \
+        OP(p[5], avg2(pix[5], pix1[5]));                                                   \
+        OP(p[6], avg2(pix[6], pix1[6]));                                                   \
+        OP(p[7], avg2(pix[7], pix1[7]));                                                   \
+        pix += line_size;                                                                \
+        pix1 += line_size;                                                               \
+        p += INCR;                                                                       \
+    } while(--h);                                                                         \
+}                                                                                        \
+                                                                                         \
+static void OPNAME ## _pixels_xy2(BTYPE *block, const UINT8 *pixels, int line_size, int h)    \
+{                                                                                        \
+    BTYPE *p;                                                                          \
+    const UINT8 *pix;                                                                    \
+    const UINT8 *pix1;                                                                   \
+                                                                                         \
+    p = block;                                                                           \
+    pix = pixels;                                                                        \
+    pix1 = pixels + line_size;                                                           \
+    do {                                                                   \
+        OP(p[0], avg4(pix[0], pix[1], pix1[0], pix1[1]));                                  \
+        OP(p[1], avg4(pix[1], pix[2], pix1[1], pix1[2]));                                  \
+        OP(p[2], avg4(pix[2], pix[3], pix1[2], pix1[3]));                                  \
+        OP(p[3], avg4(pix[3], pix[4], pix1[3], pix1[4]));                                  \
+        OP(p[4], avg4(pix[4], pix[5], pix1[4], pix1[5]));                                  \
+        OP(p[5], avg4(pix[5], pix[6], pix1[5], pix1[6]));                                  \
+        OP(p[6], avg4(pix[6], pix[7], pix1[6], pix1[7]));                                  \
+        OP(p[7], avg4(pix[7], pix[8], pix1[7], pix1[8]));                                  \
+        pix += line_size;                                                                \
+        pix1 += line_size;                                                               \
+        p += INCR;                                                                       \
+    } while(--h);                                                                         \
+}                                                                                        \
+                                                                                         \
+void (*OPNAME ## _pixels_tab[4])(BTYPE *block, const UINT8 *pixels, int line_size, int h) = { \
+    OPNAME ## _pixels,                                                                   \
+    OPNAME ## _pixels_x2,                                                                \
+    OPNAME ## _pixels_y2,                                                                \
+    OPNAME ## _pixels_xy2,                                                               \
+};
+
+
+/* rounding primitives */
+#define avg2(a,b) ((a+b+1)>>1)
+#define avg4(a,b,c,d) ((a+b+c+d+2)>>2)
+
+#define op_put(a, b) a = b
+#define op_avg(a, b) a = avg2(a, b)
+#define op_sub(a, b) a -= b
+
+PIXOP(UINT8, put, op_put, line_size)
+PIXOP(UINT8, avg, op_avg, line_size)
+
+PIXOP(DCTELEM, sub, op_sub, 8)
+
+/* not rounding primitives */
+#undef avg2
+#undef avg4
+#define avg2(a,b) ((a+b)>>1)
+#define avg4(a,b,c,d) ((a+b+c+d+1)>>2)
+
+PIXOP(UINT8, put_no_rnd, op_put, line_size)
+PIXOP(UINT8, avg_no_rnd, op_avg, line_size)
+
+/* motion estimation */
+
+#undef avg2
+#undef avg4
+#define avg2(a,b) ((a+b+1)>>1)
+#define avg4(a,b,c,d) ((a+b+c+d+2)>>2)
+
+int pix_abs16x16_c(UINT8 *pix1, UINT8 *pix2, int line_size, int h)
+{
+    int s, i;
+
+    s = 0;
+    for(i=0;i<h;i++) {
+        s += abs(pix1[0] - pix2[0]);
+        s += abs(pix1[1] - pix2[1]);
+        s += abs(pix1[2] - pix2[2]);
+        s += abs(pix1[3] - pix2[3]);
+        s += abs(pix1[4] - pix2[4]);
+        s += abs(pix1[5] - pix2[5]);
+        s += abs(pix1[6] - pix2[6]);
+        s += abs(pix1[7] - pix2[7]);
+        s += abs(pix1[8] - pix2[8]);
+        s += abs(pix1[9] - pix2[9]);
+        s += abs(pix1[10] - pix2[10]);
+        s += abs(pix1[11] - pix2[11]);
+        s += abs(pix1[12] - pix2[12]);
+        s += abs(pix1[13] - pix2[13]);
+        s += abs(pix1[14] - pix2[14]);
+        s += abs(pix1[15] - pix2[15]);
+        pix1 += line_size;
+        pix2 += line_size;
+    }
+    return s;
+}
+
+int pix_abs16x16_x2_c(UINT8 *pix1, UINT8 *pix2, int line_size, int h)
+{
+    int s, i;
+
+    s = 0;
+    for(i=0;i<h;i++) {
+        s += abs(pix1[0] - avg2(pix2[0], pix2[1]));
+        s += abs(pix1[1] - avg2(pix2[1], pix2[2]));
+        s += abs(pix1[2] - avg2(pix2[2], pix2[3]));
+        s += abs(pix1[3] - avg2(pix2[3], pix2[4]));
+        s += abs(pix1[4] - avg2(pix2[4], pix2[5]));
+        s += abs(pix1[5] - avg2(pix2[5], pix2[6]));
+        s += abs(pix1[6] - avg2(pix2[6], pix2[7]));
+        s += abs(pix1[7] - avg2(pix2[7], pix2[8]));
+        s += abs(pix1[8] - avg2(pix2[8], pix2[9]));
+        s += abs(pix1[9] - avg2(pix2[9], pix2[10]));
+        s += abs(pix1[10] - avg2(pix2[10], pix2[11]));
+        s += abs(pix1[11] - avg2(pix2[11], pix2[12]));
+        s += abs(pix1[12] - avg2(pix2[12], pix2[13]));
+        s += abs(pix1[13] - avg2(pix2[13], pix2[14]));
+        s += abs(pix1[14] - avg2(pix2[14], pix2[15]));
+        s += abs(pix1[15] - avg2(pix2[15], pix2[16]));
+        pix1 += line_size;
+        pix2 += line_size;
+    }
+    return s;
+}
+
+int pix_abs16x16_y2_c(UINT8 *pix1, UINT8 *pix2, int line_size, int h)
+{
+    int s, i;
+    UINT8 *pix3 = pix2 + line_size;
+
+    s = 0;
+    for(i=0;i<h;i++) {
+        s += abs(pix1[0] - avg2(pix2[0], pix3[0]));
+        s += abs(pix1[1] - avg2(pix2[1], pix3[1]));
+        s += abs(pix1[2] - avg2(pix2[2], pix3[2]));
+        s += abs(pix1[3] - avg2(pix2[3], pix3[3]));
+        s += abs(pix1[4] - avg2(pix2[4], pix3[4]));
+        s += abs(pix1[5] - avg2(pix2[5], pix3[5]));
+        s += abs(pix1[6] - avg2(pix2[6], pix3[6]));
+        s += abs(pix1[7] - avg2(pix2[7], pix3[7]));
+        s += abs(pix1[8] - avg2(pix2[8], pix3[8]));
+        s += abs(pix1[9] - avg2(pix2[9], pix3[9]));
+        s += abs(pix1[10] - avg2(pix2[10], pix3[10]));
+        s += abs(pix1[11] - avg2(pix2[11], pix3[11]));
+        s += abs(pix1[12] - avg2(pix2[12], pix3[12]));
+        s += abs(pix1[13] - avg2(pix2[13], pix3[13]));
+        s += abs(pix1[14] - avg2(pix2[14], pix3[14]));
+        s += abs(pix1[15] - avg2(pix2[15], pix3[15]));
+        pix1 += line_size;
+        pix2 += line_size;
+        pix3 += line_size;
+    }
+    return s;
+}
+
+int pix_abs16x16_xy2_c(UINT8 *pix1, UINT8 *pix2, int line_size, int h)
+{
+    int s, i;
+    UINT8 *pix3 = pix2 + line_size;
+
+    s = 0;
+    for(i=0;i<h;i++) {
+        s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1]));
+        s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2]));
+        s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3]));
+        s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4]));
+        s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5]));
+        s += abs(pix1[5] - avg4(pix2[5], pix2[6], pix3[5], pix3[6]));
+        s += abs(pix1[6] - avg4(pix2[6], pix2[7], pix3[6], pix3[7]));
+        s += abs(pix1[7] - avg4(pix2[7], pix2[8], pix3[7], pix3[8]));
+        s += abs(pix1[8] - avg4(pix2[8], pix2[9], pix3[8], pix3[9]));
+        s += abs(pix1[9] - avg4(pix2[9], pix2[10], pix3[9], pix3[10]));
+        s += abs(pix1[10] - avg4(pix2[10], pix2[11], pix3[10], pix3[11]));
+        s += abs(pix1[11] - avg4(pix2[11], pix2[12], pix3[11], pix3[12]));
+        s += abs(pix1[12] - avg4(pix2[12], pix2[13], pix3[12], pix3[13]));
+        s += abs(pix1[13] - avg4(pix2[13], pix2[14], pix3[13], pix3[14]));
+        s += abs(pix1[14] - avg4(pix2[14], pix2[15], pix3[14], pix3[15]));
+        s += abs(pix1[15] - avg4(pix2[15], pix2[16], pix3[15], pix3[16]));
+        pix1 += line_size;
+        pix2 += line_size;
+        pix3 += line_size;
+    }
+    return s;
+}
+
+void dsputil_init(void)
+{
+    int i;
+
+    for(i=0;i<256;i++) cropTbl[i + MAX_NEG_CROP] = i;
+    for(i=0;i<MAX_NEG_CROP;i++) {
+        cropTbl[i] = 0;
+        cropTbl[i + MAX_NEG_CROP + 256] = 255;
+    }
+
+    for(i=0;i<512;i++) {
+        squareTbl[i] = (i - 256) * (i - 256);
+    }
+
+    get_pixels = get_pixels_c;
+    put_pixels_clamped = put_pixels_clamped_c;
+    add_pixels_clamped = add_pixels_clamped_c;
+
+    pix_abs16x16 = pix_abs16x16_c;
+    pix_abs16x16_x2 = pix_abs16x16_x2_c;
+    pix_abs16x16_y2 = pix_abs16x16_y2_c;
+    pix_abs16x16_xy2 = pix_abs16x16_xy2_c;
+    av_fdct = jpeg_fdct_ifast;
+
+#ifdef CONFIG_MMX
+    dsputil_init_mmx();
+#endif
+}
diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h
new file mode 100644
index 0000000000..717653e813
--- /dev/null
+++ b/libavcodec/dsputil.h
@@ -0,0 +1,91 @@
+#ifndef DSPUTIL_H
+#define DSPUTIL_H
+
+#include "common.h"
+#include <inttypes.h>
+
+/* dct code */
+typedef short DCTELEM;
+
+void jpeg_fdct_ifast (DCTELEM *data);
+
+void j_rev_dct (DCTELEM *data);
+
+void fdct_mmx(DCTELEM *block);
+
+void (*av_fdct)(DCTELEM *block);
+
+/* pixel operations */
+#define MAX_NEG_CROP 384
+
+/* temporary */
+extern UINT32 squareTbl[512];
+
+void dsputil_init(void);
+
+/* pixel ops : interface with DCT */
+
+extern void (*get_pixels)(DCTELEM *block, const UINT8 *pixels, int line_size);
+extern void (*put_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size);
+extern void (*add_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size);
+
+void get_pixels_c(DCTELEM *block, const UINT8 *pixels, int line_size);
+void put_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size);
+void add_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size);
+
+/* add and put pixel (decoding) */
+typedef void (*op_pixels_func)(UINT8 *block, const UINT8 *pixels, int line_size, int h);
+
+extern op_pixels_func put_pixels_tab[4];
+extern op_pixels_func avg_pixels_tab[4];
+extern op_pixels_func put_no_rnd_pixels_tab[4];
+extern op_pixels_func avg_no_rnd_pixels_tab[4];
+
+/* sub pixel (encoding) */
+extern void (*sub_pixels_tab[4])(DCTELEM *block, const UINT8 *pixels, int line_size, int h);
+
+#define sub_pixels_2(block, pixels, line_size, dxy) \
+   sub_pixels_tab[dxy](block, pixels, line_size, 8)
+
+/* motion estimation */
+
+typedef int (*op_pixels_abs_func)(UINT8 *blk1, UINT8 *blk2, int line_size, int h);
+
+extern op_pixels_abs_func pix_abs16x16;
+extern op_pixels_abs_func pix_abs16x16_x2;
+extern op_pixels_abs_func pix_abs16x16_y2;
+extern op_pixels_abs_func pix_abs16x16_xy2;
+
+int pix_abs16x16_c(UINT8 *blk1, UINT8 *blk2, int lx, int h);
+int pix_abs16x16_x2_c(UINT8 *blk1, UINT8 *blk2, int lx, int h);
+int pix_abs16x16_y2_c(UINT8 *blk1, UINT8 *blk2, int lx, int h);
+int pix_abs16x16_xy2_c(UINT8 *blk1, UINT8 *blk2, int lx, int h);
+
+#ifdef CONFIG_MMX
+
+#define MM_MMX    0x0001 /* standard MMX */
+#define MM_3DNOW  0x0004 /* AMD 3DNOW */
+#define MM_MMXEXT 0x0002 /* SSE integer functions or AMD MMX ext */
+#define MM_SSE    0x0008 /* SSE functions */
+#define MM_SSE2   0x0010 /* PIV SSE2 functions */
+
+extern int mm_flags;
+
+int mm_support(void);
+
+static inline void emms(void)
+{
+	asm volatile ("emms;");
+}
+
+#define __align8 __attribute__ ((aligned (8)))
+
+void dsputil_init_mmx(void);
+
+#else
+
+#define __align8
+
+#endif
+
+#endif
diff --git a/libavcodec/fdctref.c b/libavcodec/fdctref.c
new file mode 100644
index 0000000000..5038e67538
--- /dev/null
+++ b/libavcodec/fdctref.c
@@ -0,0 +1,120 @@
+/* fdctref.c, forward discrete cosine transform, double precision           */
+
+/* Copyright (C) 1996, MPEG Software Simulation Group. All Rights Reserved. */
+
+/*
+ * Disclaimer of Warranty
+ *
+ * These software programs are available to the user without any license fee or
+ * royalty on an "as is" basis.  The MPEG Software Simulation Group disclaims
+ * any and all warranties, whether express, implied, or statuary, including any
+ * implied warranties or merchantability or of fitness for a particular
+ * purpose.  In no event shall the copyright-holder be liable for any
+ * incidental, punitive, or consequential damages of any kind whatsoever
+ * arising from the use of these programs.
+ *
+ * This disclaimer of warranty extends to the user of these programs and user's
+ * customers, employees, agents, transferees, successors, and assigns.
+ *
+ * The MPEG Software Simulation Group does not represent or warrant that the
+ * programs furnished hereunder are free of infringement of any third-party
+ * patents.
+ *
+ * Commercial implementations of MPEG-1 and MPEG-2 video, including shareware,
+ * are subject to royalty fees to patent holders.  Many of these patents are
+ * general enough such that they are unavoidable regardless of implementation
+ * design.
+ *
+ */
+
+#include <math.h>
+
+// #include "config.h"
+
+#ifndef PI
+# ifdef M_PI
+#  define PI M_PI
+# else
+#  define PI 3.14159265358979323846
+# endif
+#endif
+
+/* global declarations */
+void init_fdct (void);
+void fdct (short *block);
+
+/* private data */
+static double c[8][8]; /* transform coefficients */
+
+void init_fdct()
+{
+  int i, j;
+  double s;
+
+  for (i=0; i<8; i++)
+  {
+    s = (i==0) ? sqrt(0.125) : 0.5;
+
+    for (j=0; j<8; j++)
+      c[i][j] = s * cos((PI/8.0)*i*(j+0.5));
+  }
+}
+
+void fdct(block)
+short *block;
+{
+	register int i, j;
+	double s;
+	double tmp[64];
+
+	for(i = 0; i < 8; i++)
+    	for(j = 0; j < 8; j++)
+    	{
+    		s = 0.0;
+
+/*
+ *     		for(k = 0; k < 8; k++)
+ *         		s += c[j][k] * block[8 * i + k];
+ */
+        	s += c[j][0] * block[8 * i + 0];
+        	s += c[j][1] * block[8 * i + 1];
+        	s += c[j][2] * block[8 * i + 2];
+        	s += c[j][3] * block[8 * i + 3];
+        	s += c[j][4] * block[8 * i + 4];
+        	s += c[j][5] * block[8 * i + 5];
+        	s += c[j][6] * block[8 * i + 6];
+        	s += c[j][7] * block[8 * i + 7];
+
+    		tmp[8 * i + j] = s;
+    	}
+
+	for(j = 0; j < 8; j++)
+    	for(i = 0; i < 8; i++)
+    	{
+    		s = 0.0;
+
+/*
+ *     	  	for(k = 0; k < 8; k++)
+ *        	    s += c[i][k] * tmp[8 * k + j];
+ */
+        	s += c[i][0] * tmp[8 * 0 + j];
+        	s += c[i][1] * tmp[8 * 1 + j];
+        	s += c[i][2] * tmp[8 * 2 + j];
+        	s += c[i][3] * tmp[8 * 3 + j];
+        	s += c[i][4] * tmp[8 * 4 + j];
+        	s += c[i][5] * tmp[8 * 5 + j];
+        	s += c[i][6] * tmp[8 * 6 + j];
+        	s += c[i][7] * tmp[8 * 7 + j];
+
+    		block[8 * i + j] = (short)floor(s + 0.499999);
+/*
+ * reason for adding 0.499999 instead of 0.5:
+ * s is quite often x.5 (at least for i and/or j = 0 or 4)
+ * and setting the rounding threshold exactly to 0.5 leads to an
+ * extremely high arithmetic implementation dependency of the result;
+ * s being between x.5 and x.500001 (which is now incorrectly rounded
+ * downwards instead of upwards) is assumed to occur less often
+ * (if at all)
+ */
+      }
+}
diff --git a/libavcodec/h263.c b/libavcodec/h263.c
new file mode 100644
index 0000000000..35ecb15c2e
--- /dev/null
+++ b/libavcodec/h263.c
@@ -0,0 +1,1291 @@
+/*
+ * H263/MPEG4 backend for ffmpeg encoder and decoder
+ * Copyright (c) 2000,2001 Gerard Lantau.
+ * H263+ support for custom picture format.
+ * Copyright (c) 2001 Juan J. Sierralta P.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include "common.h"
+#include "dsputil.h"
+#include "avcodec.h"
+#include "mpegvideo.h"
+#include "h263data.h"
+#include "mpeg4data.h"
+
+#define NDEBUG
+#include <assert.h>
+
+static void h263_encode_block(MpegEncContext * s, DCTELEM * block,
+			      int n);
+static void h263_encode_motion(MpegEncContext * s, int val);
+static void mpeg4_encode_block(MpegEncContext * s, DCTELEM * block,
+			       int n);
+static int h263_decode_motion(MpegEncContext * s, int pred);
+static int h263_decode_block(MpegEncContext * s, DCTELEM * block,
+                             int n, int coded);
+static int mpeg4_decode_block(MpegEncContext * s, DCTELEM * block,
+                              int n, int coded);
+
+int h263_get_picture_format(int width, int height)
+{
+    int format;
+
+    if (width == 128 && height == 96)
+	format = 1;
+    else if (width == 176 && height == 144)
+	format = 2;
+    else if (width == 352 && height == 288)
+	format = 3;
+    else if (width == 704 && height == 576)
+	format = 4;
+    else if (width == 1408 && height == 1152)
+	format = 5;
+    else
+        format = 7;
+    return format;
+}
+
+void h263_encode_picture_header(MpegEncContext * s, int picture_number)
+{
+    int format, umvplus;
+
+    align_put_bits(&s->pb);
+    put_bits(&s->pb, 22, 0x20);
+    put_bits(&s->pb, 8, ((s->picture_number * 30 * FRAME_RATE_BASE) / 
+                         s->frame_rate) & 0xff);
+
+    put_bits(&s->pb, 1, 1);	/* marker */
+    put_bits(&s->pb, 1, 0);	/* h263 id */
+    put_bits(&s->pb, 1, 0);	/* split screen off */
+    put_bits(&s->pb, 1, 0);	/* camera  off */
+    put_bits(&s->pb, 1, 0);	/* freeze picture release off */
+
+    if (!s->h263_plus) {
+        /* H.263v1 */
+        format = h263_get_picture_format(s->width, s->height);
+        put_bits(&s->pb, 3, format);
+        put_bits(&s->pb, 1, (s->pict_type == P_TYPE));
+        /* By now UMV IS DISABLED ON H.263v1, since the restrictions
+        of H.263v1 UMV implies to check the predicted MV after
+        calculation of the current MB to see if we're on the limits */
+        put_bits(&s->pb, 1, 0);	/* unrestricted motion vector: off */
+        put_bits(&s->pb, 1, 0);	/* SAC: off */
+        put_bits(&s->pb, 1, 0);	/* advanced prediction mode: off */
+        put_bits(&s->pb, 1, 0);	/* not PB frame */
+        put_bits(&s->pb, 5, s->qscale);
+        put_bits(&s->pb, 1, 0);	/* Continuous Presence Multipoint mode: off */
+    } else {
+        /* H.263v2 */
+        /* H.263 Plus PTYPE */
+        put_bits(&s->pb, 3, 7);
+        put_bits(&s->pb,3,1); /* Update Full Extended PTYPE */
+        put_bits(&s->pb,3,6); /* Custom Source Format */
+        put_bits(&s->pb,1,0); /* Custom PCF: off */
+        umvplus = (s->pict_type == P_TYPE) && s->unrestricted_mv;
+        put_bits(&s->pb, 1, umvplus); /* Unrestricted Motion Vector */
+        put_bits(&s->pb,1,0); /* SAC: off */
+        put_bits(&s->pb,1,0); /* Advanced Prediction Mode: off */
+        put_bits(&s->pb,1,0); /* Advanced Intra Coding: off */
+        put_bits(&s->pb,1,0); /* Deblocking Filter: off */
+        put_bits(&s->pb,1,0); /* Slice Structured: off */
+        put_bits(&s->pb,1,0); /* Reference Picture Selection: off */
+        put_bits(&s->pb,1,0); /* Independent Segment Decoding: off */
+        put_bits(&s->pb,1,0); /* Alternative Inter VLC: off */
+        put_bits(&s->pb,1,0); /* Modified Quantization: off */
+        put_bits(&s->pb,1,1); /* "1" to prevent start code emulation */
+        put_bits(&s->pb,3,0); /* Reserved */
+		
+        put_bits(&s->pb, 3, s->pict_type == P_TYPE);
+		
+        put_bits(&s->pb,1,0); /* Reference Picture Resampling: off */
+        put_bits(&s->pb,1,0); /* Reduced-Resolution Update: off */
+        put_bits(&s->pb,1,0); /* Rounding Type */
+        put_bits(&s->pb,2,0); /* Reserved */
+        put_bits(&s->pb,1,1); /* "1" to prevent start code emulation */
+		
+        /* This should be here if PLUSPTYPE */
+        put_bits(&s->pb, 1, 0);	/* Continuous Presence Multipoint mode: off */
+		
+        /* Custom Picture Format (CPFMT) */
+		
+        put_bits(&s->pb,4,2); /* Aspect ratio: CIF 12:11 (4:3) picture */
+        put_bits(&s->pb,9,(s->width >> 2) - 1);
+        put_bits(&s->pb,1,1); /* "1" to prevent start code emulation */
+        put_bits(&s->pb,9,(s->height >> 2));
+        /* Unlimited Unrestricted Motion Vectors Indicator (UUI) */
+        if (umvplus)
+            put_bits(&s->pb,1,1); /* Limited according tables of Annex D */
+        put_bits(&s->pb, 5, s->qscale);
+    }
+
+    put_bits(&s->pb, 1, 0);	/* no PEI */
+}
+
+void h263_encode_mb(MpegEncContext * s,
+		    DCTELEM block[6][64],
+		    int motion_x, int motion_y)
+{
+    int cbpc, cbpy, i, cbp, pred_x, pred_y;
+
+    //    printf("**mb x=%d y=%d\n", s->mb_x, s->mb_y);
+    if (!s->mb_intra) {
+	/* compute cbp */
+	cbp = 0;
+	for (i = 0; i < 6; i++) {
+	    if (s->block_last_index[i] >= 0)
+		cbp |= 1 << (5 - i);
+	}
+	if ((cbp | motion_x | motion_y) == 0) {
+	    /* skip macroblock */
+	    put_bits(&s->pb, 1, 1);
+	    return;
+	}
+	put_bits(&s->pb, 1, 0);	/* mb coded */
+	cbpc = cbp & 3;
+	put_bits(&s->pb,
+		 inter_MCBPC_bits[cbpc],
+		 inter_MCBPC_code[cbpc]);
+	cbpy = cbp >> 2;
+	cbpy ^= 0xf;
+	put_bits(&s->pb, cbpy_tab[cbpy][1], cbpy_tab[cbpy][0]);
+
+	/* motion vectors: 16x16 mode only now */
+        h263_pred_motion(s, 0, &pred_x, &pred_y);
+        
+        h263_encode_motion(s, motion_x - pred_x);
+        h263_encode_motion(s, motion_y - pred_y);
+    } else {
+	/* compute cbp */
+	cbp = 0;
+	for (i = 0; i < 6; i++) {
+	    if (s->block_last_index[i] >= 1)
+		cbp |= 1 << (5 - i);
+	}
+
+	cbpc = cbp & 3;
+	if (s->pict_type == I_TYPE) {
+	    put_bits(&s->pb,
+		     intra_MCBPC_bits[cbpc],
+		     intra_MCBPC_code[cbpc]);
+	} else {
+	    put_bits(&s->pb, 1, 0);	/* mb coded */
+	    put_bits(&s->pb,
+		     inter_MCBPC_bits[cbpc + 4],
+		     inter_MCBPC_code[cbpc + 4]);
+	}
+	if (s->h263_pred) {
+	    /* XXX: currently, we do not try to use ac prediction */
+	    put_bits(&s->pb, 1, 0);	/* no ac prediction */
+	}
+	cbpy = cbp >> 2;
+	put_bits(&s->pb, cbpy_tab[cbpy][1], cbpy_tab[cbpy][0]);
+    }
+
+    /* encode each block */
+    if (s->h263_pred) {
+	for (i = 0; i < 6; i++) {
+	    mpeg4_encode_block(s, block[i], i);
+	}
+    } else {
+	for (i = 0; i < 6; i++) {
+	    h263_encode_block(s, block[i], i);
+	}
+    }
+}
+
+static inline int mid_pred(int a, int b, int c)
+{
+    int vmin, vmax;
+    vmin = a;
+    if (b < vmin)
+        vmin = b;
+    if (c < vmin)
+        vmin = c;
+
+    vmax = a;
+    if (b > vmax)
+        vmax = b;
+    if (c > vmax)
+        vmax = c;
+
+    return a + b + c - vmin - vmax;
+}
+
+INT16 *h263_pred_motion(MpegEncContext * s, int block, 
+                        int *px, int *py)
+{
+    int x, y, wrap;
+    INT16 *A, *B, *C, *mot_val;
+
+    x = 2 * s->mb_x + 1 + (block & 1);
+    y = 2 * s->mb_y + 1 + ((block >> 1) & 1);
+    wrap = 2 * s->mb_width + 2;
+
+    mot_val = s->motion_val[(x) + (y) * wrap];
+
+    /* special case for first line */
+    if (y == 1 || s->first_slice_line) {
+        A = s->motion_val[(x-1) + (y) * wrap];
+        *px = A[0];
+        *py = A[1];
+    } else {
+        switch(block) {
+        default:
+        case 0:
+            A = s->motion_val[(x-1) + (y) * wrap];
+            B = s->motion_val[(x) + (y-1) * wrap];
+            C = s->motion_val[(x+2) + (y-1) * wrap];
+            break;
+        case 1:
+        case 2:
+            A = s->motion_val[(x-1) + (y) * wrap];
+            B = s->motion_val[(x) + (y-1) * wrap];
+            C = s->motion_val[(x+1) + (y-1) * wrap];
+            break;
+        case 3:
+            A = s->motion_val[(x-1) + (y) * wrap];
+            B = s->motion_val[(x-1) + (y-1) * wrap];
+            C = s->motion_val[(x) + (y-1) * wrap];
+            break;
+        }
+        *px = mid_pred(A[0], B[0], C[0]);
+        *py = mid_pred(A[1], B[1], C[1]);
+    }
+    return mot_val;
+}
+
+
+static void h263_encode_motion(MpegEncContext * s, int val)
+{
+    int range, l, m, bit_size, sign, code, bits;
+
+    if (val == 0) {
+        /* zero vector */
+        code = 0;
+        put_bits(&s->pb, mvtab[code][1], mvtab[code][0]);
+    } else {
+        bit_size = s->f_code - 1;
+        range = 1 << bit_size;
+        /* modulo encoding */
+        l = range * 32;
+        m = 2 * l;
+        if (val < -l) {
+            val += m;
+        } else if (val >= l) {
+            val -= m;
+        }
+
+        if (val >= 0) {
+            val--;
+            code = (val >> bit_size) + 1;
+            bits = val & (range - 1);
+            sign = 0;
+        } else {
+            val = -val;
+            val--;
+            code = (val >> bit_size) + 1;
+            bits = val & (range - 1);
+            sign = 1;
+        }
+
+        put_bits(&s->pb, mvtab[code][1] + 1, (mvtab[code][0] << 1) | sign); 
+        if (bit_size > 0) {
+            put_bits(&s->pb, bit_size, bits);
+        }
+    }
+}
+
+void h263_encode_init_vlc(MpegEncContext *s)
+{
+    static int done = 0;
+
+    if (!done) {
+        done = 1;
+        init_rl(&rl_inter);
+        init_rl(&rl_intra);
+    }
+}
+
+static void h263_encode_block(MpegEncContext * s, DCTELEM * block, int n)
+{
+    int level, run, last, i, j, last_index, last_non_zero, sign, slevel;
+    int code;
+    RLTable *rl = &rl_inter;
+
+    if (s->mb_intra) {
+	/* DC coef */
+	level = block[0];
+        /* 255 cannot be represented, so we clamp */
+        if (level > 254) {
+            level = 254;
+            block[0] = 254;
+        }
+	if (level == 128)
+	    put_bits(&s->pb, 8, 0xff);
+	else
+	    put_bits(&s->pb, 8, level & 0xff);
+	i = 1;
+    } else {
+	i = 0;
+    }
+
+    /* AC coefs */
+    last_index = s->block_last_index[n];
+    last_non_zero = i - 1;
+    for (; i <= last_index; i++) {
+	j = zigzag_direct[i];
+	level = block[j];
+	if (level) {
+	    run = i - last_non_zero - 1;
+	    last = (i == last_index);
+	    sign = 0;
+	    slevel = level;
+	    if (level < 0) {
+		sign = 1;
+		level = -level;
+	    }
+            code = get_rl_index(rl, last, run, level);
+            put_bits(&s->pb, rl->table_vlc[code][1], rl->table_vlc[code][0]);
+            if (code == rl->n) {
+                put_bits(&s->pb, 1, last);
+                put_bits(&s->pb, 6, run);
+                put_bits(&s->pb, 8, slevel & 0xff);
+            } else {
+                put_bits(&s->pb, 1, sign);
+            }
+	    last_non_zero = i;
+	}
+    }
+}
+
+/***************************************************/
+
+/* write mpeg4 VOP header */
+void mpeg4_encode_picture_header(MpegEncContext * s, int picture_number)
+{
+    align_put_bits(&s->pb);
+
+    put_bits(&s->pb, 32, 0x1B6);	/* vop header */
+    put_bits(&s->pb, 2, s->pict_type - 1);	/* pict type: I = 0 , P = 1 */
+    /* XXX: time base + 1 not always correct */
+    put_bits(&s->pb, 1, 1);
+    put_bits(&s->pb, 1, 0);
+
+    put_bits(&s->pb, 1, 1);	/* marker */
+    put_bits(&s->pb, 4, 1);	/* XXX: correct time increment */
+    put_bits(&s->pb, 1, 1);	/* marker */
+    put_bits(&s->pb, 1, 1);	/* vop coded */
+    if (s->pict_type == P_TYPE) {
+        s->no_rounding = 0;
+	put_bits(&s->pb, 1, s->no_rounding);	/* rounding type */
+    }
+    put_bits(&s->pb, 3, 0);	/* intra dc VLC threshold */
+
+    put_bits(&s->pb, 5, s->qscale);
+
+    if (s->pict_type != I_TYPE)
+	put_bits(&s->pb, 3, s->f_code);	/* fcode_for */
+    //    printf("****frame %d\n", picture_number);
+}
+
+void h263_dc_scale(MpegEncContext * s)
+{
+    int quant;
+
+    quant = s->qscale;
+    /* luminance */
+    if (quant < 5)
+	s->y_dc_scale = 8;
+    else if (quant > 4 && quant < 9)
+	s->y_dc_scale = (2 * quant);
+    else if (quant > 8 && quant < 25)
+	s->y_dc_scale = (quant + 8);
+    else
+	s->y_dc_scale = (2 * quant - 16);
+    /* chrominance */
+    if (quant < 5)
+	s->c_dc_scale = 8;
+    else if (quant > 4 && quant < 25)
+	s->c_dc_scale = ((quant + 13) / 2);
+    else
+	s->c_dc_scale = (quant - 6);
+}
+
+static int mpeg4_pred_dc(MpegEncContext * s, int n, UINT16 **dc_val_ptr, int *dir_ptr)
+{
+    int a, b, c, x, y, wrap, pred, scale;
+    UINT16 *dc_val;
+
+    /* find prediction */
+    if (n < 4) {
+	x = 2 * s->mb_x + 1 + (n & 1);
+	y = 2 * s->mb_y + 1 + ((n & 2) >> 1);
+	wrap = s->mb_width * 2 + 2;
+	dc_val = s->dc_val[0];
+	scale = s->y_dc_scale;
+    } else {
+	x = s->mb_x + 1;
+	y = s->mb_y + 1;
+	wrap = s->mb_width + 2;
+	dc_val = s->dc_val[n - 4 + 1];
+	scale = s->c_dc_scale;
+    }
+
+    /* B C
+     * A X 
+     */
+    a = dc_val[(x - 1) + (y) * wrap];
+    b = dc_val[(x - 1) + (y - 1) * wrap];
+    c = dc_val[(x) + (y - 1) * wrap];
+
+    if (abs(a - b) < abs(b - c)) {
+	pred = c;
+        *dir_ptr = 1; /* top */
+    } else {
+	pred = a;
+        *dir_ptr = 0; /* left */
+    }
+    /* we assume pred is positive */
+    pred = (pred + (scale >> 1)) / scale;
+
+    /* prepare address for prediction update */
+    *dc_val_ptr = &dc_val[(x) + (y) * wrap];
+
+    return pred;
+}
+
+void mpeg4_pred_ac(MpegEncContext * s, INT16 *block, int n, 
+                   int dir)
+{
+    int x, y, wrap, i;
+    INT16 *ac_val, *ac_val1;
+
+    /* find prediction */
+    if (n < 4) {
+	x = 2 * s->mb_x + 1 + (n & 1);
+	y = 2 * s->mb_y + 1 + ((n & 2) >> 1);
+	wrap = s->mb_width * 2 + 2;
+	ac_val = s->ac_val[0][0];
+    } else {
+	x = s->mb_x + 1;
+	y = s->mb_y + 1;
+	wrap = s->mb_width + 2;
+	ac_val = s->ac_val[n - 4 + 1][0];
+    }
+    ac_val += ((y) * wrap + (x)) * 16;
+    ac_val1 = ac_val;
+    if (s->ac_pred) {
+        if (dir == 0) {
+            /* left prediction */
+            ac_val -= 16;
+            for(i=1;i<8;i++) {
+                block[i*8] += ac_val[i];
+            }
+        } else {
+            /* top prediction */
+            ac_val -= 16 * wrap;
+            for(i=1;i<8;i++) {
+                block[i] += ac_val[i + 8];
+            }
+        }
+    }
+    /* left copy */
+    for(i=1;i<8;i++)
+        ac_val1[i] = block[i * 8];
+    /* top copy */
+    for(i=1;i<8;i++)
+        ac_val1[8 + i] = block[i];
+}
+
+static inline void mpeg4_encode_dc(MpegEncContext * s, int level, int n, int *dir_ptr)
+{
+    int size, v, pred;
+    UINT16 *dc_val;
+
+    pred = mpeg4_pred_dc(s, n, &dc_val, dir_ptr);
+    if (n < 4) {
+        *dc_val = level * s->y_dc_scale;
+    } else {
+        *dc_val = level * s->c_dc_scale;
+    }
+
+    /* do the prediction */
+    level -= pred;
+    /* find number of bits */
+    size = 0;
+    v = abs(level);
+    while (v) {
+	v >>= 1;
+	size++;
+    }
+
+    if (n < 4) {
+	/* luminance */
+	put_bits(&s->pb, DCtab_lum[size][1], DCtab_lum[size][0]);
+    } else {
+	/* chrominance */
+	put_bits(&s->pb, DCtab_chrom[size][1], DCtab_chrom[size][0]);
+    }
+
+    /* encode remaining bits */
+    if (size > 0) {
+	if (level < 0)
+	    level = (-level) ^ ((1 << size) - 1);
+	put_bits(&s->pb, size, level);
+	if (size > 8)
+	    put_bits(&s->pb, 1, 1);
+    }
+}
+
+static void mpeg4_encode_block(MpegEncContext * s, DCTELEM * block, int n)
+{
+    int level, run, last, i, j, last_index, last_non_zero, sign, slevel;
+    int code, dc_pred_dir;
+    const RLTable *rl;
+
+    if (s->mb_intra) {
+	/* mpeg4 based DC predictor */
+	mpeg4_encode_dc(s, block[0], n, &dc_pred_dir);
+	i = 1;
+        rl = &rl_intra;
+    } else {
+	i = 0;
+        rl = &rl_inter;
+    }
+
+    /* AC coefs */
+    last_index = s->block_last_index[n];
+    last_non_zero = i - 1;
+    for (; i <= last_index; i++) {
+	j = zigzag_direct[i];
+	level = block[j];
+	if (level) {
+	    run = i - last_non_zero - 1;
+	    last = (i == last_index);
+	    sign = 0;
+	    slevel = level;
+	    if (level < 0) {
+		sign = 1;
+		level = -level;
+	    }
+            code = get_rl_index(rl, last, run, level);
+            put_bits(&s->pb, rl->table_vlc[code][1], rl->table_vlc[code][0]);
+            if (code == rl->n) {
+                int level1, run1;
+                level1 = level - rl->max_level[last][run];
+                if (level1 < 1) 
+                    goto esc2;
+                code = get_rl_index(rl, last, run, level1);
+                if (code == rl->n) {
+                esc2:
+                    put_bits(&s->pb, 1, 1);
+                    if (level > MAX_LEVEL)
+                        goto esc3;
+                    run1 = run - rl->max_run[last][level] - 1;
+                    if (run1 < 0)
+                        goto esc3;
+                    code = get_rl_index(rl, last, run1, level);
+                    if (code == rl->n) {
+                    esc3:
+                        /* third escape */
+                        put_bits(&s->pb, 1, 1);
+                        put_bits(&s->pb, 1, last);
+                        put_bits(&s->pb, 6, run);
+                        put_bits(&s->pb, 1, 1);
+                        put_bits(&s->pb, 12, slevel & 0xfff);
+                        put_bits(&s->pb, 1, 1);
+                    } else {
+                        /* second escape */
+                        put_bits(&s->pb, 1, 0);
+                        put_bits(&s->pb, rl->table_vlc[code][1], rl->table_vlc[code][0]);
+                        put_bits(&s->pb, 1, sign);
+                    }
+                } else {
+                    /* first escape */
+                    put_bits(&s->pb, 1, 0);
+                    put_bits(&s->pb, rl->table_vlc[code][1], rl->table_vlc[code][0]);
+                    put_bits(&s->pb, 1, sign);
+                }
+            } else {
+                put_bits(&s->pb, 1, sign);
+            }
+	    last_non_zero = i;
+	}
+    }
+}
+
+
+
+/***********************************************/
+/* decoding */
+
+static VLC intra_MCBPC_vlc;
+static VLC inter_MCBPC_vlc;
+static VLC cbpy_vlc;
+static VLC mv_vlc;
+static VLC dc_lum, dc_chrom;
+
+void init_rl(RLTable *rl)
+{
+    INT8 max_level[MAX_RUN+1], max_run[MAX_LEVEL+1];
+    UINT8 index_run[MAX_RUN+1];
+    int last, run, level, start, end, i;
+
+    /* compute max_level[], max_run[] and index_run[] */
+    for(last=0;last<2;last++) {
+        if (last == 0) {
+            start = 0;
+            end = rl->last;
+        } else {
+            start = rl->last;
+            end = rl->n;
+        }
+
+        memset(max_level, 0, MAX_RUN + 1);
+        memset(max_run, 0, MAX_LEVEL + 1);
+        memset(index_run, rl->n, MAX_RUN + 1);
+        for(i=start;i<end;i++) {
+            run = rl->table_run[i];
+            level = rl->table_level[i];
+            if (index_run[run] == rl->n)
+                index_run[run] = i;
+            if (level > max_level[run])
+                max_level[run] = level;
+            if (run > max_run[level])
+                max_run[level] = run;
+        }
+        rl->max_level[last] = malloc(MAX_RUN + 1);
+        memcpy(rl->max_level[last], max_level, MAX_RUN + 1);
+        rl->max_run[last] = malloc(MAX_LEVEL + 1);
+        memcpy(rl->max_run[last], max_run, MAX_LEVEL + 1);
+        rl->index_run[last] = malloc(MAX_RUN + 1);
+        memcpy(rl->index_run[last], index_run, MAX_RUN + 1);
+    }
+}
+
+void init_vlc_rl(RLTable *rl)
+{
+    init_vlc(&rl->vlc, 9, rl->n + 1, 
+             &rl->table_vlc[0][1], 4, 2,
+             &rl->table_vlc[0][0], 4, 2);
+}
+
+/* init vlcs */
+
+/* XXX: find a better solution to handle static init */
+void h263_decode_init_vlc(MpegEncContext *s)
+{
+    static int done = 0;
+
+    if (!done) {
+        done = 1;
+
+        init_vlc(&intra_MCBPC_vlc, 6, 8, 
+                 intra_MCBPC_bits, 1, 1,
+                 intra_MCBPC_code, 1, 1);
+        init_vlc(&inter_MCBPC_vlc, 9, 20, 
+                 inter_MCBPC_bits, 1, 1,
+                 inter_MCBPC_code, 1, 1);
+        init_vlc(&cbpy_vlc, 6, 16,
+                 &cbpy_tab[0][1], 2, 1,
+                 &cbpy_tab[0][0], 2, 1);
+        init_vlc(&mv_vlc, 9, 33,
+                 &mvtab[0][1], 2, 1,
+                 &mvtab[0][0], 2, 1);
+        init_rl(&rl_inter);
+        init_rl(&rl_intra);
+        init_vlc_rl(&rl_inter);
+        init_vlc_rl(&rl_intra);
+        init_vlc(&dc_lum, 9, 13,
+                 &DCtab_lum[0][1], 2, 1,
+                 &DCtab_lum[0][0], 2, 1);
+        init_vlc(&dc_chrom, 9, 13,
+                 &DCtab_chrom[0][1], 2, 1,
+                 &DCtab_chrom[0][0], 2, 1);
+    }
+}
+
+int h263_decode_mb(MpegEncContext *s,
+                   DCTELEM block[6][64])
+{
+    int cbpc, cbpy, i, cbp, pred_x, pred_y, mx, my, dquant;
+    INT16 *mot_val;
+    static UINT8 quant_tab[4] = { -1, -2, 1, 2 };
+
+    if (s->pict_type == P_TYPE) {
+        if (get_bits(&s->gb, 1)) {
+            /* skip mb */
+            s->mb_intra = 0;
+            for(i=0;i<6;i++)
+                s->block_last_index[i] = -1;
+            s->mv_dir = MV_DIR_FORWARD;
+            s->mv_type = MV_TYPE_16X16;
+            s->mv[0][0][0] = 0;
+            s->mv[0][0][1] = 0;
+            return 0;
+        }
+        cbpc = get_vlc(&s->gb, &inter_MCBPC_vlc);
+        if (cbpc < 0)
+            return -1;
+        dquant = cbpc & 8;
+        s->mb_intra = ((cbpc & 4) != 0);
+    } else {
+        cbpc = get_vlc(&s->gb, &intra_MCBPC_vlc);
+        if (cbpc < 0)
+            return -1;
+        dquant = cbpc & 4;
+        s->mb_intra = 1;
+    }
+
+    if (!s->mb_intra) {
+        cbpy = get_vlc(&s->gb, &cbpy_vlc);
+        cbp = (cbpc & 3) | ((cbpy ^ 0xf) << 2);
+        if (dquant) {
+            s->qscale += quant_tab[get_bits(&s->gb, 2)];
+            if (s->qscale < 1)
+                s->qscale = 1;
+            else if (s->qscale > 31)
+                s->qscale = 31;
+        }
+        s->mv_dir = MV_DIR_FORWARD;
+        if ((cbpc & 16) == 0) {
+            /* 16x16 motion prediction */
+            s->mv_type = MV_TYPE_16X16;
+            h263_pred_motion(s, 0, &pred_x, &pred_y);
+            mx = h263_decode_motion(s, pred_x);
+            if (mx >= 0xffff)
+                return -1;
+            my = h263_decode_motion(s, pred_y);
+            if (my >= 0xffff)
+                return -1;
+            s->mv[0][0][0] = mx;
+            s->mv[0][0][1] = my;
+        } else {
+            s->mv_type = MV_TYPE_8X8;
+            for(i=0;i<4;i++) {
+                mot_val = h263_pred_motion(s, i, &pred_x, &pred_y);
+                mx = h263_decode_motion(s, pred_x);
+                if (mx >= 0xffff)
+                    return -1;
+                my = h263_decode_motion(s, pred_y);
+                if (my >= 0xffff)
+                    return -1;
+                s->mv[0][i][0] = mx;
+                s->mv[0][i][1] = my;
+                mot_val[0] = mx;
+                mot_val[1] = my;
+            }
+        }
+    } else {
+        s->ac_pred = 0;
+	if (s->h263_pred) {
+            s->ac_pred = get_bits(&s->gb, 1);
+        }
+        cbpy = get_vlc(&s->gb, &cbpy_vlc);
+        cbp = (cbpc & 3) | (cbpy << 2);
+        if (dquant) {
+            s->qscale += quant_tab[get_bits(&s->gb, 2)];
+            if (s->qscale < 1)
+                s->qscale = 1;
+            else if (s->qscale > 31)
+                s->qscale = 31;
+        }
+    }
+
+    /* decode each block */
+    if (s->h263_pred) {
+	for (i = 0; i < 6; i++) {
+	    if (mpeg4_decode_block(s, block[i], i, (cbp >> (5 - i)) & 1) < 0)
+                return -1;
+	}
+    } else {
+	for (i = 0; i < 6; i++) {
+	    if (h263_decode_block(s, block[i], i, (cbp >> (5 - i)) & 1) < 0)
+                return -1;
+	}
+    }
+    return 0;
+}
+
+static int h263_decode_motion(MpegEncContext * s, int pred)
+{
+    int code, val, sign, shift, l, m;
+
+    code = get_vlc(&s->gb, &mv_vlc);
+    if (code < 0)
+        return 0xffff;
+
+    if (code == 0)
+        return pred;
+    sign = get_bits(&s->gb, 1);
+    shift = s->f_code - 1;
+    val = (code - 1) << shift;
+    if (shift > 0)
+        val |= get_bits(&s->gb, shift);
+    val++;
+    if (sign)
+        val = -val;
+    val += pred;
+    
+    /* modulo decoding */
+    if (!s->h263_long_vectors) {
+        l = (1 << (s->f_code - 1)) * 32;
+        m = 2 * l;
+        if (val < -l) {
+            val += m;
+        } else if (val >= l) {
+            val -= m;
+        }
+    } else {
+        /* horrible h263 long vector mode */
+        if (pred < -31 && val < -63)
+            val += 64;
+        if (pred > 32 && val > 63)
+            val -= 64;
+    }
+    return val;
+}
+
+static int h263_decode_block(MpegEncContext * s, DCTELEM * block,
+                             int n, int coded)
+{
+    int code, level, i, j, last, run;
+    RLTable *rl = &rl_inter;
+
+    if (s->mb_intra) {
+	/* DC coef */
+        if (s->h263_rv10 && s->rv10_version == 3 && s->pict_type == I_TYPE) {
+            int component, diff;
+            component = (n <= 3 ? 0 : n - 4 + 1);
+            level = s->last_dc[component];
+            if (s->rv10_first_dc_coded[component]) {
+                diff = rv_decode_dc(s, n);
+                if (diff == 0xffff)
+                    return -1;
+                level += diff;
+                level = level & 0xff; /* handle wrap round */
+                s->last_dc[component] = level;
+            } else {
+                s->rv10_first_dc_coded[component] = 1;
+            }
+        } else {
+            level = get_bits(&s->gb, 8);
+            if (level == 255)
+                level = 128;
+        }
+        block[0] = level;
+	i = 1;
+    } else {
+	i = 0;
+    }
+    if (!coded) {
+        s->block_last_index[n] = i - 1;
+        return 0;
+    }
+
+    for(;;) {
+        code = get_vlc(&s->gb, &rl->vlc);
+        if (code < 0)
+            return -1;
+        if (code == rl->n) {
+            /* escape */
+            last = get_bits(&s->gb, 1);
+            run = get_bits(&s->gb, 6);
+            level = (INT8)get_bits(&s->gb, 8);
+            if (s->h263_rv10 && level == -128) {
+                /* XXX: should patch encoder too */
+                level = get_bits(&s->gb, 12);
+                level = (level << 20) >> 20;
+            }
+        } else {
+            run = rl->table_run[code];
+            level = rl->table_level[code];
+            last = code >= rl->last;
+            if (get_bits(&s->gb, 1))
+                level = -level;
+        }
+        i += run;
+        if (i >= 64)
+            return -1;
+	j = zigzag_direct[i];
+        block[j] = level;
+        if (last)
+            break;
+        i++;
+    }
+    s->block_last_index[n] = i;
+    return 0;
+}
+
+static int mpeg4_decode_dc(MpegEncContext * s, int n, int *dir_ptr)
+{
+    int level, pred, code;
+    UINT16 *dc_val;
+
+    if (n < 4) 
+        code = get_vlc(&s->gb, &dc_lum);
+    else 
+        code = get_vlc(&s->gb, &dc_chrom);
+    if (code < 0)
+        return -1;
+    if (code == 0) {
+        level = 0;
+    } else {
+        level = get_bits(&s->gb, code);
+        if ((level >> (code - 1)) == 0) /* if MSB not set it is negative*/
+            level = - (level ^ ((1 << code) - 1));
+        if (code > 8)
+            get_bits(&s->gb, 1); /* marker */
+    }
+
+    pred = mpeg4_pred_dc(s, n, &dc_val, dir_ptr);
+    level += pred;
+    if (level < 0)
+        level = 0;
+    if (n < 4) {
+        *dc_val = level * s->y_dc_scale;
+    } else {
+        *dc_val = level * s->c_dc_scale;
+    }
+    return level;
+}
+
+static int mpeg4_decode_block(MpegEncContext * s, DCTELEM * block,
+                              int n, int coded)
+{
+    int code, level, i, j, last, run;
+    int dc_pred_dir;
+    RLTable *rl;
+    const UINT8 *scan_table;
+
+    if (s->mb_intra) {
+	/* DC coef */
+        level = mpeg4_decode_dc(s, n, &dc_pred_dir);
+        if (level < 0)
+            return -1;
+        block[0] = level;
+	i = 1;
+        if (!coded) 
+            goto not_coded;
+        rl = &rl_intra;
+        if (s->ac_pred) {
+            if (dc_pred_dir == 0) 
+                scan_table = ff_alternate_vertical_scan; /* left */
+            else
+                scan_table = ff_alternate_horizontal_scan; /* top */
+        } else {
+            scan_table = zigzag_direct;
+        }
+    } else {
+	i = 0;
+        if (!coded) {
+            s->block_last_index[n] = i - 1;
+            return 0;
+        }
+        rl = &rl_inter;
+        scan_table = zigzag_direct;
+    }
+
+    for(;;) {
+        code = get_vlc(&s->gb, &rl->vlc);
+        if (code < 0)
+            return -1;
+        if (code == rl->n) {
+            /* escape */
+            if (get_bits(&s->gb, 1) != 0) {
+                if (get_bits(&s->gb, 1) != 0) {
+                    /* third escape */
+                    last = get_bits(&s->gb, 1);
+                    run = get_bits(&s->gb, 6);
+                    get_bits(&s->gb, 1); /* marker */
+                    level = get_bits(&s->gb, 12);
+                    level = (level << 20) >> 20; /* sign extend */
+                    get_bits(&s->gb, 1); /* marker */
+                } else {
+                    /* second escape */
+                    code = get_vlc(&s->gb, &rl->vlc);
+                    if (code < 0 || code >= rl->n)
+                        return -1;
+                    run = rl->table_run[code];
+                    level = rl->table_level[code];
+                    last = code >= rl->last;
+                    run += rl->max_run[last][level] + 1;
+                    if (get_bits(&s->gb, 1))
+                        level = -level;
+                }
+            } else {
+                /* first escape */
+                code = get_vlc(&s->gb, &rl->vlc);
+                if (code < 0 || code >= rl->n)
+                    return -1;
+                run = rl->table_run[code];
+                level = rl->table_level[code];
+                last = code >= rl->last;
+                level += rl->max_level[last][run];
+                if (get_bits(&s->gb, 1))
+                    level = -level;
+            }
+        } else {
+            run = rl->table_run[code];
+            level = rl->table_level[code];
+            last = code >= rl->last;
+            if (get_bits(&s->gb, 1))
+                level = -level;
+        }
+        i += run;
+        if (i >= 64)
+            return -1;
+	j = scan_table[i];
+        block[j] = level;
+        i++;
+        if (last)
+            break;
+    }
+ not_coded:
+    if (s->mb_intra) {
+        mpeg4_pred_ac(s, block, n, dc_pred_dir);
+        if (s->ac_pred) {
+            i = 64; /* XXX: not optimal */
+        }
+    }
+    s->block_last_index[n] = i - 1;
+    return 0;
+}
+
+/* most is hardcoded. should extend to handle all h263 streams */
+int h263_decode_picture_header(MpegEncContext *s)
+{
+    int format, width, height;
+
+    /* picture header */
+    if (get_bits(&s->gb, 22) != 0x20)
+        return -1;
+    get_bits(&s->gb, 8); /* picture timestamp */
+
+    if (get_bits(&s->gb, 1) != 1)
+        return -1;	/* marker */
+    if (get_bits(&s->gb, 1) != 0)
+        return -1;	/* h263 id */
+    get_bits(&s->gb, 1);	/* split screen off */
+    get_bits(&s->gb, 1);	/* camera  off */
+    get_bits(&s->gb, 1);	/* freeze picture release off */
+
+    format = get_bits(&s->gb, 3);
+
+    if (format != 7) {
+        s->h263_plus = 0;
+        /* H.263v1 */
+        width = h263_format[format][0];
+        height = h263_format[format][1];
+        if (!width)
+            return -1;
+
+        s->pict_type = I_TYPE + get_bits(&s->gb, 1);
+
+        s->unrestricted_mv = get_bits(&s->gb, 1); 
+        s->h263_long_vectors = s->unrestricted_mv;
+
+        if (get_bits(&s->gb, 1) != 0)
+            return -1;	/* SAC: off */
+        if (get_bits(&s->gb, 1) != 0)
+            return -1;	/* advanced prediction mode: off */
+        if (get_bits(&s->gb, 1) != 0)
+            return -1;	/* not PB frame */
+
+        s->qscale = get_bits(&s->gb, 5);
+        get_bits(&s->gb, 1);	/* Continuous Presence Multipoint mode: off */
+    } else {
+        s->h263_plus = 1;
+        /* H.263v2 */
+        if (get_bits(&s->gb, 3) != 1)
+            return -1;
+        if (get_bits(&s->gb, 3) != 6) /* custom source format */
+            return -1;
+        get_bits(&s->gb, 12);
+        get_bits(&s->gb, 3);
+        s->pict_type = get_bits(&s->gb, 3) + 1;
+        if (s->pict_type != I_TYPE &&
+            s->pict_type != P_TYPE)
+            return -1;
+        get_bits(&s->gb, 7);
+        get_bits(&s->gb, 4); /* aspect ratio */
+        width = (get_bits(&s->gb, 9) + 1) * 4;
+        get_bits(&s->gb, 1);
+        height = get_bits(&s->gb, 9) * 4;
+        if (height == 0)
+            return -1;
+        s->qscale = get_bits(&s->gb, 5);
+    }
+    /* PEI */
+    while (get_bits(&s->gb, 1) != 0) {
+        get_bits(&s->gb, 8);
+    }
+    s->f_code = 1;
+    s->width = width;
+    s->height = height;
+    return 0;
+}
+
+/* decode mpeg4 VOP header */
+int mpeg4_decode_picture_header(MpegEncContext * s)
+{
+    int time_incr, startcode, state, v;
+
+ redo:
+    /* search next start code */
+    align_get_bits(&s->gb);
+    state = 0xff;
+    for(;;) {
+        v = get_bits(&s->gb, 8);
+        if (state == 0x000001) {
+            state = ((state << 8) | v) & 0xffffff;
+            startcode = state;
+            break;
+        }
+        state = ((state << 8) | v) & 0xffffff;
+        /* XXX: really detect end of frame */
+        if (state == 0)
+            return -1;
+    }
+
+    if (startcode == 0x120) {
+        int time_increment_resolution, width, height;
+
+        /* vol header */
+        get_bits(&s->gb, 1); /* random access */
+        get_bits(&s->gb, 8); /* vo_type */
+        get_bits(&s->gb, 1); /* is_ol_id */
+        get_bits(&s->gb, 4); /* vo_ver_id */
+        get_bits(&s->gb, 3); /* vo_priority */
+        
+        get_bits(&s->gb, 4); /* aspect_ratio_info */
+        get_bits(&s->gb, 1); /* vol control parameter */
+        get_bits(&s->gb, 2); /* vol shape */
+        get_bits(&s->gb, 1); /* marker */
+        
+        time_increment_resolution = get_bits(&s->gb, 16);
+        s->time_increment_bits = log2(time_increment_resolution - 1) + 1;
+        get_bits(&s->gb, 1); /* marker */
+
+        get_bits(&s->gb, 1); /* vop rate  */
+        get_bits(&s->gb, s->time_increment_bits);
+        get_bits(&s->gb, 1); /* marker */
+        
+        width = get_bits(&s->gb, 13);
+        get_bits(&s->gb, 1); /* marker */
+        height = get_bits(&s->gb, 13);
+        get_bits(&s->gb, 1); /* marker */
+        
+        get_bits(&s->gb, 1); /* interfaced */
+        get_bits(&s->gb, 1); /* OBMC */
+        get_bits(&s->gb, 2); /* vol_sprite_usage */
+        get_bits(&s->gb, 1); /* not_8_bit */
+
+        get_bits(&s->gb, 1); /* vol_quant_type */
+        get_bits(&s->gb, 1); /* vol_quarter_pixel */
+        get_bits(&s->gb, 1); /* complexity_estimation_disabled */
+        get_bits(&s->gb, 1); /* resync_marker_disabled */
+        get_bits(&s->gb, 1); /* data_partioning_enabled */
+        goto redo;
+    } else if (startcode != 0x1b6) {
+        goto redo;
+    }
+
+    s->pict_type = get_bits(&s->gb, 2) + 1;	/* pict type: I = 0 , P = 1 */
+    if (s->pict_type != I_TYPE &&
+        s->pict_type != P_TYPE)
+        return -1;
+    
+    /* XXX: parse time base */
+    time_incr = 0;
+    while (get_bits(&s->gb, 1) != 0) 
+        time_incr++;
+
+    get_bits(&s->gb, 1);	/* marker */
+    get_bits(&s->gb, s->time_increment_bits);
+    get_bits(&s->gb, 1);	/* marker */
+    /* vop coded */
+    if (get_bits(&s->gb, 1) != 1)
+	return -1; 
+    
+    if (s->pict_type == P_TYPE) {
+        /* rounding type for motion estimation */
+	s->no_rounding = get_bits(&s->gb, 1);
+    }
+        
+    if (get_bits(&s->gb, 3) != 0)
+	return -1; /* intra dc VLC threshold */
+
+    s->qscale = get_bits(&s->gb, 5);
+
+    if (s->pict_type != I_TYPE) {
+	s->f_code = get_bits(&s->gb, 3);	/* fcode_for */
+    }
+    return 0;
+}
+
+/* don't understand why they choose a different header ! */
+int intel_h263_decode_picture_header(MpegEncContext *s)
+{
+    int format;
+
+    /* picture header */
+    if (get_bits(&s->gb, 22) != 0x20)
+        return -1;
+    get_bits(&s->gb, 8); /* picture timestamp */
+
+    if (get_bits(&s->gb, 1) != 1)
+        return -1;	/* marker */
+    if (get_bits(&s->gb, 1) != 0)
+        return -1;	/* h263 id */
+    get_bits(&s->gb, 1);	/* split screen off */
+    get_bits(&s->gb, 1);	/* camera  off */
+    get_bits(&s->gb, 1);	/* freeze picture release off */
+
+    format = get_bits(&s->gb, 3);
+    if (format != 7)
+        return -1;
+
+    s->h263_plus = 0;
+
+    s->pict_type = I_TYPE + get_bits(&s->gb, 1);
+    
+    s->unrestricted_mv = get_bits(&s->gb, 1); 
+    s->h263_long_vectors = s->unrestricted_mv;
+
+    if (get_bits(&s->gb, 1) != 0)
+        return -1;	/* SAC: off */
+    if (get_bits(&s->gb, 1) != 0)
+        return -1;	/* advanced prediction mode: off */
+    if (get_bits(&s->gb, 1) != 0)
+        return -1;	/* not PB frame */
+
+    /* skip unknown header garbage */
+    get_bits(&s->gb, 41);
+
+    s->qscale = get_bits(&s->gb, 5);
+    get_bits(&s->gb, 1);	/* Continuous Presence Multipoint mode: off */
+
+    /* PEI */
+    while (get_bits(&s->gb, 1) != 0) {
+        get_bits(&s->gb, 8);
+    }
+    s->f_code = 1;
+    return 0;
+}
diff --git a/libavcodec/h263data.h b/libavcodec/h263data.h
new file mode 100644
index 0000000000..4fd9d36298
--- /dev/null
+++ b/libavcodec/h263data.h
@@ -0,0 +1,115 @@
+
+/* intra MCBPC, mb_type = (intra), then (intraq) */
+static const UINT8 intra_MCBPC_code[8] = { 1, 1, 2, 3, 1, 1, 2, 3 };
+static const UINT8 intra_MCBPC_bits[8] = { 1, 3, 3, 3, 4, 6, 6, 6 };
+
+/* inter MCBPC, mb_type = (inter), (intra), (interq), (intraq), (inter4v) */
+static const UINT8 inter_MCBPC_code[20] = { 
+    1, 3, 2, 5, 
+    3, 4, 3, 3, 
+    0, 1, 2, 3,
+    4, 4, 3, 2,
+    2, 5, 4, 5,
+};
+static const UINT8 inter_MCBPC_bits[20] = { 
+    1, 4, 4, 6, 
+    5, 8, 8, 7,
+    12, 12, 12, 12,
+    6, 9, 9, 9,
+    3, 7, 7, 8,
+};
+
+static const UINT8 cbpy_tab[16][2] =
+{
+  {3,4}, {5,5}, {4,5}, {9,4}, {3,5}, {7,4}, {2,6}, {11,4},
+  {2,5}, {3,6}, {5,4}, {10,4}, {4,4}, {8,4}, {6,4}, {3,2}
+};
+
+static const UINT8 mvtab[33][2] =
+{
+  {1,1}, {1,2}, {1,3}, {1,4}, {3,6}, {5,7}, {4,7}, {3,7},
+  {11,9}, {10,9}, {9,9}, {17,10}, {16,10}, {15,10}, {14,10}, {13,10},
+  {12,10}, {11,10}, {10,10}, {9,10}, {8,10}, {7,10}, {6,10}, {5,10},
+  {4,10}, {7,11}, {6,11}, {5,11}, {4,11}, {3,11}, {2,11}, {3,12},
+  {2,12}
+};
+
+/* third non intra table */
+const UINT16 inter_vlc[103][2] = {
+{ 0x2, 2 },{ 0xf, 4 },{ 0x15, 6 },{ 0x17, 7 },
+{ 0x1f, 8 },{ 0x25, 9 },{ 0x24, 9 },{ 0x21, 10 },
+{ 0x20, 10 },{ 0x7, 11 },{ 0x6, 11 },{ 0x20, 11 },
+{ 0x6, 3 },{ 0x14, 6 },{ 0x1e, 8 },{ 0xf, 10 },
+{ 0x21, 11 },{ 0x50, 12 },{ 0xe, 4 },{ 0x1d, 8 },
+{ 0xe, 10 },{ 0x51, 12 },{ 0xd, 5 },{ 0x23, 9 },
+{ 0xd, 10 },{ 0xc, 5 },{ 0x22, 9 },{ 0x52, 12 },
+{ 0xb, 5 },{ 0xc, 10 },{ 0x53, 12 },{ 0x13, 6 },
+{ 0xb, 10 },{ 0x54, 12 },{ 0x12, 6 },{ 0xa, 10 },
+{ 0x11, 6 },{ 0x9, 10 },{ 0x10, 6 },{ 0x8, 10 },
+{ 0x16, 7 },{ 0x55, 12 },{ 0x15, 7 },{ 0x14, 7 },
+{ 0x1c, 8 },{ 0x1b, 8 },{ 0x21, 9 },{ 0x20, 9 },
+{ 0x1f, 9 },{ 0x1e, 9 },{ 0x1d, 9 },{ 0x1c, 9 },
+{ 0x1b, 9 },{ 0x1a, 9 },{ 0x22, 11 },{ 0x23, 11 },
+{ 0x56, 12 },{ 0x57, 12 },{ 0x7, 4 },{ 0x19, 9 },
+{ 0x5, 11 },{ 0xf, 6 },{ 0x4, 11 },{ 0xe, 6 },
+{ 0xd, 6 },{ 0xc, 6 },{ 0x13, 7 },{ 0x12, 7 },
+{ 0x11, 7 },{ 0x10, 7 },{ 0x1a, 8 },{ 0x19, 8 },
+{ 0x18, 8 },{ 0x17, 8 },{ 0x16, 8 },{ 0x15, 8 },
+{ 0x14, 8 },{ 0x13, 8 },{ 0x18, 9 },{ 0x17, 9 },
+{ 0x16, 9 },{ 0x15, 9 },{ 0x14, 9 },{ 0x13, 9 },
+{ 0x12, 9 },{ 0x11, 9 },{ 0x7, 10 },{ 0x6, 10 },
+{ 0x5, 10 },{ 0x4, 10 },{ 0x24, 11 },{ 0x25, 11 },
+{ 0x26, 11 },{ 0x27, 11 },{ 0x58, 12 },{ 0x59, 12 },
+{ 0x5a, 12 },{ 0x5b, 12 },{ 0x5c, 12 },{ 0x5d, 12 },
+{ 0x5e, 12 },{ 0x5f, 12 },{ 0x3, 7 },
+};
+
+const INT8 inter_level[102] = {
+  1,  2,  3,  4,  5,  6,  7,  8,
+  9, 10, 11, 12,  1,  2,  3,  4,
+  5,  6,  1,  2,  3,  4,  1,  2,
+  3,  1,  2,  3,  1,  2,  3,  1,
+  2,  3,  1,  2,  1,  2,  1,  2,
+  1,  2,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  2,  3,  1,  2,  1,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  1,
+};
+
+const INT8 inter_run[102] = {
+  0,  0,  0,  0,  0,  0,  0,  0,
+  0,  0,  0,  0,  1,  1,  1,  1,
+  1,  1,  2,  2,  2,  2,  3,  3,
+  3,  4,  4,  4,  5,  5,  5,  6,
+  6,  6,  7,  7,  8,  8,  9,  9,
+ 10, 10, 11, 12, 13, 14, 15, 16,
+ 17, 18, 19, 20, 21, 22, 23, 24,
+ 25, 26,  0,  0,  0,  1,  1,  2,
+  3,  4,  5,  6,  7,  8,  9, 10,
+ 11, 12, 13, 14, 15, 16, 17, 18,
+ 19, 20, 21, 22, 23, 24, 25, 26,
+ 27, 28, 29, 30, 31, 32, 33, 34,
+ 35, 36, 37, 38, 39, 40,
+};
+
+static RLTable rl_inter = {
+    102,
+    58,
+    inter_vlc,
+    inter_run,
+    inter_level,
+};
+
+static const UINT16 h263_format[8][2] = {
+    { 0, 0 },
+    { 128, 96 },
+    { 176, 144 },
+    { 352, 288 },
+    { 704, 576 },
+    { 1408, 1152 },
+};
+
diff --git a/libavcodec/h263dec.c b/libavcodec/h263dec.c
new file mode 100644
index 0000000000..2af76199ab
--- /dev/null
+++ b/libavcodec/h263dec.c
@@ -0,0 +1,205 @@
+/*
+ * H263 decoder
+ * Copyright (c) 2001 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include "dsputil.h"
+#include "avcodec.h"
+#include "mpegvideo.h"
+
+//#define DEBUG
+
+static int h263_decode_init(AVCodecContext *avctx)
+{
+    MpegEncContext *s = avctx->priv_data;
+    
+    s->out_format = FMT_H263;
+
+    s->width = avctx->width;
+    s->height = avctx->height;
+
+    /* select sub codec */
+    switch(avctx->codec->id) {
+    case CODEC_ID_H263:
+        break;
+    case CODEC_ID_OPENDIVX:
+        s->time_increment_bits = 4; /* default value for broken headers */
+        s->h263_pred = 1;
+        break;
+    case CODEC_ID_MSMPEG4:
+        s->h263_msmpeg4 = 1;
+        s->h263_pred = 1;
+        break;
+    case CODEC_ID_H263I:
+        s->h263_intel = 1;
+        break;
+    default:
+        return -1;
+    }
+
+    /* for h263, we allocate the images after having read the header */
+    if (MPV_common_init(s) < 0)
+        return -1;
+
+    if (s->h263_msmpeg4)
+        msmpeg4_decode_init_vlc(s);
+    else
+        h263_decode_init_vlc(s);
+    
+    return 0;
+}
+
+static int h263_decode_end(AVCodecContext *avctx)
+{
+    MpegEncContext *s = avctx->priv_data;
+
+    MPV_common_end(s);
+    return 0;
+}
+
+static int h263_decode_frame(AVCodecContext *avctx, 
+                             void *data, int *data_size,
+                             UINT8 *buf, int buf_size)
+{
+    MpegEncContext *s = avctx->priv_data;
+    int ret;
+    DCTELEM block[6][64];
+    AVPicture *pict = data; 
+
+#ifdef DEBUG
+    printf("*****frame %d size=%d\n", avctx->frame_number, buf_size);
+    printf("bytes=%x %x %x %x\n", buf[0], buf[1], buf[2], buf[3]);
+#endif
+
+    /* no supplementary picture */
+    if (buf_size == 0) {
+        *data_size = 0;
+        return 0;
+    }
+
+    init_get_bits(&s->gb, buf, buf_size);
+
+    /* let's go :-) */
+    if (s->h263_msmpeg4) {
+        ret = msmpeg4_decode_picture_header(s);
+    } else if (s->h263_pred) {
+        ret = mpeg4_decode_picture_header(s);
+    } else if (s->h263_intel) {
+        ret = intel_h263_decode_picture_header(s);
+    } else {
+        ret = h263_decode_picture_header(s);
+    }
+    if (ret < 0)
+        return -1;
+
+    MPV_frame_start(s);
+
+#ifdef DEBUG
+    printf("qscale=%d\n", s->qscale);
+#endif
+
+    /* decode each macroblock */
+    for(s->mb_y=0; s->mb_y < s->mb_height; s->mb_y++) {
+        for(s->mb_x=0; s->mb_x < s->mb_width; s->mb_x++) {
+#ifdef DEBUG
+            printf("**mb x=%d y=%d\n", s->mb_x, s->mb_y);
+#endif
+            /* DCT & quantize */
+            if (s->h263_msmpeg4) {
+                msmpeg4_dc_scale(s);
+            } else if (s->h263_pred) {
+                h263_dc_scale(s);
+            } else {
+                /* default quantization values */
+                s->y_dc_scale = 8;
+                s->c_dc_scale = 8;
+            }
+
+            memset(block, 0, sizeof(block));
+            s->mv_dir = MV_DIR_FORWARD;
+            s->mv_type = MV_TYPE_16X16; 
+            if (s->h263_msmpeg4) {
+                if (msmpeg4_decode_mb(s, block) < 0)
+                    return -1;
+            } else {
+                if (h263_decode_mb(s, block) < 0)
+                    return -1;
+            }
+            MPV_decode_mb(s, block);
+        }
+    }
+
+    MPV_frame_end(s);
+    
+    pict->data[0] = s->current_picture[0];
+    pict->data[1] = s->current_picture[1];
+    pict->data[2] = s->current_picture[2];
+    pict->linesize[0] = s->linesize;
+    pict->linesize[1] = s->linesize / 2;
+    pict->linesize[2] = s->linesize / 2;
+
+    avctx->quality = s->qscale;
+    *data_size = sizeof(AVPicture);
+    return buf_size;
+}
+
+AVCodec opendivx_decoder = {
+    "opendivx",
+    CODEC_TYPE_VIDEO,
+    CODEC_ID_OPENDIVX,
+    sizeof(MpegEncContext),
+    h263_decode_init,
+    NULL,
+    h263_decode_end,
+    h263_decode_frame,
+};
+
+AVCodec h263_decoder = {
+    "h263",
+    CODEC_TYPE_VIDEO,
+    CODEC_ID_H263,
+    sizeof(MpegEncContext),
+    h263_decode_init,
+    NULL,
+    h263_decode_end,
+    h263_decode_frame,
+};
+
+AVCodec msmpeg4_decoder = {
+    "msmpeg4",
+    CODEC_TYPE_VIDEO,
+    CODEC_ID_MSMPEG4,
+    sizeof(MpegEncContext),
+    h263_decode_init,
+    NULL,
+    h263_decode_end,
+    h263_decode_frame,
+};
+
+AVCodec h263i_decoder = {
+    "h263i",
+    CODEC_TYPE_VIDEO,
+    CODEC_ID_H263I,
+    sizeof(MpegEncContext),
+    h263_decode_init,
+    NULL,
+    h263_decode_end,
+    h263_decode_frame,
+};
+
diff --git a/libavcodec/i386/cputest.c b/libavcodec/i386/cputest.c
new file mode 100644
index 0000000000..9181f413d5
--- /dev/null
+++ b/libavcodec/i386/cputest.c
@@ -0,0 +1,102 @@
+/* Cpu detection code, extracted from mmx.h ((c)1997-99 by H. Dietz
+   and R. Fisher). Converted to C and improved by Gerard Lantau */
+
+#include <stdlib.h>
+#include "../dsputil.h"
+
+#define cpuid(index,eax,ebx,ecx,edx) \
+    asm ("cpuid" \
+         : "=a" (eax), "=b" (ebx), \
+           "=c" (ecx), "=d" (edx) \
+         : "a" (index) \
+         : "cc")
+
+/* Function to test if multimedia instructions are supported...  */
+int mm_support(void)
+{
+    int rval;
+    int eax, ebx, ecx, edx;
+    
+    
+    __asm__ __volatile__ (
+                          /* See if CPUID instruction is supported ... */
+                          /* ... Get copies of EFLAGS into eax and ecx */
+                          "pushf\n\t"
+                          "popl %0\n\t"
+                          "movl %0, %1\n\t"
+                          
+                          /* ... Toggle the ID bit in one copy and store */
+                          /*     to the EFLAGS reg */
+                          "xorl $0x200000, %0\n\t"
+                          "push %0\n\t"
+                          "popf\n\t"
+                          
+                          /* ... Get the (hopefully modified) EFLAGS */
+                          "pushf\n\t"
+                          "popl %0\n\t"
+                          : "=a" (eax), "=c" (ecx)
+                          :
+                          : "cc" 
+                          );
+    
+    if (eax == ecx)
+        return 0; /* CPUID not supported */
+    
+    cpuid(0, eax, ebx, ecx, edx);
+
+    if (ebx == 0x756e6547 &&
+        edx == 0x49656e69 &&
+        ecx == 0x6c65746e) {
+        
+        /* intel */
+    inteltest:
+        cpuid(1, eax, ebx, ecx, edx);
+        if ((edx & 0x00800000) == 0)
+            return 0;
+        rval = MM_MMX;
+        if (edx & 0x02000000) 
+            rval |= MM_MMXEXT | MM_SSE;
+        if (edx & 0x04000000) 
+            rval |= MM_SSE2;
+        return rval;
+    } else if (ebx == 0x68747541 &&
+               edx == 0x69746e65 &&
+               ecx == 0x444d4163) {
+        /* AMD */
+        cpuid(0x80000000, eax, ebx, ecx, edx);
+        if ((unsigned)eax < 0x80000001)
+            goto inteltest;
+        cpuid(0x80000001, eax, ebx, ecx, edx);
+        if ((edx & 0x00800000) == 0)
+            return 0;
+        rval = MM_MMX;
+        if (edx & 0x80000000)
+            rval |= MM_3DNOW;
+        if (edx & 0x00400000)
+            rval |= MM_MMXEXT;
+        return rval;
+    } else if (ebx == 0x69727943 &&
+               edx == 0x736e4978 &&
+               ecx == 0x64616574) {
+        /* Cyrix Section */
+        /* See if extended CPUID level 80000001 is supported */
+        /* The value of CPUID/80000001 for the 6x86MX is undefined
+           according to the Cyrix CPU Detection Guide (Preliminary
+           Rev. 1.01 table 1), so we'll check the value of eax for
+           CPUID/0 to see if standard CPUID level 2 is supported.
+           According to the table, the only CPU which supports level
+           2 is also the only one which supports extended CPUID levels.
+        */
+        if (eax != 2) 
+            goto inteltest;
+        cpuid(0x80000001, eax, ebx, ecx, edx);
+        if ((eax & 0x00800000) == 0)
+            return 0;
+        rval = MM_MMX;
+        if (eax & 0x01000000)
+            rval |= MM_MMXEXT;
+        return rval;
+    } else {
+        return 0;
+    }
+}
diff --git a/libavcodec/i386/dsputil_mmx.c b/libavcodec/i386/dsputil_mmx.c
new file mode 100644
index 0000000000..701b70283d
--- /dev/null
+++ b/libavcodec/i386/dsputil_mmx.c
@@ -0,0 +1,1061 @@
+/*
+ * MMX optimized DSP utils
+ * Copyright (c) 2000, 2001 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ *
+ * MMX optimization by Nick Kurshev <nickols_k@mail.ru>
+ */
+
+#include "../dsputil.h"
+
+int pix_abs16x16_mmx(UINT8 *blk1, UINT8 *blk2, int lx, int h);
+int pix_abs16x16_sse(UINT8 *blk1, UINT8 *blk2, int lx, int h);
+int pix_abs16x16_x2_mmx(UINT8 *blk1, UINT8 *blk2, int lx, int h);
+int pix_abs16x16_y2_mmx(UINT8 *blk1, UINT8 *blk2, int lx, int h);
+int pix_abs16x16_xy2_mmx(UINT8 *blk1, UINT8 *blk2, int lx, int h);
+
+/* pixel operations */
+static const unsigned short mm_wone[4] __attribute__ ((aligned(8))) = { 0x1, 0x1, 0x1, 0x1 };
+static const unsigned short mm_wtwo[4] __attribute__ ((aligned(8))) = { 0x2, 0x2, 0x2, 0x2 };
+
+/***********************************/
+/* 3Dnow specific */
+
+#define DEF(x) x ## _3dnow
+/* for Athlons PAVGUSB is prefered */
+#define PAVGB "pavgusb"
+
+#include "dsputil_mmx_avg.h"
+
+#undef DEF
+#undef PAVGB
+
+/***********************************/
+/* MMX2 specific */
+
+#define DEF(x) x ## _sse
+
+/* Introduced only in MMX2 set */
+#define PAVGB "pavgb"
+
+#include "dsputil_mmx_avg.h"
+
+#undef DEF
+#undef PAVGB
+
+/***********************************/
+/* standard MMX */
+
+static void get_pixels_mmx(DCTELEM *block, const UINT8 *pixels, int line_size)
+{
+    DCTELEM *p;
+    const UINT8 *pix;
+    int i;
+
+    /* read the pixels */
+    p = block;
+    pix = pixels;
+    __asm __volatile("pxor %%mm7, %%mm7":::"memory");
+    for(i=0;i<4;i++) {
+	__asm __volatile(
+		"movq	%1, %%mm0\n\t"
+		"movq	%2, %%mm1\n\t"
+		"movq	%%mm0, %%mm2\n\t"
+		"movq	%%mm1, %%mm3\n\t"
+		"punpcklbw %%mm7, %%mm0\n\t"
+		"punpckhbw %%mm7, %%mm2\n\t"
+		"punpcklbw %%mm7, %%mm1\n\t"
+		"punpckhbw %%mm7, %%mm3\n\t"
+		"movq	%%mm0, %0\n\t"
+		"movq	%%mm2, 8%0\n\t"
+		"movq	%%mm1, 16%0\n\t"
+		"movq	%%mm3, 24%0\n\t"
+		:"=m"(*p)
+		:"m"(*pix), "m"(*(pix+line_size))
+		:"memory");
+        pix += line_size*2;
+        p += 16;
+    }
+    emms();
+}
+
+static void put_pixels_clamped_mmx(const DCTELEM *block, UINT8 *pixels, int line_size)
+{
+    const DCTELEM *p;
+    UINT8 *pix;
+    int i;
+
+    /* read the pixels */
+    p = block;
+    pix = pixels;
+    for(i=0;i<2;i++) {
+	__asm __volatile(
+		"movq	%4, %%mm0\n\t"
+		"movq	8%4, %%mm1\n\t"
+		"movq	16%4, %%mm2\n\t"
+		"movq	24%4, %%mm3\n\t"
+		"movq	32%4, %%mm4\n\t"
+		"movq	40%4, %%mm5\n\t"
+		"movq	48%4, %%mm6\n\t"
+		"movq	56%4, %%mm7\n\t"
+		"packuswb %%mm1, %%mm0\n\t"
+		"packuswb %%mm3, %%mm2\n\t"
+		"packuswb %%mm5, %%mm4\n\t"
+		"packuswb %%mm7, %%mm6\n\t"
+		"movq	%%mm0, %0\n\t"
+		"movq	%%mm2, %1\n\t"
+		"movq	%%mm4, %2\n\t"
+		"movq	%%mm6, %3\n\t"
+		:"=m"(*pix), "=m"(*(pix+line_size))
+		,"=m"(*(pix+line_size*2)), "=m"(*(pix+line_size*3))
+		:"m"(*p)
+		:"memory");
+        pix += line_size*4;
+        p += 32;
+    }
+    emms();
+}
+
+static void add_pixels_clamped_mmx(const DCTELEM *block, UINT8 *pixels, int line_size)
+{
+    const DCTELEM *p;
+    UINT8 *pix;
+    int i;
+
+    /* read the pixels */
+    p = block;
+    pix = pixels;
+	__asm __volatile("pxor	%%mm7, %%mm7":::"memory");
+    for(i=0;i<4;i++) {
+	__asm __volatile(
+		"movq	%2, %%mm0\n\t"
+		"movq	8%2, %%mm1\n\t"
+		"movq	16%2, %%mm2\n\t"
+		"movq	24%2, %%mm3\n\t"
+		"movq	%0, %%mm4\n\t"
+		"movq	%1, %%mm6\n\t"
+		"movq	%%mm4, %%mm5\n\t"
+		"punpcklbw %%mm7, %%mm4\n\t"
+		"punpckhbw %%mm7, %%mm5\n\t"
+		"paddsw	%%mm4, %%mm0\n\t"
+		"paddsw	%%mm5, %%mm1\n\t"
+		"movq	%%mm6, %%mm5\n\t"
+		"punpcklbw %%mm7, %%mm6\n\t"
+		"punpckhbw %%mm7, %%mm5\n\t"
+		"paddsw	%%mm6, %%mm2\n\t"
+		"paddsw	%%mm5, %%mm3\n\t"
+		"packuswb %%mm1, %%mm0\n\t"
+		"packuswb %%mm3, %%mm2\n\t"
+		"movq	%%mm0, %0\n\t"
+		"movq	%%mm2, %1\n\t"
+		:"=m"(*pix), "=m"(*(pix+line_size))
+		:"m"(*p)
+		:"memory");
+        pix += line_size*2;
+        p += 16;
+    }
+    emms();
+}
+
+static void put_pixels_mmx(UINT8 *block, const UINT8 *pixels, int line_size, int h)
+{
+    int dh, hh;
+    UINT8 *p;
+    const UINT8 *pix;
+    p   = block;
+    pix = pixels;
+    hh=h>>2;
+    dh=h&3;
+    while(hh--) {
+    __asm __volatile(
+	"movq	%4, %%mm0\n\t"
+	"movq	%5, %%mm1\n\t"
+	"movq	%6, %%mm2\n\t"
+	"movq	%7, %%mm3\n\t"
+	"movq	%%mm0, %0\n\t"
+	"movq	%%mm1, %1\n\t"
+	"movq	%%mm2, %2\n\t"
+	"movq	%%mm3, %3\n\t"
+	:"=m"(*p), "=m"(*(p+line_size)), "=m"(*(p+line_size*2)), "=m"(*(p+line_size*3))
+	:"m"(*pix), "m"(*(pix+line_size)), "m"(*(pix+line_size*2)), "m"(*(pix+line_size*3))
+	:"memory");
+        pix = pix + line_size*4;
+        p =   p   + line_size*4;
+    }
+    while(dh--) {
+     __asm __volatile(
+	"movq	%1, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix)
+	:"memory");
+        pix = pix + line_size;
+        p =   p   + line_size;
+    }
+    emms();
+}
+
+static void put_pixels_x2_mmx(UINT8 *block, const UINT8 *pixels, int line_size, int h)
+{
+  UINT8 *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile(
+	"pxor	%%mm7, %%mm7\n\t"
+	"movq	%0, %%mm4\n\t"
+	::"m"(mm_wone[0]):"memory");
+  do {
+    __asm __volatile(
+	"movq	%1, %%mm0\n\t"
+	"movq	1%1, %%mm1\n\t"
+	"movq	%%mm0, %%mm2\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm0\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"paddusw %%mm4, %%mm0\n\t"
+	"paddusw %%mm4, %%mm2\n\t"
+	"psrlw	$1, %%mm0\n\t"
+	"psrlw	$1, %%mm2\n\t"
+	"packuswb  %%mm2, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix)
+	:"memory");
+   pix += line_size; p += line_size;
+  } while (--h);
+  emms();
+}
+
+static void put_pixels_y2_mmx(UINT8 *block, const UINT8 *pixels, int line_size, int h)
+{
+  UINT8 *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile(
+	"pxor	%%mm7, %%mm7\n\t"
+	"movq	%0, %%mm4\n\t"
+	::"m"(mm_wone[0]):"memory");
+  do {
+    __asm __volatile(
+	"movq	%1, %%mm0\n\t"
+	"movq	%2, %%mm1\n\t"
+	"movq	%%mm0, %%mm2\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm0\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"paddusw %%mm4, %%mm0\n\t"
+	"paddusw %%mm4, %%mm2\n\t"
+	"psrlw	$1, %%mm0\n\t"
+	"psrlw	$1, %%mm2\n\t"
+	"packuswb  %%mm2, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix),
+	 "m"(*(pix+line_size))
+	:"memory");
+   pix += line_size;
+   p += line_size;
+  } while (--h);
+  emms();
+}
+
+static void put_pixels_xy2_mmx(UINT8 *block, const UINT8 *pixels, int line_size, int h)
+{
+  UINT8 *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile(
+	"pxor	%%mm7, %%mm7\n\t"
+	"movq	%0, %%mm6\n\t"
+	::"m"(mm_wtwo[0]):"memory");
+  do {
+    __asm __volatile(
+	"movq	%1, %%mm0\n\t"
+	"movq	%2, %%mm1\n\t"
+	"movq	1%1, %%mm4\n\t"
+	"movq	1%2, %%mm5\n\t"
+	"movq	%%mm0, %%mm2\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm0\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"movq	%%mm4, %%mm1\n\t"
+	"movq	%%mm5, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm4\n\t"
+	"punpcklbw %%mm7, %%mm5\n\t"
+	"punpckhbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm5, %%mm4\n\t"
+	"paddusw %%mm3, %%mm1\n\t"
+	"paddusw %%mm6, %%mm4\n\t"
+	"paddusw %%mm6, %%mm1\n\t"
+	"paddusw %%mm4, %%mm0\n\t"
+	"paddusw %%mm1, %%mm2\n\t"
+	"psrlw	$2, %%mm0\n\t"
+	"psrlw	$2, %%mm2\n\t"
+	"packuswb  %%mm2, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix),
+	 "m"(*(pix+line_size))
+	:"memory");
+   pix += line_size;
+   p += line_size;
+  } while(--h);
+  emms();
+}
+
+static void   put_no_rnd_pixels_x2_mmx( UINT8  *block, const UINT8 *pixels, int line_size, int h)
+{
+  UINT8  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile("pxor %%mm7, %%mm7\n\t":::"memory");
+  do {
+    __asm __volatile(
+	"movq	%1, %%mm0\n\t"
+	"movq	1%1, %%mm1\n\t"
+	"movq	%%mm0, %%mm2\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm0\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"psrlw	$1, %%mm0\n\t"
+	"psrlw	$1, %%mm2\n\t"
+	"packuswb  %%mm2, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix)
+	:"memory");
+   pix += line_size;
+   p +=   line_size;
+  } while (--h);
+  emms();
+}
+
+static void put_no_rnd_pixels_y2_mmx( UINT8  *block, const UINT8 *pixels, int line_size, int h)
+{
+  UINT8  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile("pxor %%mm7, %%mm7\n\t":::"memory");
+  do {
+    __asm __volatile(
+	"movq	%1, %%mm0\n\t"
+	"movq	%2, %%mm1\n\t"
+	"movq	%%mm0, %%mm2\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm0\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"psrlw	$1, %%mm0\n\t"
+	"psrlw	$1, %%mm2\n\t"
+	"packuswb  %%mm2, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix),
+	 "m"(*(pix+line_size))
+	:"memory");
+   pix += line_size;
+   p +=   line_size;
+  } while(--h);
+  emms();
+}
+
+static void   put_no_rnd_pixels_xy2_mmx( UINT8  *block, const UINT8 *pixels, int line_size, int h)
+{
+  UINT8  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile(
+	"pxor	%%mm7, %%mm7\n\t"
+	"movq	%0, %%mm6\n\t"
+	::"m"(mm_wone[0]):"memory");
+  do {
+    __asm __volatile(
+	"movq	%1, %%mm0\n\t"
+	"movq	%2, %%mm1\n\t"
+	"movq	1%1, %%mm4\n\t"
+	"movq	1%2, %%mm5\n\t"
+	"movq	%%mm0, %%mm2\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm0\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"movq	%%mm4, %%mm1\n\t"
+	"movq	%%mm5, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm4\n\t"
+	"punpcklbw %%mm7, %%mm5\n\t"
+	"punpckhbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm5, %%mm4\n\t"
+	"paddusw %%mm3, %%mm1\n\t"
+	"paddusw %%mm6, %%mm4\n\t"
+	"paddusw %%mm6, %%mm1\n\t"
+	"paddusw %%mm4, %%mm0\n\t"
+	"paddusw %%mm1, %%mm2\n\t"
+	"psrlw	$2, %%mm0\n\t"
+	"psrlw	$2, %%mm2\n\t"
+	"packuswb  %%mm2, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix),
+	 "m"(*(pix+line_size))
+	:"memory");
+   pix += line_size;
+   p +=   line_size;
+  } while(--h);
+  emms();
+}
+
+static void avg_pixels_mmx(UINT8 *block, const UINT8 *pixels, int line_size, int h)
+{
+  UINT8  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile(
+	"pxor	%%mm7, %%mm7\n\t"
+	"movq	%0, %%mm6\n\t"
+	::"m"(mm_wone[0]):"memory");
+  do {
+    __asm __volatile(
+	"movq	%0, %%mm0\n\t"
+	"movq	%1, %%mm1\n\t"
+	"movq	%%mm0, %%mm2\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm0\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"paddusw %%mm6, %%mm0\n\t"
+	"paddusw %%mm6, %%mm2\n\t"
+	"psrlw	$1, %%mm0\n\t"
+	"psrlw	$1, %%mm2\n\t"
+	"packuswb  %%mm2, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix)
+	:"memory");
+   pix += line_size;
+   p +=   line_size;
+  }
+  while (--h);
+  emms();
+}
+
+static void   avg_pixels_x2_mmx( UINT8  *block, const UINT8 *pixels, int line_size, int h)
+{
+  UINT8  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile(
+	"pxor	%%mm7, %%mm7\n\t"
+	"movq	%0, %%mm6\n\t"
+	::"m"(mm_wone[0]):"memory");
+  do {
+    __asm __volatile(
+	"movq	%1, %%mm1\n\t"
+	"movq	%0, %%mm0\n\t"
+	"movq	1%1, %%mm4\n\t"
+	"movq	%%mm0, %%mm2\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"movq	%%mm4, %%mm5\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm4\n\t"
+	"punpckhbw %%mm7, %%mm5\n\t"
+	"punpcklbw %%mm7, %%mm0\n\t"
+	"punpckhbw %%mm7, %%mm2\n\t"
+	"paddusw %%mm4, %%mm1\n\t"
+	"paddusw %%mm5, %%mm3\n\t"
+	"paddusw %%mm6, %%mm1\n\t"
+	"paddusw %%mm6, %%mm3\n\t"
+	"psrlw	$1, %%mm1\n\t"
+	"psrlw	$1, %%mm3\n\t"
+	"paddusw %%mm6, %%mm0\n\t"
+	"paddusw %%mm6, %%mm2\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"psrlw	$1, %%mm0\n\t"
+	"psrlw	$1, %%mm2\n\t"
+	"packuswb  %%mm2, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix)
+	:"memory");
+   pix += line_size;
+   p +=   line_size;
+  } while (--h);
+  emms();
+}
+
+static void   avg_pixels_y2_mmx( UINT8  *block, const UINT8 *pixels, int line_size, int h)
+{
+  UINT8  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile(
+	"pxor	%%mm7, %%mm7\n\t"
+	"movq	%0, %%mm6\n\t"
+	::"m"(mm_wone[0]):"memory");
+  do {
+    __asm __volatile(
+	"movq	%1, %%mm1\n\t"
+	"movq	%0, %%mm0\n\t"
+	"movq	%2, %%mm4\n\t"
+	"movq	%%mm0, %%mm2\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"movq	%%mm4, %%mm5\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm4\n\t"
+	"punpckhbw %%mm7, %%mm5\n\t"
+	"punpcklbw %%mm7, %%mm0\n\t"
+	"punpckhbw %%mm7, %%mm2\n\t"
+	"paddusw %%mm4, %%mm1\n\t"
+	"paddusw %%mm5, %%mm3\n\t"
+	"paddusw %%mm6, %%mm1\n\t"
+	"paddusw %%mm6, %%mm3\n\t"
+	"psrlw	$1, %%mm1\n\t"
+	"psrlw	$1, %%mm3\n\t"
+	"paddusw %%mm6, %%mm0\n\t"
+	"paddusw %%mm6, %%mm2\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"psrlw	$1, %%mm0\n\t"
+	"psrlw	$1, %%mm2\n\t"
+	"packuswb  %%mm2, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix), "m"(*(pix+line_size))
+	:"memory");
+   pix += line_size;
+   p +=   line_size ;
+  } while(--h);
+  emms();
+}
+
+static void   avg_pixels_xy2_mmx( UINT8  *block, const UINT8 *pixels, int line_size, int h)
+{
+  UINT8  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile(
+	"pxor	%%mm7, %%mm7\n\t"
+	"movq	%0, %%mm6\n\t"
+	::"m"(mm_wtwo[0]):"memory");
+  do {
+    __asm __volatile(
+	"movq	%1, %%mm0\n\t"
+	"movq	%2, %%mm1\n\t"
+	"movq	1%1, %%mm4\n\t"
+	"movq	1%2, %%mm5\n\t"
+	"movq	%%mm0, %%mm2\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm0\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"movq	%%mm4, %%mm1\n\t"
+	"movq	%%mm5, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm4\n\t"
+	"punpcklbw %%mm7, %%mm5\n\t"
+	"punpckhbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm5, %%mm4\n\t"
+	"paddusw %%mm3, %%mm1\n\t"
+	"paddusw %%mm6, %%mm4\n\t"
+	"paddusw %%mm6, %%mm1\n\t"
+	"paddusw %%mm4, %%mm0\n\t"
+	"paddusw %%mm1, %%mm2\n\t"
+	"movq	%3, %%mm5\n\t"
+	"psrlw	$2, %%mm0\n\t"
+	"movq	%0, %%mm1\n\t"
+	"psrlw	$2, %%mm2\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"paddusw %%mm5, %%mm0\n\t"
+	"paddusw %%mm5, %%mm2\n\t"
+	"psrlw	$1, %%mm0\n\t"
+	"psrlw	$1, %%mm2\n\t"
+	"packuswb  %%mm2, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix),
+	 "m"(*(pix+line_size)), "m"(mm_wone[0])
+	:"memory");
+   pix += line_size;
+   p +=   line_size ;
+  } while(--h);
+  emms();
+}
+
+static void avg_no_rnd_pixels_mmx( UINT8  *block, const UINT8 *pixels, int line_size, int h)
+{
+  UINT8  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile("pxor %%mm7, %%mm7\n\t":::"memory");
+  do {
+    __asm __volatile(
+	"movq	%1, %%mm0\n\t"
+	"movq	%0, %%mm1\n\t"
+	"movq	%%mm0, %%mm2\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm0\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"psrlw	$1, %%mm0\n\t"
+	"psrlw	$1, %%mm2\n\t"
+	"packuswb  %%mm2, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix)
+	:"memory");
+   pix += line_size;
+   p +=   line_size ;
+  } while (--h);
+  emms();
+}
+
+static void   avg_no_rnd_pixels_x2_mmx( UINT8  *block, const UINT8 *pixels, int line_size, int h)
+{
+  UINT8  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile(
+      "pxor	%%mm7, %%mm7\n\t":::"memory");
+  do {
+    __asm __volatile(
+	"movq	%1, %%mm0\n\t"
+	"movq	1%1, %%mm1\n\t"
+	"movq	%0, %%mm4\n\t"
+	"movq	%%mm0, %%mm2\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"movq	%%mm4, %%mm5\n\t"
+	"punpcklbw %%mm7, %%mm0\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm4\n\t"
+	"punpckhbw %%mm7, %%mm5\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"psrlw	$1, %%mm0\n\t"
+	"psrlw	$1, %%mm2\n\t"
+	"paddusw %%mm4, %%mm0\n\t"
+	"paddusw %%mm5, %%mm2\n\t"
+	"psrlw	$1, %%mm0\n\t"
+	"psrlw	$1, %%mm2\n\t"
+	"packuswb  %%mm2, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix)
+	:"memory");
+   pix += line_size;
+   p +=   line_size;
+ } while (--h);
+ emms();
+}
+
+static void   avg_no_rnd_pixels_y2_mmx( UINT8  *block, const UINT8 *pixels, int line_size, int h)
+{
+  UINT8  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile(
+      "pxor	%%mm7, %%mm7\n\t":::"memory");
+  do {
+    __asm __volatile(
+	"movq	%1, %%mm0\n\t"
+	"movq	%2, %%mm1\n\t"
+	"movq	%0, %%mm4\n\t"
+	"movq	%%mm0, %%mm2\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"movq	%%mm4, %%mm5\n\t"
+	"punpcklbw %%mm7, %%mm0\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm4\n\t"
+	"punpckhbw %%mm7, %%mm5\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"psrlw	$1, %%mm0\n\t"
+	"psrlw	$1, %%mm2\n\t"
+	"paddusw %%mm4, %%mm0\n\t"
+	"paddusw %%mm5, %%mm2\n\t"
+	"psrlw	$1, %%mm0\n\t"
+	"psrlw	$1, %%mm2\n\t"
+	"packuswb  %%mm2, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix), "m"(*(pix+line_size))
+	:"memory");
+   pix += line_size;
+   p +=   line_size ;
+  } while(--h);
+  emms();
+}
+
+static void   avg_no_rnd_pixels_xy2_mmx( UINT8  *block, const UINT8 *pixels, int line_size, int h)
+{
+  UINT8  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile(
+	"pxor	%%mm7, %%mm7\n\t"
+	"movq	%0, %%mm6\n\t"
+	::"m"(mm_wone[0]):"memory");
+  do {
+    __asm __volatile(
+	"movq	%1, %%mm0\n\t"
+	"movq	%2, %%mm1\n\t"
+	"movq	1%1, %%mm4\n\t"
+	"movq	1%2, %%mm5\n\t"
+	"movq	%%mm0, %%mm2\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm0\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"movq	%%mm4, %%mm1\n\t"
+	"movq	%%mm5, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm4\n\t"
+	"punpcklbw %%mm7, %%mm5\n\t"
+	"punpckhbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm5, %%mm4\n\t"
+	"paddusw %%mm3, %%mm1\n\t"
+	"paddusw %%mm6, %%mm4\n\t"
+	"paddusw %%mm6, %%mm1\n\t"
+	"paddusw %%mm4, %%mm0\n\t"
+	"paddusw %%mm1, %%mm2\n\t"
+	"movq	%0, %%mm1\n\t"
+	"psrlw	$2, %%mm0\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"psrlw	$2, %%mm2\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"psrlw	$1, %%mm0\n\t"
+	"psrlw	$1, %%mm2\n\t"
+	"packuswb  %%mm2, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix),
+	 "m"(*(pix+line_size))
+	:"memory");
+   pix += line_size;
+   p += line_size;
+  } while(--h);
+  emms();
+}
+
+static void sub_pixels_mmx( DCTELEM  *block, const UINT8 *pixels, int line_size, int h)
+{
+  DCTELEM  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile("pxor %%mm7, %%mm7":::"memory");
+  do {
+    __asm __volatile(
+	"movq	%0, %%mm0\n\t"
+	"movq	%1, %%mm2\n\t"
+	"movq	8%0, %%mm1\n\t"
+	"movq	%%mm2, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"psubsw %%mm2, %%mm0\n\t"
+	"psubsw %%mm3, %%mm1\n\t"
+	"movq	%%mm0, %0\n\t"
+	"movq	%%mm1, 8%0\n\t"
+	:"=m"(*p)
+	:"m"(*pix)
+	:"memory");
+   pix += line_size;
+   p +=   8;
+  } while (--h);
+  emms();
+}
+
+static void sub_pixels_x2_mmx( DCTELEM  *block, const UINT8 *pixels, int line_size, int h)
+{
+  DCTELEM  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile(
+      "pxor	%%mm7, %%mm7\n\t"
+      "movq	%0, %%mm6"
+      ::"m"(mm_wone[0]):"memory");
+  do {
+    __asm __volatile(
+	"movq	%0, %%mm0\n\t"
+	"movq	%1, %%mm2\n\t"
+	"movq	8%0, %%mm1\n\t"
+	"movq	1%1, %%mm4\n\t"
+	"movq	%%mm2, %%mm3\n\t"
+	"movq	%%mm4, %%mm5\n\t"
+	"punpcklbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm4\n\t"
+	"punpckhbw %%mm7, %%mm5\n\t"
+	"paddusw %%mm4, %%mm2\n\t"
+	"paddusw %%mm5, %%mm3\n\t"
+	"paddusw %%mm6, %%mm2\n\t"
+	"paddusw %%mm6, %%mm3\n\t"
+	"psrlw	$1, %%mm2\n\t"
+	"psrlw	$1, %%mm3\n\t"
+	"psubsw %%mm2, %%mm0\n\t"
+	"psubsw %%mm3, %%mm1\n\t"
+	"movq	%%mm0, %0\n\t"
+	"movq	%%mm1, 8%0\n\t"
+	:"=m"(*p)
+	:"m"(*pix)
+	:"memory");
+   pix += line_size;
+   p +=   8;
+ } while (--h);
+ emms();
+}
+
+static void sub_pixels_y2_mmx( DCTELEM  *block, const UINT8 *pixels, int line_size, int h)
+{
+  DCTELEM  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile(
+      "pxor	%%mm7, %%mm7\n\t"
+      "movq	%0, %%mm6"
+      ::"m"(mm_wone[0]):"memory");
+  do {
+    __asm __volatile(
+	"movq	%0, %%mm0\n\t"
+	"movq	%1, %%mm2\n\t"
+	"movq	8%0, %%mm1\n\t"
+	"movq	%2, %%mm4\n\t"
+	"movq	%%mm2, %%mm3\n\t"
+	"movq	%%mm4, %%mm5\n\t"
+	"punpcklbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm4\n\t"
+	"punpckhbw %%mm7, %%mm5\n\t"
+	"paddusw %%mm4, %%mm2\n\t"
+	"paddusw %%mm5, %%mm3\n\t"
+	"paddusw %%mm6, %%mm2\n\t"
+	"paddusw %%mm6, %%mm3\n\t"
+	"psrlw	$1, %%mm2\n\t"
+	"psrlw	$1, %%mm3\n\t"
+	"psubsw %%mm2, %%mm0\n\t"
+	"psubsw %%mm3, %%mm1\n\t"
+	"movq	%%mm0, %0\n\t"
+	"movq	%%mm1, 8%0\n\t"
+	:"=m"(*p)
+	:"m"(*pix), "m"(*(pix+line_size))
+	:"memory");
+   pix += line_size;
+   p +=   8;
+ } while (--h);
+ emms();
+}
+
+static void   sub_pixels_xy2_mmx( DCTELEM  *block, const UINT8 *pixels, int line_size, int h)
+{
+  DCTELEM  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile(
+	"pxor	%%mm7, %%mm7\n\t"
+	"movq	%0, %%mm6\n\t"
+	::"m"(mm_wtwo[0]):"memory");
+  do {
+    __asm __volatile(
+	"movq	%1, %%mm0\n\t"
+	"movq	%2, %%mm1\n\t"
+	"movq	1%1, %%mm4\n\t"
+	"movq	1%2, %%mm5\n\t"
+	"movq	%%mm0, %%mm2\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm0\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"movq	%%mm4, %%mm1\n\t"
+	"movq	%%mm5, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm4\n\t"
+	"punpcklbw %%mm7, %%mm5\n\t"
+	"punpckhbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm5, %%mm4\n\t"
+	"paddusw %%mm3, %%mm1\n\t"
+	"paddusw %%mm6, %%mm4\n\t"
+	"paddusw %%mm6, %%mm1\n\t"
+	"paddusw %%mm4, %%mm0\n\t"
+	"paddusw %%mm1, %%mm2\n\t"
+	"movq	%0, %%mm1\n\t"
+	"movq	8%0, %%mm3\n\t"
+	"psrlw	$2, %%mm0\n\t"
+	"psrlw	$2, %%mm2\n\t"
+	"psubsw %%mm0, %%mm1\n\t"
+	"psubsw %%mm2, %%mm3\n\t"
+	"movq	%%mm1, %0\n\t"
+	"movq	%%mm3, 8%0\n\t"
+	:"=m"(*p)
+	:"m"(*pix),
+	 "m"(*(pix+line_size))
+	:"memory");
+   pix += line_size;
+   p +=   8 ;
+  } while(--h);
+  emms();
+}
+
+void dsputil_init_mmx(void)
+{
+    mm_flags = mm_support();
+#if 0
+    printf("CPU flags:");
+    if (mm_flags & MM_MMX)
+        printf(" mmx");
+    if (mm_flags & MM_MMXEXT)
+        printf(" mmxext");
+    if (mm_flags & MM_3DNOW)
+        printf(" 3dnow");
+    if (mm_flags & MM_SSE)
+        printf(" sse");
+    if (mm_flags & MM_SSE2)
+        printf(" sse2");
+    printf("\n");
+#endif
+
+    if (mm_flags & MM_MMX) {
+        get_pixels = get_pixels_mmx;
+        put_pixels_clamped = put_pixels_clamped_mmx;
+        add_pixels_clamped = add_pixels_clamped_mmx;
+        
+        pix_abs16x16 = pix_abs16x16_mmx;
+        pix_abs16x16_x2 = pix_abs16x16_x2_mmx;
+        pix_abs16x16_y2 = pix_abs16x16_y2_mmx;
+        pix_abs16x16_xy2 = pix_abs16x16_xy2_mmx;
+        av_fdct = fdct_mmx;
+
+        put_pixels_tab[0] = put_pixels_mmx;
+        put_pixels_tab[1] = put_pixels_x2_mmx;
+        put_pixels_tab[2] = put_pixels_y2_mmx;
+        put_pixels_tab[3] = put_pixels_xy2_mmx;
+
+        put_no_rnd_pixels_tab[0] = put_pixels_mmx;
+        put_no_rnd_pixels_tab[1] = put_no_rnd_pixels_x2_mmx;
+        put_no_rnd_pixels_tab[2] = put_no_rnd_pixels_y2_mmx;
+        put_no_rnd_pixels_tab[3] = put_no_rnd_pixels_xy2_mmx;
+        
+        avg_pixels_tab[0] = avg_pixels_mmx;
+        avg_pixels_tab[1] = avg_pixels_x2_mmx;
+        avg_pixels_tab[2] = avg_pixels_y2_mmx;
+        avg_pixels_tab[3] = avg_pixels_xy2_mmx;
+
+        avg_no_rnd_pixels_tab[0] = avg_no_rnd_pixels_mmx;
+        avg_no_rnd_pixels_tab[1] = avg_no_rnd_pixels_x2_mmx;
+        avg_no_rnd_pixels_tab[2] = avg_no_rnd_pixels_y2_mmx;
+        avg_no_rnd_pixels_tab[3] = avg_no_rnd_pixels_xy2_mmx;
+        
+        sub_pixels_tab[0] = sub_pixels_mmx;
+        sub_pixels_tab[1] = sub_pixels_x2_mmx;
+        sub_pixels_tab[2] = sub_pixels_y2_mmx;
+        sub_pixels_tab[3] = sub_pixels_xy2_mmx;
+
+        if (mm_flags & MM_MMXEXT) {
+            pix_abs16x16 = pix_abs16x16_sse;
+        }
+
+        if (mm_flags & MM_SSE) {
+            put_pixels_tab[1] = put_pixels_x2_sse;
+            put_pixels_tab[2] = put_pixels_y2_sse;
+            
+            avg_pixels_tab[0] = avg_pixels_sse;
+            avg_pixels_tab[1] = avg_pixels_x2_sse;
+            avg_pixels_tab[2] = avg_pixels_y2_sse;
+            avg_pixels_tab[3] = avg_pixels_xy2_sse;
+
+            sub_pixels_tab[1] = sub_pixels_x2_sse;
+            sub_pixels_tab[2] = sub_pixels_y2_sse;
+        } else if (mm_flags & MM_3DNOW) {
+            put_pixels_tab[1] = put_pixels_x2_3dnow;
+            put_pixels_tab[2] = put_pixels_y2_3dnow;
+            
+            avg_pixels_tab[0] = avg_pixels_3dnow;
+            avg_pixels_tab[1] = avg_pixels_x2_3dnow;
+            avg_pixels_tab[2] = avg_pixels_y2_3dnow;
+            avg_pixels_tab[3] = avg_pixels_xy2_3dnow;
+
+            sub_pixels_tab[1] = sub_pixels_x2_3dnow;
+            sub_pixels_tab[2] = sub_pixels_y2_3dnow;
+        }
+    }
+}
diff --git a/libavcodec/i386/dsputil_mmx_avg.h b/libavcodec/i386/dsputil_mmx_avg.h
new file mode 100644
index 0000000000..e47b83f62d
--- /dev/null
+++ b/libavcodec/i386/dsputil_mmx_avg.h
@@ -0,0 +1,352 @@
+/*
+ * DSP utils : average functions are compiled twice for 3dnow/mmx2
+ * Copyright (c) 2000, 2001 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ *
+ * MMX optimization by Nick Kurshev <nickols_k@mail.ru>
+ */
+
+static void DEF(put_pixels_x2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)
+{
+  int dh, hh;
+  UINT8 *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  hh=h>>2;
+  dh=h&3;
+  while(hh--) {
+    __asm __volatile(
+	"movq	%4, %%mm0\n\t"
+	"movq	1%4, %%mm1\n\t"
+	"movq	%5, %%mm2\n\t"
+	"movq	1%5, %%mm3\n\t"
+	"movq	%6, %%mm4\n\t"
+	"movq	1%6, %%mm5\n\t"
+	"movq	%7, %%mm6\n\t"
+	"movq	1%7, %%mm7\n\t"
+	PAVGB"  %%mm1, %%mm0\n\t"
+	PAVGB"  %%mm3, %%mm2\n\t"
+	PAVGB"  %%mm5, %%mm4\n\t"
+	PAVGB"  %%mm7, %%mm6\n\t"
+	"movq	%%mm0, %0\n\t"
+	"movq	%%mm2, %1\n\t"
+	"movq	%%mm4, %2\n\t"
+	"movq	%%mm6, %3\n\t"
+	:"=m"(*p), "=m"(*(p+line_size)), "=m"(*(p+line_size*2)), "=m"(*(p+line_size*3))
+	:"m"(*pix), "m"(*(pix+line_size)), "m"(*(pix+line_size*2)), "m"(*(pix+line_size*3))
+	:"memory");
+     pix += line_size*4; p += line_size*4;
+  }
+  while(dh--) {
+    __asm __volatile(
+	"movq	%1, %%mm0\n\t"
+	"movq	1%1, %%mm1\n\t"
+	PAVGB"  %%mm1, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix)
+	:"memory");
+     pix += line_size; p += line_size;
+  }
+  emms();
+}
+
+static void DEF(put_pixels_y2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)
+{
+  int dh, hh;
+  UINT8 *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+
+  hh=h>>1;
+  dh=h&1;
+  while(hh--) {
+    __asm __volatile(
+	"movq	%2, %%mm0\n\t"
+	"movq	%3, %%mm1\n\t"
+	"movq	%4, %%mm2\n\t"
+	PAVGB"  %%mm1, %%mm0\n\t"
+	PAVGB"  %%mm2, %%mm1\n\t"
+	"movq	%%mm0, %0\n\t"
+	"movq	%%mm1, %1\n\t"
+	:"=m"(*p), "=m"(*(p+line_size))
+	:"m"(*pix), "m"(*(pix+line_size)),
+	 "m"(*(pix+line_size*2))
+	:"memory");
+     pix += line_size*2;
+     p += line_size*2;
+  }
+  if(dh) {
+    __asm __volatile(
+	"movq	%1, %%mm0\n\t"
+	"movq	%2, %%mm1\n\t"
+	PAVGB"  %%mm1, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix),
+	 "m"(*(pix+line_size))
+	:"memory");
+  }
+  emms();
+}
+
+static void DEF(avg_pixels)(UINT8 *block, const UINT8 *pixels, int line_size, int h)
+{
+  int dh, hh;
+  UINT8 *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  hh=h>>2;
+  dh=h&3;
+  while(hh--) {
+    __asm __volatile(
+	"movq	%0, %%mm0\n\t"
+	"movq	%4, %%mm1\n\t"
+	"movq	%1, %%mm2\n\t"
+	"movq	%5, %%mm3\n\t"
+	"movq	%2, %%mm4\n\t"
+	"movq	%6, %%mm5\n\t"
+	"movq	%3, %%mm6\n\t"
+	"movq	%7, %%mm7\n\t"
+	PAVGB"  %%mm1, %%mm0\n\t"
+	PAVGB"  %%mm3, %%mm2\n\t"
+	PAVGB"  %%mm5, %%mm4\n\t"
+	PAVGB"  %%mm7, %%mm6\n\t"
+	"movq	%%mm0, %0\n\t"
+	"movq	%%mm2, %1\n\t"
+	"movq	%%mm4, %2\n\t"
+	"movq	%%mm6, %3\n\t"
+	:"=m"(*p), "=m"(*(p+line_size)), "=m"(*(p+line_size*2)), "=m"(*(p+line_size*3))
+	:"m"(*pix), "m"(*(pix+line_size)), "m"(*(pix+line_size*2)), "m"(*(pix+line_size*3))
+	:"memory");
+     pix += line_size*4; p += line_size*4;
+  }
+  while(dh--) {
+    __asm __volatile(
+	"movq	%0, %%mm0\n\t"
+	"movq	%1, %%mm1\n\t"
+	PAVGB"  %%mm1, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix)
+	:"memory");
+     pix += line_size; p += line_size;
+  }
+  emms();
+}
+
+static void DEF(avg_pixels_x2)( UINT8  *block, const UINT8 *pixels, int line_size, int h)
+{
+  int dh, hh;
+  UINT8  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  hh=h>>1;
+  dh=h&1;
+  while(hh--) {
+    __asm __volatile(
+	"movq	%2, %%mm2\n\t"
+	"movq	1%2, %%mm3\n\t"
+	"movq	%3, %%mm4\n\t"
+	"movq	1%3, %%mm5\n\t"
+	"movq	%0, %%mm0\n\t"
+	"movq	%1, %%mm1\n\t"
+	PAVGB"	%%mm3, %%mm2\n\t"
+	PAVGB"	%%mm2, %%mm0\n\t"
+	PAVGB"	%%mm5, %%mm4\n\t"
+	PAVGB"	%%mm4, %%mm1\n\t"
+	"movq	%%mm0, %0\n\t"
+	"movq	%%mm1, %1\n\t"
+	:"=m"(*p), "=m"(*(p+line_size))
+	:"m"(*pix), "m"(*(pix+line_size))
+	:"memory");
+   pix += line_size*2;
+   p +=   line_size*2;
+  }
+  if(dh) {
+    __asm __volatile(
+	"movq	%1, %%mm1\n\t"
+	"movq	1%1, %%mm2\n\t"
+	"movq	%0, %%mm0\n\t"
+	PAVGB"	%%mm2, %%mm1\n\t"
+	PAVGB"	%%mm1, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix)
+	:"memory");
+  }
+  emms();
+}
+
+static void  DEF(avg_pixels_y2)( UINT8  *block, const UINT8 *pixels, int line_size, int h)
+{
+  int dh, hh;
+  UINT8  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  hh=h>>1;
+  dh=h&1;
+  while(hh--) {
+    __asm __volatile(
+	"movq	%2, %%mm2\n\t"
+	"movq	%3, %%mm3\n\t"
+	"movq	%3, %%mm4\n\t"
+	"movq	%4, %%mm5\n\t"
+	"movq	%0, %%mm0\n\t"
+	"movq	%1, %%mm1\n\t"
+	PAVGB"	%%mm3, %%mm2\n\t"
+	PAVGB"	%%mm2, %%mm0\n\t"
+	PAVGB"	%%mm5, %%mm4\n\t"
+	PAVGB"	%%mm4, %%mm1\n\t"
+	"movq	%%mm0, %0\n\t"
+	"movq	%%mm1, %1\n\t"
+	:"=m"(*p), "=m"(*(p+line_size))
+	:"m"(*pix), "m"(*(pix+line_size)), "m"(*(pix+line_size*2))
+	:"memory");
+   pix += line_size*2;
+   p +=   line_size*2;
+  }
+  if(dh) {
+    __asm __volatile(
+	"movq	%1, %%mm1\n\t"
+	"movq	%2, %%mm2\n\t"
+	"movq	%0, %%mm0\n\t"
+	PAVGB"	%%mm2, %%mm1\n\t"
+	PAVGB"	%%mm1, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix), "m"(*(pix+line_size))
+	:"memory");
+  }
+  emms();
+}
+
+static void DEF(avg_pixels_xy2)( UINT8  *block, const UINT8 *pixels, int line_size, int h)
+{
+  UINT8  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile(
+	"pxor	%%mm7, %%mm7\n\t"
+	"movq	%0, %%mm6\n\t"
+	::"m"(mm_wtwo[0]):"memory");
+  do {
+    __asm __volatile(
+	"movq	%1, %%mm0\n\t"
+	"movq	%2, %%mm1\n\t"
+	"movq	1%1, %%mm4\n\t"
+	"movq	1%2, %%mm5\n\t"
+	"movq	%%mm0, %%mm2\n\t"
+	"movq	%%mm1, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm0\n\t"
+	"punpcklbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm1, %%mm0\n\t"
+	"paddusw %%mm3, %%mm2\n\t"
+	"movq	%%mm4, %%mm1\n\t"
+	"movq	%%mm5, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm4\n\t"
+	"punpcklbw %%mm7, %%mm5\n\t"
+	"punpckhbw %%mm7, %%mm1\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"paddusw %%mm5, %%mm4\n\t"
+	"paddusw %%mm3, %%mm1\n\t"
+	"paddusw %%mm6, %%mm4\n\t"
+	"paddusw %%mm6, %%mm1\n\t"
+	"paddusw %%mm4, %%mm0\n\t"
+	"paddusw %%mm1, %%mm2\n\t"
+	"psrlw	$2, %%mm0\n\t"
+	"psrlw	$2, %%mm2\n\t"
+	"packuswb  %%mm2, %%mm0\n\t"
+	PAVGB"	%0, %%mm0\n\t"
+	"movq	%%mm0, %0\n\t"
+	:"=m"(*p)
+	:"m"(*pix),
+	 "m"(*(pix+line_size))
+	:"memory");
+   pix += line_size;
+   p +=   line_size ;
+  } while(--h);
+  emms();
+}
+
+static void DEF(sub_pixels_x2)( DCTELEM  *block, const UINT8 *pixels, int line_size, int h)
+{
+  DCTELEM  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile(
+      "pxor	%%mm7, %%mm7":::"memory");
+  do {
+    __asm __volatile(
+	"movq	1%1, %%mm2\n\t"
+	"movq	%0, %%mm0\n\t"
+	PAVGB"	%1, %%mm2\n\t"
+	"movq	8%0, %%mm1\n\t"
+	"movq	%%mm2, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"psubsw %%mm2, %%mm0\n\t"
+	"psubsw %%mm3, %%mm1\n\t"
+	"movq	%%mm0, %0\n\t"
+	"movq	%%mm1, 8%0\n\t"
+	:"=m"(*p)
+	:"m"(*pix)
+	:"memory");
+   pix += line_size;
+   p +=   8;
+ } while (--h);
+ emms();
+}
+
+static void DEF(sub_pixels_y2)( DCTELEM  *block, const UINT8 *pixels, int line_size, int h)
+{
+  DCTELEM  *p;
+  const UINT8 *pix;
+  p = block;
+  pix = pixels;
+  __asm __volatile(
+      "pxor	%%mm7, %%mm7":::"memory");
+  do {
+    __asm __volatile(
+	"movq	%2, %%mm2\n\t"
+	"movq	%0, %%mm0\n\t"
+	PAVGB"	%1, %%mm2\n\t"
+	"movq	8%0, %%mm1\n\t"
+	"movq	%%mm2, %%mm3\n\t"
+	"punpcklbw %%mm7, %%mm2\n\t"
+	"punpckhbw %%mm7, %%mm3\n\t"
+	"psubsw %%mm2, %%mm0\n\t"
+	"psubsw %%mm3, %%mm1\n\t"
+	"movq	%%mm0, %0\n\t"
+	"movq	%%mm1, 8%0\n\t"
+	:"=m"(*p)
+	:"m"(*pix), "m"(*(pix+line_size))
+	:"memory");
+   pix += line_size;
+   p +=   8;
+ } while (--h);
+ emms();
+}
+
diff --git a/libavcodec/i386/fdct_mmx.s b/libavcodec/i386/fdct_mmx.s
new file mode 100644
index 0000000000..75c67bfe5e
--- /dev/null
+++ b/libavcodec/i386/fdct_mmx.s
@@ -0,0 +1,507 @@
+; //////////////////////////////////////////////////////////////////////////////
+; //
+; //  fdctam32.c - AP922 MMX(3D-Now) forward-DCT
+; //  ----------
+; //  Intel Application Note AP-922 - fast, precise implementation of DCT
+; //        http://developer.intel.com/vtune/cbts/appnotes.htm
+; //  ----------
+; //  
+; //       This routine can use a 3D-Now/MMX enhancement to increase the
+; //  accuracy of the fdct_col_4 macro.  The dct_col function uses 3D-Now's
+; //  PMHULHRW instead of MMX's PMHULHW(and POR).  The substitution improves
+; //  accuracy very slightly with performance penalty.  If the target CPU
+; //  does not support 3D-Now, then this function cannot be executed.
+; //  
+; //  For a fast, precise MMX implementation of inverse-DCT 
+; //              visit http://www.elecard.com/peter
+; //
+; //  v1.0 07/22/2000 (initial release)
+; //     
+; //  liaor@iname.com  http://members.tripod.com/~liaor  
+; //////////////////////////////////////////////////////////////////////////////
+
+;;;
+;;; A.Stevens Jul 2000:	 ported to nasm syntax and disentangled from
+;;; from Win**** compiler specific stuff.
+;;; All the real work was done above though.
+;;; See above for how to optimise quality on 3DNow! CPU's
+
+		;;
+		;;		Macros for code-readability...
+		;; 
+%define INP eax		;	 pointer to (short *blk) 
+%define OUT ecx		;	 pointer to output (temporary store space qwTemp[])
+%define TABLE ebx	; pointer to tab_frw_01234567[]
+%define TABLEF ebx  ; pointer to tg_all_16
+%define round_frw_row edx
+
+
+%define x0 INP + 0*16
+%define x1 INP + 1*16
+%define x2 INP + 2*16
+%define x3 INP + 3*16
+%define x4 INP + 4*16
+%define x5 INP + 5*16
+%define x6 INP + 6*16
+%define x7 INP + 7*16
+%define y0 OUT + 0*16
+%define y1 OUT + 1*16
+%define y2 OUT + 2*16
+%define y3 OUT + 3*16
+%define y4 OUT + 4*16
+%define y5 OUT + 5*16
+%define y6 OUT + 6*16
+%define y7 OUT + 7*16
+				
+		;;
+		;; Constants for DCT
+		;;
+%define BITS_FRW_ACC	3 ; 2 or 3 for accuracy
+%define SHIFT_FRW_COL	BITS_FRW_ACC
+%define SHIFT_FRW_ROW	(BITS_FRW_ACC + 17)
+%define RND_FRW_ROW		(1 << (SHIFT_FRW_ROW-1))
+%define RND_FRW_COL		(1 << (SHIFT_FRW_COL-1))
+
+extern fdct_one_corr		
+extern fdct_r_row				;  Defined in C for convenience
+		;;
+		;; Concatenated table of forward dct transformation coeffs.
+		;; 
+extern  fdct_tg_all_16			; Defined in C for convenience
+		;; Offsets into table..
+		
+%define tg_1_16 (TABLEF + 0)
+%define tg_2_16 (TABLEF + 8)
+%define tg_3_16 (TABLEF + 16)
+%define cos_4_16 (TABLEF + 24)
+%define ocos_4_16 (TABLEF + 32)		
+
+		;;
+		;; Concatenated table of forward dct coefficients
+		;; 
+extern tab_frw_01234567		; Defined in C for convenience
+
+		;; Offsets into table..
+SECTION .text
+		
+global fdct_mmx
+		
+;;; 
+;;; void fdct_mmx( short *blk )
+;;; 
+
+
+
+;     ////////////////////////////////////////////////////////////////////////
+;     //
+;     // The high-level pseudocode for the fdct_am32() routine :
+;     //
+;     // fdct_am32()
+;     // {
+;     //    forward_dct_col03(); // dct_column transform on cols 0-3
+;     //    forward_dct_col47(); // dct_column transform on cols 4-7
+;     //    for ( j = 0; j < 8; j=j+1 )
+;     //      forward_dct_row1(j); // dct_row transform on row #j
+;     // }
+;     //
+;    
+
+align 32
+fdct_mmx:
+	push ebp			; save stack pointer
+	mov ebp, esp		; link
+
+	push ebx		
+	push ecx		
+	push edx
+	push edi
+					
+	mov INP, [ebp+8];		; input data is row 0 of blk[]
+    ;// transform the left half of the matrix (4 columns)
+
+    lea TABLEF,  [fdct_tg_all_16];
+    mov OUT, INP;
+
+;	lea round_frw_col,  [r_frw_col]
+    ; for ( i = 0; i < 2; i = i + 1)
+    ; the for-loop is executed twice.  We are better off unrolling the 
+    ; loop to avoid branch misprediction.
+.mmx32_fdct_col03: 
+    movq mm0, [x1] ; 0 ; x1
+     ;;
+
+    movq mm1, [x6] ; 1 ; x6
+    movq mm2, mm0 ; 2 ; x1
+
+    movq mm3, [x2] ; 3 ; x2
+    paddsw mm0, mm1 ; t1 = x[1] + x[6]
+
+    movq mm4, [x5] ; 4 ; x5
+    psllw mm0, SHIFT_FRW_COL ; t1
+
+    movq mm5, [x0] ; 5 ; x0
+    paddsw mm4, mm3 ; t2 = x[2] + x[5]
+
+    paddsw mm5, [x7] ; t0 = x[0] + x[7]
+    psllw mm4, SHIFT_FRW_COL ; t2
+
+    movq mm6, mm0 ; 6 ; t1
+    psubsw mm2, mm1 ; 1 ; t6 = x[1] - x[6]
+
+    movq mm1,  [tg_2_16] ; 1 ; tg_2_16
+    psubsw mm0, mm4 ; tm12 = t1 - t2
+
+    movq mm7, [x3] ; 7 ; x3
+    pmulhw mm1, mm0 ; tm12*tg_2_16
+
+    paddsw mm7, [x4] ; t3 = x[3] + x[4]
+    psllw mm5, SHIFT_FRW_COL ; t0
+
+    paddsw mm6, mm4 ; 4 ; tp12 = t1 + t2
+    psllw mm7, SHIFT_FRW_COL ; t3
+
+    movq mm4, mm5 ; 4 ; t0
+    psubsw mm5, mm7 ; tm03 = t0 - t3
+
+    paddsw mm1, mm5 ; y2 = tm03 + tm12*tg_2_16
+    paddsw mm4, mm7 ; 7 ; tp03 = t0 + t3
+
+    por mm1,  [fdct_one_corr] ; correction y2 +0.5
+    psllw mm2, SHIFT_FRW_COL+1 ; t6
+
+    pmulhw mm5,  [tg_2_16] ; tm03*tg_2_16
+    movq mm7, mm4 ; 7 ; tp03
+
+    psubsw mm3, [x5] ; t5 = x[2] - x[5]
+    psubsw mm4, mm6 ; y4 = tp03 - tp12
+
+    movq [y2], mm1 ; 1 ; save y2
+    paddsw mm7, mm6 ; 6 ; y0 = tp03 + tp12
+    
+    movq mm1, [x3] ; 1 ; x3
+    psllw mm3, SHIFT_FRW_COL+1 ; t5
+
+    psubsw mm1, [x4] ; t4 = x[3] - x[4]
+    movq mm6, mm2 ; 6 ; t6
+    
+    movq [y4], mm4 ; 4 ; save y4
+    paddsw mm2, mm3 ; t6 + t5
+
+    pmulhw mm2,  [ocos_4_16] ; tp65 = (t6 + t5)*cos_4_16
+    psubsw mm6, mm3 ; 3 ; t6 - t5
+
+    pmulhw mm6,  [ocos_4_16] ; tm65 = (t6 - t5)*cos_4_16
+    psubsw mm5, mm0 ; 0 ; y6 = tm03*tg_2_16 - tm12
+
+    por mm5,  [fdct_one_corr] ; correction y6 +0.5
+    psllw mm1, SHIFT_FRW_COL ; t4
+
+    por mm2,  [fdct_one_corr] ; correction tp65 +0.5
+    movq mm4, mm1 ; 4 ; t4
+
+    movq mm3, [x0] ; 3 ; x0
+    paddsw mm1, mm6 ; tp465 = t4 + tm65
+
+    psubsw mm3, [x7] ; t7 = x[0] - x[7]
+    psubsw mm4, mm6 ; 6 ; tm465 = t4 - tm65
+
+    movq mm0,  [tg_1_16] ; 0 ; tg_1_16
+    psllw mm3, SHIFT_FRW_COL ; t7
+
+    movq mm6,  [tg_3_16] ; 6 ; tg_3_16
+    pmulhw mm0, mm1 ; tp465*tg_1_16
+
+    movq [y0], mm7 ; 7 ; save y0
+    pmulhw mm6, mm4 ; tm465*tg_3_16
+
+    movq [y6], mm5 ; 5 ; save y6
+    movq mm7, mm3 ; 7 ; t7
+
+    movq mm5,  [tg_3_16] ; 5 ; tg_3_16
+    psubsw mm7, mm2 ; tm765 = t7 - tp65
+
+    paddsw mm3, mm2 ; 2 ; tp765 = t7 + tp65
+    pmulhw mm5, mm7 ; tm765*tg_3_16
+
+    paddsw mm0, mm3 ; y1 = tp765 + tp465*tg_1_16
+    paddsw mm6, mm4 ; tm465*tg_3_16
+
+    pmulhw mm3,  [tg_1_16] ; tp765*tg_1_16
+    ;;
+
+    por mm0,  [fdct_one_corr] ; correction y1 +0.5
+    paddsw mm5, mm7 ; tm765*tg_3_16
+
+    psubsw mm7, mm6 ; 6 ; y3 = tm765 - tm465*tg_3_16
+    add INP, 0x08   ; ; increment pointer
+
+    movq [y1], mm0 ; 0 ; save y1
+    paddsw mm5, mm4 ; 4 ; y5 = tm765*tg_3_16 + tm465
+
+    movq [y3], mm7 ; 7 ; save y3
+    psubsw mm3, mm1 ; 1 ; y7 = tp765*tg_1_16 - tp465
+
+    movq [y5], mm5 ; 5 ; save y5
+
+
+.mmx32_fdct_col47: ; begin processing last four columns
+    movq mm0, [x1] ; 0 ; x1
+    ;;
+    movq [y7], mm3 ; 3 ; save y7 (columns 0-4)
+    ;;
+
+    movq mm1, [x6] ; 1 ; x6
+    movq mm2, mm0 ; 2 ; x1
+
+    movq mm3, [x2] ; 3 ; x2
+    paddsw mm0, mm1 ; t1 = x[1] + x[6]
+
+    movq mm4, [x5] ; 4 ; x5
+    psllw mm0, SHIFT_FRW_COL ; t1
+
+    movq mm5, [x0] ; 5 ; x0
+    paddsw mm4, mm3 ; t2 = x[2] + x[5]
+
+    paddsw mm5, [x7] ; t0 = x[0] + x[7]
+    psllw mm4, SHIFT_FRW_COL ; t2
+
+    movq mm6, mm0 ; 6 ; t1
+    psubsw mm2, mm1 ; 1 ; t6 = x[1] - x[6]
+
+    movq mm1,  [tg_2_16] ; 1 ; tg_2_16
+    psubsw mm0, mm4 ; tm12 = t1 - t2
+
+    movq mm7, [x3] ; 7 ; x3
+    pmulhw mm1, mm0 ; tm12*tg_2_16
+
+    paddsw mm7, [x4] ; t3 = x[3] + x[4]
+    psllw mm5, SHIFT_FRW_COL ; t0
+
+    paddsw mm6, mm4 ; 4 ; tp12 = t1 + t2
+    psllw mm7, SHIFT_FRW_COL ; t3
+
+    movq mm4, mm5 ; 4 ; t0
+    psubsw mm5, mm7 ; tm03 = t0 - t3
+
+    paddsw mm1, mm5 ; y2 = tm03 + tm12*tg_2_16
+    paddsw mm4, mm7 ; 7 ; tp03 = t0 + t3
+
+    por mm1,  [fdct_one_corr] ; correction y2 +0.5
+    psllw mm2, SHIFT_FRW_COL+1 ; t6
+
+    pmulhw mm5,  [tg_2_16] ; tm03*tg_2_16
+    movq mm7, mm4 ; 7 ; tp03
+
+    psubsw mm3, [x5] ; t5 = x[2] - x[5]
+    psubsw mm4, mm6 ; y4 = tp03 - tp12
+
+    movq [y2+8], mm1 ; 1 ; save y2
+    paddsw mm7, mm6 ; 6 ; y0 = tp03 + tp12
+    
+    movq mm1, [x3] ; 1 ; x3
+    psllw mm3, SHIFT_FRW_COL+1 ; t5
+
+    psubsw mm1, [x4] ; t4 = x[3] - x[4]
+    movq mm6, mm2 ; 6 ; t6
+    
+    movq [y4+8], mm4 ; 4 ; save y4
+    paddsw mm2, mm3 ; t6 + t5
+
+    pmulhw mm2,  [ocos_4_16] ; tp65 = (t6 + t5)*cos_4_16
+    psubsw mm6, mm3 ; 3 ; t6 - t5
+
+    pmulhw mm6,  [ocos_4_16] ; tm65 = (t6 - t5)*cos_4_16
+    psubsw mm5, mm0 ; 0 ; y6 = tm03*tg_2_16 - tm12
+
+    por mm5,  [fdct_one_corr] ; correction y6 +0.5
+    psllw mm1, SHIFT_FRW_COL ; t4
+
+    por mm2,  [fdct_one_corr] ; correction tp65 +0.5
+    movq mm4, mm1 ; 4 ; t4
+
+    movq mm3, [x0] ; 3 ; x0
+    paddsw mm1, mm6 ; tp465 = t4 + tm65
+
+    psubsw mm3, [x7] ; t7 = x[0] - x[7]
+    psubsw mm4, mm6 ; 6 ; tm465 = t4 - tm65
+
+    movq mm0,  [tg_1_16] ; 0 ; tg_1_16
+    psllw mm3, SHIFT_FRW_COL ; t7
+
+    movq mm6,  [tg_3_16] ; 6 ; tg_3_16
+    pmulhw mm0, mm1 ; tp465*tg_1_16
+
+    movq [y0+8], mm7 ; 7 ; save y0
+    pmulhw mm6, mm4 ; tm465*tg_3_16
+
+    movq [y6+8], mm5 ; 5 ; save y6
+    movq mm7, mm3 ; 7 ; t7
+
+    movq mm5,  [tg_3_16] ; 5 ; tg_3_16
+    psubsw mm7, mm2 ; tm765 = t7 - tp65
+
+    paddsw mm3, mm2 ; 2 ; tp765 = t7 + tp65
+    pmulhw mm5, mm7 ; tm765*tg_3_16
+
+    paddsw mm0, mm3 ; y1 = tp765 + tp465*tg_1_16
+    paddsw mm6, mm4 ; tm465*tg_3_16
+
+    pmulhw mm3,  [tg_1_16] ; tp765*tg_1_16
+    ;;
+
+    por mm0, [fdct_one_corr] ; correction y1 +0.5
+    paddsw mm5, mm7 ; tm765*tg_3_16
+
+    psubsw mm7, mm6 ; 6 ; y3 = tm765 - tm465*tg_3_16
+    ;;
+
+    movq [y1+8], mm0 ; 0 ; save y1
+    paddsw mm5, mm4 ; 4 ; y5 = tm765*tg_3_16 + tm465
+
+    movq [y3+8], mm7 ; 7 ; save y3
+    psubsw mm3, mm1 ; 1 ; y7 = tp765*tg_1_16 - tp465
+
+    movq [y5+8], mm5 ; 5 ; save y5
+
+    movq [y7+8], mm3 ; 3 ; save y7
+
+;    emms;
+;    }   ; end of forward_dct_col07() 
+    ;  done with dct_row transform
+
+  
+  ; fdct_mmx32_cols() --
+  ; the following subroutine repeats the row-transform operation, 
+  ; except with different shift&round constants.  This version
+  ; does NOT transpose the output again.  Thus the final output
+  ; is transposed with respect to the source.
+  ;
+  ;  The output is stored into blk[], which destroys the original
+  ;  input data.
+	mov INP,  [ebp+8];		;; row 0
+	 mov edi, 0x08;	;x = 8
+
+	lea TABLE,  [tab_frw_01234567]; ; row 0
+	 mov OUT, INP;
+
+	lea round_frw_row,  [fdct_r_row];
+	; for ( x = 8; x > 0; --x )  ; transform one row per iteration
+
+; ---------- loop begin
+  .lp_mmx_fdct_row1:
+    movd mm5,  [INP+12]; ; mm5 = 7 6
+
+    punpcklwd mm5,  [INP+8] ; mm5 =  5 7 4 6
+
+    movq mm2, mm5;     ; mm2 = 5 7 4 6
+    psrlq mm5, 32;     ; mm5 = _ _ 5 7
+
+    movq mm0,  [INP]; ; mm0 = 3 2 1 0
+    punpcklwd mm5, mm2;; mm5 = 4 5 6 7
+
+    movq mm1, mm0;     ; mm1 = 3 2 1 0
+    paddsw mm0, mm5;   ; mm0 = [3+4, 2+5, 1+6, 0+7] (xt3, xt2, xt1, xt0)
+
+    psubsw mm1, mm5;   ; mm1 = [3-4, 2-5, 1-6, 0-7] (xt7, xt6, xt5, xt4)
+    movq mm2, mm0;     ; mm2 = [ xt3 xt2 xt1 xt0 ]
+
+    ;movq [ xt3xt2xt1xt0 ], mm0;
+    ;movq [ xt7xt6xt5xt4 ], mm1;
+
+    punpcklwd mm0, mm1;; mm0 = [ xt5 xt1 xt4 xt0 ]
+
+    punpckhwd mm2, mm1;; mm2 = [ xt7 xt3 xt6 xt2 ]
+    movq mm1, mm2;     ; mm1
+
+    ;; shuffle bytes around
+
+;  movq mm0,  [INP] ; 0 ; x3 x2 x1 x0
+
+;  movq mm1,  [INP+8] ; 1 ; x7 x6 x5 x4
+    movq mm2, mm0 ; 2 ; x3 x2 x1 x0
+
+    movq mm3,  [TABLE] ; 3 ; w06 w04 w02 w00
+    punpcklwd mm0, mm1 ; x5 x1 x4 x0
+
+    movq mm5, mm0 ; 5 ; x5 x1 x4 x0
+    punpckldq mm0, mm0 ; x4 x0 x4 x0  [ xt2 xt0 xt2 xt0 ]
+
+    movq mm4,  [TABLE+8] ; 4 ; w07 w05 w03 w01
+    punpckhwd mm2, mm1 ; 1 ; x7 x3 x6 x2
+
+    pmaddwd mm3, mm0 ; x4*w06+x0*w04 x4*w02+x0*w00
+    movq mm6, mm2 ; 6 ; x7 x3 x6 x2
+
+    movq mm1,  [TABLE+32] ; 1 ; w22 w20 w18 w16
+    punpckldq mm2, mm2 ; x6 x2 x6 x2  [ xt3 xt1 xt3 xt1 ]
+
+    pmaddwd mm4, mm2 ; x6*w07+x2*w05 x6*w03+x2*w01
+    punpckhdq mm5, mm5 ; x5 x1 x5 x1  [ xt6 xt4 xt6 xt4 ]
+
+    pmaddwd mm0,  [TABLE+16] ; x4*w14+x0*w12 x4*w10+x0*w08
+    punpckhdq mm6, mm6 ; x7 x3 x7 x3  [ xt7 xt5 xt7 xt5 ]
+
+    movq mm7,  [TABLE+40] ; 7 ; w23 w21 w19 w17
+    pmaddwd mm1, mm5 ; x5*w22+x1*w20 x5*w18+x1*w16
+;mm3 = a1, a0 (y2,y0)
+;mm1 = b1, b0 (y3,y1)
+;mm0 = a3,a2  (y6,y4)
+;mm5 = b3,b2  (y7,y5)
+
+    paddd mm3,  [round_frw_row] ; +rounder (y2,y0)
+    pmaddwd mm7, mm6 ; x7*w23+x3*w21 x7*w19+x3*w17
+
+    pmaddwd mm2,  [TABLE+24] ; x6*w15+x2*w13 x6*w11+x2*w09
+    paddd mm3, mm4 ; 4 ; a1=sum(even1) a0=sum(even0) ; now ( y2, y0)
+
+    pmaddwd mm5,  [TABLE+48] ; x5*w30+x1*w28 x5*w26+x1*w24
+    ;;
+
+    pmaddwd mm6,  [TABLE+56] ; x7*w31+x3*w29 x7*w27+x3*w25
+    paddd mm1, mm7 ; 7 ; b1=sum(odd1) b0=sum(odd0) ; now ( y3, y1)
+
+    paddd mm0,  [round_frw_row] ; +rounder (y6,y4)
+    psrad mm3, SHIFT_FRW_ROW ; (y2, y0)
+
+    paddd mm1,  [round_frw_row] ; +rounder (y3,y1)
+    paddd mm0, mm2 ; 2 ; a3=sum(even3) a2=sum(even2) ; now (y6, y4)
+
+    paddd mm5,  [round_frw_row] ; +rounder (y7,y5)
+    psrad mm1, SHIFT_FRW_ROW ; y1=a1+b1 y0=a0+b0
+
+    paddd mm5, mm6 ; 6 ; b3=sum(odd3) b2=sum(odd2) ; now ( y7, y5)
+    psrad mm0, SHIFT_FRW_ROW ;y3=a3+b3 y2=a2+b2
+
+    add OUT, 16;  ; increment row-output address by 1 row
+    psrad mm5, SHIFT_FRW_ROW ; y4=a3-b3 y5=a2-b2
+
+    add INP, 16;  ; increment row-address by 1 row
+    packssdw mm3, mm0 ; 0 ; y6 y4 y2 y0
+
+    packssdw mm1, mm5 ; 3 ; y7 y5 y3 y1
+    movq mm6, mm3;    ; mm0 = y6 y4 y2 y0
+
+    punpcklwd mm3, mm1; ; y3 y2 y1 y0
+    sub edi, 0x01;   ; i = i - 1
+    
+    punpckhwd mm6, mm1; ; y7 y6 y5 y4
+    add TABLE,64;  ; increment to next table
+
+    movq  [OUT-16], mm3 ; 1 ; save y3 y2 y1 y0
+
+    movq  [OUT-8], mm6 ; 7 ; save y7 y6 y5 y4
+
+    cmp edi, 0x00;
+    jg near .lp_mmx_fdct_row1;  ; begin fdct processing on next row
+		;; 
+		;; Tidy up and return
+		;;
+	pop edi
+	pop edx			
+	pop ecx			
+	pop ebx			
+
+	pop ebp			; restore stack pointer
+	emms
+	ret		
+  
\ No newline at end of file
diff --git a/libavcodec/i386/fdctdata.c b/libavcodec/i386/fdctdata.c
new file mode 100644
index 0000000000..e095d0f4e4
--- /dev/null
+++ b/libavcodec/i386/fdctdata.c
@@ -0,0 +1,143 @@
+//////////////////////////////////////////////////////////////////////////////
+//
+//  fdctam32.c - AP922 MMX(3D-Now) forward-DCT
+//  ----------
+//  Intel Application Note AP-922 - fast, precise implementation of DCT
+//        http://developer.intel.com/vtune/cbts/appnotes.htm
+//  ----------
+//  
+//       This routine uses a 3D-Now/MMX enhancement to increase the
+//  accuracy of the fdct_col_4 macro.  The dct_col function uses 3D-Now's
+//  PMHULHRW instead of MMX's PMHULHW(and POR).  The substitution improves
+//  accuracy very slightly with performance penalty.  If the target CPU
+//  does not support 3D-Now, then this function cannot be executed.  
+//  fdctmm32.c contains the standard MMX implementation of AP-922.
+//  
+//  For a fast, precise MMX implementation of inverse-DCT 
+//              visit http://www.elecard.com/peter
+//
+//  v1.0 07/22/2000 (initial release)
+//       Initial release of AP922 MMX(3D-Now) forward_DCT.
+//       This code was tested with Visual C++ 6.0Pro + service_pack4 + 
+//       processor_pack_beta!  If you have the processor_pack_beta, you can
+//       remove the #include for amd3dx.h, and substitute the 'normal'
+//       assembly lines for the macro'd versions.  Otherwise, this
+//       code should compile 'as is', under Visual C++ 6.0 Pro.
+//     
+//  liaor@iname.com  http://members.tripod.com/~liaor  
+//////////////////////////////////////////////////////////////////////////////
+
+#include <inttypes.h>
+
+//////////////////////////////////////////////////////////////////////
+//
+// constants for the forward DCT
+// -----------------------------
+//
+// Be sure to check that your compiler is aligning all constants to QWORD
+// (8-byte) memory boundaries!  Otherwise the unaligned memory access will
+// severely stall MMX execution.
+//
+//////////////////////////////////////////////////////////////////////
+
+#define BITS_FRW_ACC	3 //; 2 or 3 for accuracy
+#define SHIFT_FRW_COL	BITS_FRW_ACC
+#define SHIFT_FRW_ROW	(BITS_FRW_ACC + 17)
+//#define RND_FRW_ROW		(262144 * (BITS_FRW_ACC - 1)) //; 1 << (SHIFT_FRW_ROW-1)
+#define RND_FRW_ROW		(1 << (SHIFT_FRW_ROW-1))
+//#define RND_FRW_COL		(2 * (BITS_FRW_ACC - 1)) //; 1 << (SHIFT_FRW_COL-1)
+#define RND_FRW_COL		(1 << (SHIFT_FRW_COL-1))
+
+//concatenated table, for forward DCT transformation
+const int16_t fdct_tg_all_16[] = {
+	13036, 13036, 13036, 13036,		// tg * (2<<16) + 0.5
+	27146, 27146, 27146, 27146,		// tg * (2<<16) + 0.5
+	-21746, -21746, -21746, -21746,	// tg * (2<<16) + 0.5
+	-19195, -19195, -19195, -19195,	//cos * (2<<16) + 0.5
+	23170, 23170, 23170, 23170 };	//cos * (2<<15) + 0.5
+const long long  fdct_one_corr = 0x0001000100010001LL;
+const long fdct_r_row[2] = {RND_FRW_ROW, RND_FRW_ROW };
+
+const int16_t tab_frw_01234567[] = {  // forward_dct coeff table
+    //row0
+    16384, 16384, 21407, -8867,     //    w09 w01 w08 w00
+    16384, 16384, 8867, -21407,     //    w13 w05 w12 w04
+    16384, -16384, 8867, 21407,     //    w11 w03 w10 w02
+    -16384, 16384, -21407, -8867,   //    w15 w07 w14 w06
+    22725, 12873, 19266, -22725,    //    w22 w20 w18 w16
+    19266, 4520, -4520, -12873,     //    w23 w21 w19 w17
+    12873, 4520, 4520, 19266,       //    w30 w28 w26 w24
+    -22725, 19266, -12873, -22725,  //    w31 w29 w27 w25
+
+    //row1
+    22725, 22725, 29692, -12299,    //    w09 w01 w08 w00
+    22725, 22725, 12299, -29692,    //    w13 w05 w12 w04
+    22725, -22725, 12299, 29692,    //    w11 w03 w10 w02
+    -22725, 22725, -29692, -12299,  //    w15 w07 w14 w06
+    31521, 17855, 26722, -31521,    //    w22 w20 w18 w16
+    26722, 6270, -6270, -17855,     //    w23 w21 w19 w17
+    17855, 6270, 6270, 26722,       //    w30 w28 w26 w24
+    -31521, 26722, -17855, -31521,  //    w31 w29 w27 w25
+
+    //row2
+    21407, 21407, 27969, -11585,    //    w09 w01 w08 w00
+    21407, 21407, 11585, -27969,    //    w13 w05 w12 w04
+    21407, -21407, 11585, 27969,    //    w11 w03 w10 w02
+    -21407, 21407, -27969, -11585,  //    w15 w07 w14 w06
+    29692, 16819, 25172, -29692,    //    w22 w20 w18 w16
+    25172, 5906, -5906, -16819,     //    w23 w21 w19 w17
+    16819, 5906, 5906, 25172,       //    w30 w28 w26 w24
+    -29692, 25172, -16819, -29692,  //    w31 w29 w27 w25
+
+    //row3
+    19266, 19266, 25172, -10426,    //    w09 w01 w08 w00
+    19266, 19266, 10426, -25172,    //    w13 w05 w12 w04
+    19266, -19266, 10426, 25172,    //    w11 w03 w10 w02
+    -19266, 19266, -25172, -10426,  //    w15 w07 w14 w06, 
+    26722, 15137, 22654, -26722,    //    w22 w20 w18 w16
+    22654, 5315, -5315, -15137,     //    w23 w21 w19 w17
+    15137, 5315, 5315, 22654,       //    w30 w28 w26 w24
+    -26722, 22654, -15137, -26722,  //    w31 w29 w27 w25, 
+
+    //row4
+    16384, 16384, 21407, -8867,     //    w09 w01 w08 w00
+    16384, 16384, 8867, -21407,     //    w13 w05 w12 w04
+    16384, -16384, 8867, 21407,     //    w11 w03 w10 w02
+    -16384, 16384, -21407, -8867,   //    w15 w07 w14 w06
+    22725, 12873, 19266, -22725,    //    w22 w20 w18 w16
+    19266, 4520, -4520, -12873,     //    w23 w21 w19 w17
+    12873, 4520, 4520, 19266,       //    w30 w28 w26 w24
+    -22725, 19266, -12873, -22725,  //    w31 w29 w27 w25 
+
+    //row5
+    19266, 19266, 25172, -10426,    //    w09 w01 w08 w00
+    19266, 19266, 10426, -25172,    //    w13 w05 w12 w04
+    19266, -19266, 10426, 25172,    //    w11 w03 w10 w02
+    -19266, 19266, -25172, -10426,  //    w15 w07 w14 w06
+    26722, 15137, 22654, -26722,    //    w22 w20 w18 w16
+    22654, 5315, -5315, -15137,     //    w23 w21 w19 w17
+    15137, 5315, 5315, 22654,       //    w30 w28 w26 w24
+    -26722, 22654, -15137, -26722,  //    w31 w29 w27 w25
+
+    //row6
+    21407, 21407, 27969, -11585,    //    w09 w01 w08 w00
+    21407, 21407, 11585, -27969,    //    w13 w05 w12 w04
+    21407, -21407, 11585, 27969,    //    w11 w03 w10 w02
+    -21407, 21407, -27969, -11585,  //    w15 w07 w14 w06, 
+    29692, 16819, 25172, -29692,    //    w22 w20 w18 w16
+    25172, 5906, -5906, -16819,     //    w23 w21 w19 w17
+    16819, 5906, 5906, 25172,       //    w30 w28 w26 w24
+    -29692, 25172, -16819, -29692,  //    w31 w29 w27 w25, 
+
+    //row7
+    22725, 22725, 29692, -12299,    //    w09 w01 w08 w00
+    22725, 22725, 12299, -29692,    //    w13 w05 w12 w04
+    22725, -22725, 12299, 29692,    //    w11 w03 w10 w02
+    -22725, 22725, -29692, -12299,  //    w15 w07 w14 w06, 
+    31521, 17855, 26722, -31521,    //    w22 w20 w18 w16
+    26722, 6270, -6270, -17855,     //    w23 w21 w19 w17
+    17855, 6270, 6270, 26722,       //    w30 w28 w26 w24
+    -31521, 26722, -17855, -31521   //    w31 w29 w27 w25
+};
+
+
diff --git a/libavcodec/i386/mmx.h b/libavcodec/i386/mmx.h
new file mode 100644
index 0000000000..a82c0eda32
--- /dev/null
+++ b/libavcodec/i386/mmx.h
@@ -0,0 +1,536 @@
+/*	mmx.h
+
+	MultiMedia eXtensions GCC interface library for IA32.
+
+	To use this library, simply include this header file
+	and compile with GCC.  You MUST have inlining enabled
+	in order for mmx_ok() to work; this can be done by
+	simply using -O on the GCC command line.
+
+	Compiling with -DMMX_TRACE will cause detailed trace
+	output to be sent to stderr for each mmx operation.
+	This adds lots of code, and obviously slows execution to
+	a crawl, but can be very useful for debugging.
+
+	THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY
+	EXPRESS OR IMPLIED WARRANTIES, INCLUDING, WITHOUT
+	LIMITATION, THE IMPLIED WARRANTIES OF MERCHANTABILITY
+	AND FITNESS FOR ANY PARTICULAR PURPOSE.
+
+	1997-99 by H. Dietz and R. Fisher
+
+ Notes:
+	It appears that the latest gas has the pand problem fixed, therefore
+	  I'll undefine BROKEN_PAND by default.
+*/
+
+#ifndef _MMX_H
+#define _MMX_H
+
+
+/*	Warning:  at this writing, the version of GAS packaged
+	with most Linux distributions does not handle the
+	parallel AND operation mnemonic correctly.  If the
+	symbol BROKEN_PAND is defined, a slower alternative
+	coding will be used.  If execution of mmxtest results
+	in an illegal instruction fault, define this symbol.
+*/
+#undef	BROKEN_PAND
+
+
+/*	The type of an value that fits in an MMX register
+	(note that long long constant values MUST be suffixed
+	 by LL and unsigned long long values by ULL, lest
+	 they be truncated by the compiler)
+*/
+typedef	union {
+	long long		q;	/* Quadword (64-bit) value */
+	unsigned long long	uq;	/* Unsigned Quadword */
+	int			d[2];	/* 2 Doubleword (32-bit) values */
+	unsigned int		ud[2];	/* 2 Unsigned Doubleword */
+	short			w[4];	/* 4 Word (16-bit) values */
+	unsigned short		uw[4];	/* 4 Unsigned Word */
+	char			b[8];	/* 8 Byte (8-bit) values */
+	unsigned char		ub[8];	/* 8 Unsigned Byte */
+	float			s[2];	/* Single-precision (32-bit) value */
+} __attribute__ ((aligned (8))) mmx_t;	/* On an 8-byte (64-bit) boundary */
+
+
+/*	Helper functions for the instruction macros that follow...
+	(note that memory-to-register, m2r, instructions are nearly
+	 as efficient as register-to-register, r2r, instructions;
+	 however, memory-to-memory instructions are really simulated
+	 as a convenience, and are only 1/3 as efficient)
+*/
+#ifdef	MMX_TRACE
+
+/*	Include the stuff for printing a trace to stderr...
+*/
+
+#include <stdio.h>
+
+#define	mmx_i2r(op, imm, reg) \
+	{ \
+		mmx_t mmx_trace; \
+		mmx_trace.uq = (imm); \
+		fprintf(stderr, #op "_i2r(" #imm "=0x%08x%08x, ", \
+			mmx_trace.d[1], mmx_trace.d[0]); \
+		__asm__ __volatile__ ("movq %%" #reg ", %0" \
+				      : "=X" (mmx_trace) \
+				      : /* nothing */ ); \
+		fprintf(stderr, #reg "=0x%08x%08x) => ", \
+			mmx_trace.d[1], mmx_trace.d[0]); \
+		__asm__ __volatile__ (#op " %0, %%" #reg \
+				      : /* nothing */ \
+				      : "X" (imm)); \
+		__asm__ __volatile__ ("movq %%" #reg ", %0" \
+				      : "=X" (mmx_trace) \
+				      : /* nothing */ ); \
+		fprintf(stderr, #reg "=0x%08x%08x\n", \
+			mmx_trace.d[1], mmx_trace.d[0]); \
+	}
+
+#define	mmx_m2r(op, mem, reg) \
+	{ \
+		mmx_t mmx_trace; \
+		mmx_trace = (mem); \
+		fprintf(stderr, #op "_m2r(" #mem "=0x%08x%08x, ", \
+			mmx_trace.d[1], mmx_trace.d[0]); \
+		__asm__ __volatile__ ("movq %%" #reg ", %0" \
+				      : "=X" (mmx_trace) \
+				      : /* nothing */ ); \
+		fprintf(stderr, #reg "=0x%08x%08x) => ", \
+			mmx_trace.d[1], mmx_trace.d[0]); \
+		__asm__ __volatile__ (#op " %0, %%" #reg \
+				      : /* nothing */ \
+				      : "X" (mem)); \
+		__asm__ __volatile__ ("movq %%" #reg ", %0" \
+				      : "=X" (mmx_trace) \
+				      : /* nothing */ ); \
+		fprintf(stderr, #reg "=0x%08x%08x\n", \
+			mmx_trace.d[1], mmx_trace.d[0]); \
+	}
+
+#define	mmx_r2m(op, reg, mem) \
+	{ \
+		mmx_t mmx_trace; \
+		__asm__ __volatile__ ("movq %%" #reg ", %0" \
+				      : "=X" (mmx_trace) \
+				      : /* nothing */ ); \
+		fprintf(stderr, #op "_r2m(" #reg "=0x%08x%08x, ", \
+			mmx_trace.d[1], mmx_trace.d[0]); \
+		mmx_trace = (mem); \
+		fprintf(stderr, #mem "=0x%08x%08x) => ", \
+			mmx_trace.d[1], mmx_trace.d[0]); \
+		__asm__ __volatile__ (#op " %%" #reg ", %0" \
+				      : "=X" (mem) \
+				      : /* nothing */ ); \
+		mmx_trace = (mem); \
+		fprintf(stderr, #mem "=0x%08x%08x\n", \
+			mmx_trace.d[1], mmx_trace.d[0]); \
+	}
+
+#define	mmx_r2r(op, regs, regd) \
+	{ \
+		mmx_t mmx_trace; \
+		__asm__ __volatile__ ("movq %%" #regs ", %0" \
+				      : "=X" (mmx_trace) \
+				      : /* nothing */ ); \
+		fprintf(stderr, #op "_r2r(" #regs "=0x%08x%08x, ", \
+			mmx_trace.d[1], mmx_trace.d[0]); \
+		__asm__ __volatile__ ("movq %%" #regd ", %0" \
+				      : "=X" (mmx_trace) \
+				      : /* nothing */ ); \
+		fprintf(stderr, #regd "=0x%08x%08x) => ", \
+			mmx_trace.d[1], mmx_trace.d[0]); \
+		__asm__ __volatile__ (#op " %" #regs ", %" #regd); \
+		__asm__ __volatile__ ("movq %%" #regd ", %0" \
+				      : "=X" (mmx_trace) \
+				      : /* nothing */ ); \
+		fprintf(stderr, #regd "=0x%08x%08x\n", \
+			mmx_trace.d[1], mmx_trace.d[0]); \
+	}
+
+#define	mmx_m2m(op, mems, memd) \
+	{ \
+		mmx_t mmx_trace; \
+		mmx_trace = (mems); \
+		fprintf(stderr, #op "_m2m(" #mems "=0x%08x%08x, ", \
+			mmx_trace.d[1], mmx_trace.d[0]); \
+		mmx_trace = (memd); \
+		fprintf(stderr, #memd "=0x%08x%08x) => ", \
+			mmx_trace.d[1], mmx_trace.d[0]); \
+		__asm__ __volatile__ ("movq %0, %%mm0\n\t" \
+				      #op " %1, %%mm0\n\t" \
+				      "movq %%mm0, %0" \
+				      : "=X" (memd) \
+				      : "X" (mems)); \
+		mmx_trace = (memd); \
+		fprintf(stderr, #memd "=0x%08x%08x\n", \
+			mmx_trace.d[1], mmx_trace.d[0]); \
+	}
+
+#else
+
+/*	These macros are a lot simpler without the tracing...
+*/
+
+#define	mmx_i2r(op, imm, reg) \
+	__asm__ __volatile__ (#op " %0, %%" #reg \
+			      : /* nothing */ \
+			      : "i" (imm) )
+
+#define	mmx_m2r(op, mem, reg) \
+	__asm__ __volatile__ (#op " %0, %%" #reg \
+			      : /* nothing */ \
+			      : "m" (mem))
+
+#define	mmx_r2m(op, reg, mem) \
+	__asm__ __volatile__ (#op " %%" #reg ", %0" \
+			      : "=m" (mem) \
+			      : /* nothing */ )
+
+#define	mmx_r2r(op, regs, regd) \
+	__asm__ __volatile__ (#op " %" #regs ", %" #regd)
+
+#define	mmx_m2m(op, mems, memd) \
+	__asm__ __volatile__ ("movq %0, %%mm0\n\t" \
+			      #op " %1, %%mm0\n\t" \
+			      "movq %%mm0, %0" \
+			      : "=m" (memd) \
+			      : "m" (mems))
+
+#endif
+
+
+/*	1x64 MOVe Quadword
+	(this is both a load and a store...
+	 in fact, it is the only way to store)
+*/
+#define	movq_m2r(var, reg)	mmx_m2r(movq, var, reg)
+#define	movq_r2m(reg, var)	mmx_r2m(movq, reg, var)
+#define	movq_r2r(regs, regd)	mmx_r2r(movq, regs, regd)
+#define	movq(vars, vard) \
+	__asm__ __volatile__ ("movq %1, %%mm0\n\t" \
+			      "movq %%mm0, %0" \
+			      : "=X" (vard) \
+			      : "X" (vars))
+
+
+/*	1x32 MOVe Doubleword
+	(like movq, this is both load and store...
+	 but is most useful for moving things between
+	 mmx registers and ordinary registers)
+*/
+#define	movd_m2r(var, reg)	mmx_m2r(movd, var, reg)
+#define	movd_r2m(reg, var)	mmx_r2m(movd, reg, var)
+#define	movd_r2r(regs, regd)	mmx_r2r(movd, regs, regd)
+#define	movd(vars, vard) \
+	__asm__ __volatile__ ("movd %1, %%mm0\n\t" \
+			      "movd %%mm0, %0" \
+			      : "=X" (vard) \
+			      : "X" (vars))
+
+
+/*	2x32, 4x16, and 8x8 Parallel ADDs
+*/
+#define	paddd_m2r(var, reg)	mmx_m2r(paddd, var, reg)
+#define	paddd_r2r(regs, regd)	mmx_r2r(paddd, regs, regd)
+#define	paddd(vars, vard)	mmx_m2m(paddd, vars, vard)
+
+#define	paddw_m2r(var, reg)	mmx_m2r(paddw, var, reg)
+#define	paddw_r2r(regs, regd)	mmx_r2r(paddw, regs, regd)
+#define	paddw(vars, vard)	mmx_m2m(paddw, vars, vard)
+
+#define	paddb_m2r(var, reg)	mmx_m2r(paddb, var, reg)
+#define	paddb_r2r(regs, regd)	mmx_r2r(paddb, regs, regd)
+#define	paddb(vars, vard)	mmx_m2m(paddb, vars, vard)
+
+
+/*	4x16 and 8x8 Parallel ADDs using Saturation arithmetic
+*/
+#define	paddsw_m2r(var, reg)	mmx_m2r(paddsw, var, reg)
+#define	paddsw_r2r(regs, regd)	mmx_r2r(paddsw, regs, regd)
+#define	paddsw(vars, vard)	mmx_m2m(paddsw, vars, vard)
+
+#define	paddsb_m2r(var, reg)	mmx_m2r(paddsb, var, reg)
+#define	paddsb_r2r(regs, regd)	mmx_r2r(paddsb, regs, regd)
+#define	paddsb(vars, vard)	mmx_m2m(paddsb, vars, vard)
+
+
+/*	4x16 and 8x8 Parallel ADDs using Unsigned Saturation arithmetic
+*/
+#define	paddusw_m2r(var, reg)	mmx_m2r(paddusw, var, reg)
+#define	paddusw_r2r(regs, regd)	mmx_r2r(paddusw, regs, regd)
+#define	paddusw(vars, vard)	mmx_m2m(paddusw, vars, vard)
+
+#define	paddusb_m2r(var, reg)	mmx_m2r(paddusb, var, reg)
+#define	paddusb_r2r(regs, regd)	mmx_r2r(paddusb, regs, regd)
+#define	paddusb(vars, vard)	mmx_m2m(paddusb, vars, vard)
+
+
+/*	2x32, 4x16, and 8x8 Parallel SUBs
+*/
+#define	psubd_m2r(var, reg)	mmx_m2r(psubd, var, reg)
+#define	psubd_r2r(regs, regd)	mmx_r2r(psubd, regs, regd)
+#define	psubd(vars, vard)	mmx_m2m(psubd, vars, vard)
+
+#define	psubw_m2r(var, reg)	mmx_m2r(psubw, var, reg)
+#define	psubw_r2r(regs, regd)	mmx_r2r(psubw, regs, regd)
+#define	psubw(vars, vard)	mmx_m2m(psubw, vars, vard)
+
+#define	psubb_m2r(var, reg)	mmx_m2r(psubb, var, reg)
+#define	psubb_r2r(regs, regd)	mmx_r2r(psubb, regs, regd)
+#define	psubb(vars, vard)	mmx_m2m(psubb, vars, vard)
+
+
+/*	4x16 and 8x8 Parallel SUBs using Saturation arithmetic
+*/
+#define	psubsw_m2r(var, reg)	mmx_m2r(psubsw, var, reg)
+#define	psubsw_r2r(regs, regd)	mmx_r2r(psubsw, regs, regd)
+#define	psubsw(vars, vard)	mmx_m2m(psubsw, vars, vard)
+
+#define	psubsb_m2r(var, reg)	mmx_m2r(psubsb, var, reg)
+#define	psubsb_r2r(regs, regd)	mmx_r2r(psubsb, regs, regd)
+#define	psubsb(vars, vard)	mmx_m2m(psubsb, vars, vard)
+
+
+/*	4x16 and 8x8 Parallel SUBs using Unsigned Saturation arithmetic
+*/
+#define	psubusw_m2r(var, reg)	mmx_m2r(psubusw, var, reg)
+#define	psubusw_r2r(regs, regd)	mmx_r2r(psubusw, regs, regd)
+#define	psubusw(vars, vard)	mmx_m2m(psubusw, vars, vard)
+
+#define	psubusb_m2r(var, reg)	mmx_m2r(psubusb, var, reg)
+#define	psubusb_r2r(regs, regd)	mmx_r2r(psubusb, regs, regd)
+#define	psubusb(vars, vard)	mmx_m2m(psubusb, vars, vard)
+
+
+/*	4x16 Parallel MULs giving Low 4x16 portions of results
+*/
+#define	pmullw_m2r(var, reg)	mmx_m2r(pmullw, var, reg)
+#define	pmullw_r2r(regs, regd)	mmx_r2r(pmullw, regs, regd)
+#define	pmullw(vars, vard)	mmx_m2m(pmullw, vars, vard)
+
+
+/*	4x16 Parallel MULs giving High 4x16 portions of results
+*/
+#define	pmulhw_m2r(var, reg)	mmx_m2r(pmulhw, var, reg)
+#define	pmulhw_r2r(regs, regd)	mmx_r2r(pmulhw, regs, regd)
+#define	pmulhw(vars, vard)	mmx_m2m(pmulhw, vars, vard)
+
+
+/*	4x16->2x32 Parallel Mul-ADD
+	(muls like pmullw, then adds adjacent 16-bit fields
+	 in the multiply result to make the final 2x32 result)
+*/
+#define	pmaddwd_m2r(var, reg)	mmx_m2r(pmaddwd, var, reg)
+#define	pmaddwd_r2r(regs, regd)	mmx_r2r(pmaddwd, regs, regd)
+#define	pmaddwd(vars, vard)	mmx_m2m(pmaddwd, vars, vard)
+
+
+/*	1x64 bitwise AND
+*/
+#ifdef	BROKEN_PAND
+#define	pand_m2r(var, reg) \
+	{ \
+		mmx_m2r(pandn, (mmx_t) -1LL, reg); \
+		mmx_m2r(pandn, var, reg); \
+	}
+#define	pand_r2r(regs, regd) \
+	{ \
+		mmx_m2r(pandn, (mmx_t) -1LL, regd); \
+		mmx_r2r(pandn, regs, regd) \
+	}
+#define	pand(vars, vard) \
+	{ \
+		movq_m2r(vard, mm0); \
+		mmx_m2r(pandn, (mmx_t) -1LL, mm0); \
+		mmx_m2r(pandn, vars, mm0); \
+		movq_r2m(mm0, vard); \
+	}
+#else
+#define	pand_m2r(var, reg)	mmx_m2r(pand, var, reg)
+#define	pand_r2r(regs, regd)	mmx_r2r(pand, regs, regd)
+#define	pand(vars, vard)	mmx_m2m(pand, vars, vard)
+#endif
+
+
+/*	1x64 bitwise AND with Not the destination
+*/
+#define	pandn_m2r(var, reg)	mmx_m2r(pandn, var, reg)
+#define	pandn_r2r(regs, regd)	mmx_r2r(pandn, regs, regd)
+#define	pandn(vars, vard)	mmx_m2m(pandn, vars, vard)
+
+
+/*	1x64 bitwise OR
+*/
+#define	por_m2r(var, reg)	mmx_m2r(por, var, reg)
+#define	por_r2r(regs, regd)	mmx_r2r(por, regs, regd)
+#define	por(vars, vard)	mmx_m2m(por, vars, vard)
+
+
+/*	1x64 bitwise eXclusive OR
+*/
+#define	pxor_m2r(var, reg)	mmx_m2r(pxor, var, reg)
+#define	pxor_r2r(regs, regd)	mmx_r2r(pxor, regs, regd)
+#define	pxor(vars, vard)	mmx_m2m(pxor, vars, vard)
+
+
+/*	2x32, 4x16, and 8x8 Parallel CoMPare for EQuality
+	(resulting fields are either 0 or -1)
+*/
+#define	pcmpeqd_m2r(var, reg)	mmx_m2r(pcmpeqd, var, reg)
+#define	pcmpeqd_r2r(regs, regd)	mmx_r2r(pcmpeqd, regs, regd)
+#define	pcmpeqd(vars, vard)	mmx_m2m(pcmpeqd, vars, vard)
+
+#define	pcmpeqw_m2r(var, reg)	mmx_m2r(pcmpeqw, var, reg)
+#define	pcmpeqw_r2r(regs, regd)	mmx_r2r(pcmpeqw, regs, regd)
+#define	pcmpeqw(vars, vard)	mmx_m2m(pcmpeqw, vars, vard)
+
+#define	pcmpeqb_m2r(var, reg)	mmx_m2r(pcmpeqb, var, reg)
+#define	pcmpeqb_r2r(regs, regd)	mmx_r2r(pcmpeqb, regs, regd)
+#define	pcmpeqb(vars, vard)	mmx_m2m(pcmpeqb, vars, vard)
+
+
+/*	2x32, 4x16, and 8x8 Parallel CoMPare for Greater Than
+	(resulting fields are either 0 or -1)
+*/
+#define	pcmpgtd_m2r(var, reg)	mmx_m2r(pcmpgtd, var, reg)
+#define	pcmpgtd_r2r(regs, regd)	mmx_r2r(pcmpgtd, regs, regd)
+#define	pcmpgtd(vars, vard)	mmx_m2m(pcmpgtd, vars, vard)
+
+#define	pcmpgtw_m2r(var, reg)	mmx_m2r(pcmpgtw, var, reg)
+#define	pcmpgtw_r2r(regs, regd)	mmx_r2r(pcmpgtw, regs, regd)
+#define	pcmpgtw(vars, vard)	mmx_m2m(pcmpgtw, vars, vard)
+
+#define	pcmpgtb_m2r(var, reg)	mmx_m2r(pcmpgtb, var, reg)
+#define	pcmpgtb_r2r(regs, regd)	mmx_r2r(pcmpgtb, regs, regd)
+#define	pcmpgtb(vars, vard)	mmx_m2m(pcmpgtb, vars, vard)
+
+
+/*	1x64, 2x32, and 4x16 Parallel Shift Left Logical
+*/
+#define	psllq_i2r(imm, reg)	mmx_i2r(psllq, imm, reg)
+#define	psllq_m2r(var, reg)	mmx_m2r(psllq, var, reg)
+#define	psllq_r2r(regs, regd)	mmx_r2r(psllq, regs, regd)
+#define	psllq(vars, vard)	mmx_m2m(psllq, vars, vard)
+
+#define	pslld_i2r(imm, reg)	mmx_i2r(pslld, imm, reg)
+#define	pslld_m2r(var, reg)	mmx_m2r(pslld, var, reg)
+#define	pslld_r2r(regs, regd)	mmx_r2r(pslld, regs, regd)
+#define	pslld(vars, vard)	mmx_m2m(pslld, vars, vard)
+
+#define	psllw_i2r(imm, reg)	mmx_i2r(psllw, imm, reg)
+#define	psllw_m2r(var, reg)	mmx_m2r(psllw, var, reg)
+#define	psllw_r2r(regs, regd)	mmx_r2r(psllw, regs, regd)
+#define	psllw(vars, vard)	mmx_m2m(psllw, vars, vard)
+
+
+/*	1x64, 2x32, and 4x16 Parallel Shift Right Logical
+*/
+#define	psrlq_i2r(imm, reg)	mmx_i2r(psrlq, imm, reg)
+#define	psrlq_m2r(var, reg)	mmx_m2r(psrlq, var, reg)
+#define	psrlq_r2r(regs, regd)	mmx_r2r(psrlq, regs, regd)
+#define	psrlq(vars, vard)	mmx_m2m(psrlq, vars, vard)
+
+#define	psrld_i2r(imm, reg)	mmx_i2r(psrld, imm, reg)
+#define	psrld_m2r(var, reg)	mmx_m2r(psrld, var, reg)
+#define	psrld_r2r(regs, regd)	mmx_r2r(psrld, regs, regd)
+#define	psrld(vars, vard)	mmx_m2m(psrld, vars, vard)
+
+#define	psrlw_i2r(imm, reg)	mmx_i2r(psrlw, imm, reg)
+#define	psrlw_m2r(var, reg)	mmx_m2r(psrlw, var, reg)
+#define	psrlw_r2r(regs, regd)	mmx_r2r(psrlw, regs, regd)
+#define	psrlw(vars, vard)	mmx_m2m(psrlw, vars, vard)
+
+
+/*	2x32 and 4x16 Parallel Shift Right Arithmetic
+*/
+#define	psrad_i2r(imm, reg)	mmx_i2r(psrad, imm, reg)
+#define	psrad_m2r(var, reg)	mmx_m2r(psrad, var, reg)
+#define	psrad_r2r(regs, regd)	mmx_r2r(psrad, regs, regd)
+#define	psrad(vars, vard)	mmx_m2m(psrad, vars, vard)
+
+#define	psraw_i2r(imm, reg)	mmx_i2r(psraw, imm, reg)
+#define	psraw_m2r(var, reg)	mmx_m2r(psraw, var, reg)
+#define	psraw_r2r(regs, regd)	mmx_r2r(psraw, regs, regd)
+#define	psraw(vars, vard)	mmx_m2m(psraw, vars, vard)
+
+
+/*	2x32->4x16 and 4x16->8x8 PACK and Signed Saturate
+	(packs source and dest fields into dest in that order)
+*/
+#define	packssdw_m2r(var, reg)	mmx_m2r(packssdw, var, reg)
+#define	packssdw_r2r(regs, regd) mmx_r2r(packssdw, regs, regd)
+#define	packssdw(vars, vard)	mmx_m2m(packssdw, vars, vard)
+
+#define	packsswb_m2r(var, reg)	mmx_m2r(packsswb, var, reg)
+#define	packsswb_r2r(regs, regd) mmx_r2r(packsswb, regs, regd)
+#define	packsswb(vars, vard)	mmx_m2m(packsswb, vars, vard)
+
+
+/*	4x16->8x8 PACK and Unsigned Saturate
+	(packs source and dest fields into dest in that order)
+*/
+#define	packuswb_m2r(var, reg)	mmx_m2r(packuswb, var, reg)
+#define	packuswb_r2r(regs, regd) mmx_r2r(packuswb, regs, regd)
+#define	packuswb(vars, vard)	mmx_m2m(packuswb, vars, vard)
+
+
+/*	2x32->1x64, 4x16->2x32, and 8x8->4x16 UNPaCK Low
+	(interleaves low half of dest with low half of source
+	 as padding in each result field)
+*/
+#define	punpckldq_m2r(var, reg)	mmx_m2r(punpckldq, var, reg)
+#define	punpckldq_r2r(regs, regd) mmx_r2r(punpckldq, regs, regd)
+#define	punpckldq(vars, vard)	mmx_m2m(punpckldq, vars, vard)
+
+#define	punpcklwd_m2r(var, reg)	mmx_m2r(punpcklwd, var, reg)
+#define	punpcklwd_r2r(regs, regd) mmx_r2r(punpcklwd, regs, regd)
+#define	punpcklwd(vars, vard)	mmx_m2m(punpcklwd, vars, vard)
+
+#define	punpcklbw_m2r(var, reg)	mmx_m2r(punpcklbw, var, reg)
+#define	punpcklbw_r2r(regs, regd) mmx_r2r(punpcklbw, regs, regd)
+#define	punpcklbw(vars, vard)	mmx_m2m(punpcklbw, vars, vard)
+
+
+/*	2x32->1x64, 4x16->2x32, and 8x8->4x16 UNPaCK High
+	(interleaves high half of dest with high half of source
+	 as padding in each result field)
+*/
+#define	punpckhdq_m2r(var, reg)	mmx_m2r(punpckhdq, var, reg)
+#define	punpckhdq_r2r(regs, regd) mmx_r2r(punpckhdq, regs, regd)
+#define	punpckhdq(vars, vard)	mmx_m2m(punpckhdq, vars, vard)
+
+#define	punpckhwd_m2r(var, reg)	mmx_m2r(punpckhwd, var, reg)
+#define	punpckhwd_r2r(regs, regd) mmx_r2r(punpckhwd, regs, regd)
+#define	punpckhwd(vars, vard)	mmx_m2m(punpckhwd, vars, vard)
+
+#define	punpckhbw_m2r(var, reg)	mmx_m2r(punpckhbw, var, reg)
+#define	punpckhbw_r2r(regs, regd) mmx_r2r(punpckhbw, regs, regd)
+#define	punpckhbw(vars, vard)	mmx_m2m(punpckhbw, vars, vard)
+
+
+/*	Empty MMx State
+	(used to clean-up when going from mmx to float use
+	 of the registers that are shared by both; note that
+	 there is no float-to-mmx operation needed, because
+	 only the float tag word info is corruptible)
+*/
+#ifdef	MMX_TRACE
+
+#define	emms() \
+	{ \
+		fprintf(stderr, "emms()\n"); \
+		__asm__ __volatile__ ("emms"); \
+	}
+
+#else
+
+#define	emms()			__asm__ __volatile__ ("emms")
+
+#endif
+
+#endif
+
diff --git a/libavcodec/i386/sad_mmx.s b/libavcodec/i386/sad_mmx.s
new file mode 100644
index 0000000000..b512639c25
--- /dev/null
+++ b/libavcodec/i386/sad_mmx.s
@@ -0,0 +1,798 @@
+;  MMX/SSE optimized routines for SAD of 16*16 macroblocks
+;	Copyright (C) Juan J. Sierralta P. <juanjo@atmlab.utfsm.cl>
+;
+;  dist1_* Original Copyright (C) 2000 Chris Atenasio <chris@crud.net>
+;  Enhancements and rest Copyright (C) 2000 Andrew Stevens <as@comlab.ox.ac.uk>
+
+;
+;  This program is free software; you can redistribute it and/or
+;  modify it under the terms of the GNU General Public License
+;  as published by the Free Software Foundation; either version 2
+;  of the License, or (at your option) any later version.
+;
+;  This program is distributed in the hope that it will be useful,
+;  but WITHOUT ANY WARRANTY; without even the implied warranty of
+;  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+;  GNU General Public License for more details.
+;
+;  You should have received a copy of the GNU General Public License
+;  along with this program; if not, write to the Free Software
+;  Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.
+;
+
+global pix_abs16x16_mmx
+
+; int  pix_abs16x16_mmx(unsigned char *pix1,unsigned char *pix2, int lx, int h);
+; esi = p1 (init:		blk1)
+; edi = p2 (init:		blk2)
+; ecx = rowsleft (init:	 h)
+; edx = lx;
+
+; mm0 = distance accumulators (4 words)
+; mm1 = distance accumulators (4 words) 
+; mm2 = temp 
+; mm3 = temp
+; mm4 = temp
+; mm5 = temp 
+; mm6 = 0
+; mm7 = temp
+
+
+align 32
+pix_abs16x16_mmx:
+	push				ebp					; save frame pointer
+	mov				ebp, esp
+
+	push				ebx					; Saves registers (called saves convention in
+	push				ecx					; x86 GCC it seems)
+	push				edx					; 
+	push				esi
+	push				edi
+		
+	pxor				mm0, mm0				; zero acculumators
+	pxor				mm1, mm1
+	pxor				mm6, mm6
+	mov				esi, [ebp+8]		; get pix1
+	mov				edi, [ebp+12]		; get pix2
+	mov				edx, [ebp+16]		; get lx
+	mov				ecx, [ebp+20]		; get rowsleft
+	jmp				.nextrow
+align 32
+
+.nextrow:
+	; First 8 bytes of the row
+	
+	movq				mm4, [edi]	; load first 8 bytes of pix2 row 
+	movq				mm5, [esi]	; load first 8 bytes of pix1 row
+	movq				mm3, mm4		; mm4 := abs(mm4-mm5)
+	movq				mm2,[esi+8]	; load last 8 bytes of pix1 row	
+	psubusb			mm4, mm5
+	movq				mm7,[edi+8]	; load last 8 bytes of pix2 row 		
+	psubusb			mm5, mm3
+	por				mm4, mm5
+	
+	; Last 8 bytes of the row	
+
+	movq				mm3, mm7		; mm7 := abs(mm7-mm2)
+	psubusb			mm7, mm2
+	psubusb			mm2, mm3
+	por				mm7, mm2
+	
+	; Now mm4 and mm7 have 16 absdiffs to add
+	
+	; First 8 bytes of the row2
+	
+	
+	add				edi, edx
+	movq				mm2, [edi]	; load first 8 bytes of pix2 row 
+	add				esi, edx
+	movq				mm5, [esi]	; load first 8 bytes of pix1 row
+	
+	
+	
+	movq				mm3, mm2		; mm2 := abs(mm2-mm5)
+	psubusb			mm2, mm5
+	movq				mm6,[esi+8]	; load last 8 bytes of pix1 row
+	psubusb			mm5, mm3
+	por				mm2, mm5
+		
+	; Last 8 bytes of the row2
+	
+	movq				mm5,[edi+8]	; load last 8 bytes of pix2 row 
+	
+	
+	movq				mm3, mm5		; mm5 := abs(mm5-mm6)
+	psubusb			mm5, mm6
+	psubusb			mm6, mm3
+	por				mm5, mm6
+
+	; Now mm2, mm4, mm5, mm7 have 32 absdiffs
+	
+	movq				mm3, mm7
+	
+	pxor				mm6, mm6		; Zero mm6
+		
+	punpcklbw		mm3, mm6		; Unpack to words and add
+	punpckhbw		mm7, mm6
+	paddusw			mm7, mm3
+	
+	movq				mm3, mm5
+	
+	punpcklbw		mm3, mm6		; Unpack to words and add
+	punpckhbw		mm5, mm6
+	paddusw			mm5, mm3
+		
+	paddusw			mm0, mm7		; Add to the acumulator (mm0)
+	paddusw			mm1, mm5		; Add to the acumulator (mm1)
+		
+	movq				mm3, mm4
+	
+	punpcklbw		mm3, mm6		; Unpack to words and add
+	punpckhbw		mm4, mm6
+	movq				mm5, mm2
+	paddusw			mm4, mm3
+
+	
+	
+	punpcklbw		mm5, mm6		; Unpack to words and add
+	punpckhbw		mm2, mm6
+	paddusw			mm2, mm5
+
+	; Loop termination
+	
+	add				esi, edx		; update pointers to next row
+	paddusw			mm0, mm4		; Add to the acumulator (mm0)
+	add				edi, edx		
+	sub				ecx,2
+	paddusw			mm1, mm2		; Add to the acumulator (mm1)
+	test				ecx, ecx		; check rowsleft
+	jnz				near .nextrow
+	
+	paddusw			mm0, mm1
+	movq				mm2, mm0		; Copy mm0 to mm2
+	psrlq				mm2, 32
+	paddusw			mm0, mm2		; Add 
+	movq				mm3, mm0
+	psrlq				mm3, 16
+	paddusw			mm0, mm3
+	movd				eax, mm0		; Store return value
+	and				eax, 0xffff
+
+	pop edi
+	pop esi	
+	pop edx			
+	pop ecx			
+	pop ebx			
+
+	pop ebp							; restore stack pointer
+
+	;emms								; clear mmx registers
+	ret								; return
+
+global pix_abs16x16_sse
+
+; int  pix_abs16x16_mmx(unsigned char *pix1,unsigned char *pix2, int lx, int h);
+; esi = p1 (init:		blk1)
+; edi = p2 (init:		blk2)
+; ecx = rowsleft (init:	 h)
+; edx = lx;
+
+; mm0 = distance accumulators (4 words)
+; mm1 = distance accumulators (4 words) 
+; mm2 = temp 
+; mm3 = temp
+; mm4 = temp
+; mm5 = temp 
+; mm6 = temp
+; mm7 = temp
+
+
+align 32
+pix_abs16x16_sse:
+	push				ebp					; save frame pointer
+	mov				ebp, esp
+
+	push				ebx					; Saves registers (called saves convention in
+	push				ecx					; x86 GCC it seems)
+	push				edx					; 
+	push				esi
+	push				edi
+		
+	pxor				mm0, mm0				; zero acculumators
+	pxor				mm1, mm1
+	mov				esi, [ebp+8]		; get pix1
+	mov				edi, [ebp+12]		; get pix2
+	mov				edx, [ebp+16]		; get lx
+	mov				ecx, [ebp+20]		; get rowsleft
+	jmp				.next4row
+align 32
+
+.next4row:
+	; First row
+	
+	movq				mm4, [edi]		; load first 8 bytes of pix2 row 
+	movq				mm5, [edi+8]	; load last 8 bytes of pix2 row
+	psadbw			mm4, [esi]		; SAD of first 8 bytes
+	psadbw			mm5, [esi+8]	; SAD of last 8 bytes
+	paddw				mm0, mm4			; Add to acumulators
+	paddw				mm1, mm5
+		
+	; Second row	
+
+	add				edi, edx;
+	add				esi, edx;
+	
+	movq				mm6, [edi]		; load first 8 bytes of pix2 row 
+	movq				mm7, [edi+8]	; load last 8 bytes of pix2 row
+	psadbw			mm6, [esi]		; SAD of first 8 bytes
+	psadbw			mm7, [esi+8]	; SAD of last 8 bytes
+	paddw				mm0, mm6			; Add to acumulators
+	paddw				mm1, mm7
+		
+	; Third row
+	
+	add				edi, edx;
+	add				esi, edx;
+	
+	movq				mm4, [edi]		; load first 8 bytes of pix2 row 
+	movq				mm5, [edi+8]	; load last 8 bytes of pix2 row
+	psadbw			mm4, [esi]		; SAD of first 8 bytes
+	psadbw			mm5, [esi+8]	; SAD of last 8 bytes
+	paddw				mm0, mm4			; Add to acumulators
+	paddw				mm1, mm5
+		
+	; Fourth row	
+
+	add				edi, edx;
+	add				esi, edx;
+	
+	movq				mm6, [edi]		; load first 8 bytes of pix2 row 
+	movq				mm7, [edi+8]	; load last 8 bytes of pix2 row
+	psadbw			mm6, [esi]		; SAD of first 8 bytes
+	psadbw			mm7, [esi+8]	; SAD of last 8 bytes
+	paddw				mm0, mm6			; Add to acumulators
+	paddw				mm1, mm7
+	
+	; Loop termination
+	
+	add				esi, edx		; update pointers to next row
+	add				edi, edx		
+	sub				ecx,4
+	test				ecx, ecx		; check rowsleft
+	jnz				near .next4row
+	
+	paddd				mm0, mm1		; Sum acumulators
+	movd				eax, mm0		; Store return value
+
+	pop edi
+	pop esi	
+	pop edx			
+	pop ecx			
+	pop ebx			
+
+	pop ebp							; restore stack pointer
+
+	;emms								; clear mmx registers
+	ret								; return
+		
+global pix_abs16x16_x2_mmx
+
+; int  pix_abs16x16_x2_mmx(unsigned char *pix1,unsigned char *pix2, int lx, int h);
+; esi = p1 (init:		blk1)
+; edi = p2 (init:		blk2)
+; ecx = rowsleft (init:	 h)
+; edx = lx;
+
+; mm0 = distance accumulators (4 words)
+; mm1 = distance accumulators (4 words) 
+; mm2 = temp 
+; mm3 = temp
+; mm4 = temp
+; mm5 = temp 
+; mm6 = 0
+; mm7 = temp
+
+
+align 32
+pix_abs16x16_x2_mmx:
+	push				ebp					; save frame pointer
+	mov				ebp, esp
+
+	push				ebx					; Saves registers (called saves convention in
+	push				ecx					; x86 GCC it seems)
+	push				edx					; 
+	push				esi
+	push				edi
+		
+	pxor				mm0, mm0				; zero acculumators
+	pxor				mm1, mm1
+	pxor				mm6, mm6
+	mov				esi, [ebp+8]		; get pix1
+	mov				edi, [ebp+12]		; get pix2
+	mov				edx, [ebp+16]		; get lx
+	mov				ecx, [ebp+20]		; get rowsleft
+	jmp				.nextrow_x2
+align 32
+
+.nextrow_x2:
+	; First 8 bytes of the row
+	
+	movq				mm4, [edi]			; load first 8 bytes of pix2 row 
+	movq				mm5, [edi+1]		; load bytes 1-8 of pix2 row
+
+	movq				mm2, mm4		; copy mm4 on mm2
+	movq				mm3, mm5		; copy mm5 on mm3
+	punpcklbw		mm4, mm6		; first 4 bytes of [edi] on mm4
+	punpcklbw		mm5, mm6		; first 4 bytes of [edi+1] on mm5
+	paddusw			mm4, mm5		; mm4 := first 4 bytes interpolated in words
+	psrlw				mm4, 1
+
+	punpckhbw		mm2, mm6		; last 4 bytes of [edi] on mm2
+	punpckhbw		mm3, mm6		; last 4 bytes of [edi+1] on mm3
+	paddusw			mm2, mm3		; mm2 := last 4 bytes interpolated in words
+	psrlw				mm2, 1
+	
+	packuswb			mm4, mm2 	; pack 8 bytes interpolated on mm4
+	movq				mm5,[esi]	; load first 8 bytes of pix1 row
+	
+	movq				mm3, mm4		; mm4 := abs(mm4-mm5)
+	psubusb			mm4, mm5
+	psubusb			mm5, mm3
+	por				mm4, mm5
+		
+	; Last 8 bytes of the row	
+
+	movq	mm7, [edi+8]			; load last 8 bytes of pix2 row 
+	movq	mm5, [edi+9]			; load bytes 10-17 of pix2 row
+
+	movq				mm2, mm7		; copy mm7 on mm2
+	movq				mm3, mm5		; copy mm5 on mm3
+	punpcklbw		mm7, mm6		; first 4 bytes of [edi+8] on mm7
+	punpcklbw		mm5, mm6		; first 4 bytes of [edi+9] on mm5
+	paddusw			mm7, mm5		; mm1 := first 4 bytes interpolated in words
+	psrlw				mm7, 1
+
+	punpckhbw		mm2, mm6		; last 4 bytes of [edi] on mm2
+	punpckhbw		mm3, mm6		; last 4 bytes of [edi+1] on mm3
+	paddusw			mm2, mm3		; mm2 := last 4 bytes interpolated in words
+	psrlw				mm2, 1
+	
+	packuswb			mm7, mm2 	; pack 8 bytes interpolated on mm1
+	movq				mm5,[esi+8]	; load last 8 bytes of pix1 row
+	
+	movq				mm3, mm7		; mm7 := abs(mm1-mm5)
+	psubusb			mm7, mm5
+	psubusb			mm5, mm3
+	por				mm7, mm5
+	
+	; Now mm4 and mm7 have 16 absdiffs to add
+	
+	movq				mm3, mm4		; Make copies of these bytes
+	movq				mm2, mm7
+		
+	punpcklbw		mm4, mm6		; Unpack to words and add
+	punpcklbw		mm7, mm6
+	paddusw			mm4, mm7
+	paddusw			mm0, mm4		; Add to the acumulator (mm0)
+	
+	punpckhbw		mm3, mm6		; Unpack to words and add
+	punpckhbw		mm2, mm6
+	paddusw			mm3, mm2
+	paddusw			mm1, mm3		; Add to the acumulator (mm1)
+	
+	; Loop termination
+	
+	add				esi, edx		; update pointers to next row
+	add				edi, edx		
+			
+	sub				ecx,1
+	test				ecx, ecx		; check rowsleft
+	jnz				near .nextrow_x2
+	
+	paddusw			mm0, mm1 
+	
+	movq				mm1, mm0		; Copy mm0 to mm1
+	psrlq				mm1, 32
+	paddusw			mm0, mm1		; Add 
+	movq				mm2, mm0
+	psrlq				mm2, 16
+	paddusw			mm0, mm2
+	movd				eax, mm0		; Store return value
+	and				eax, 0xffff
+
+	pop edi
+	pop esi	
+	pop edx			
+	pop ecx			
+	pop ebx			
+
+	pop ebp							; restore stack pointer
+
+	emms								; clear mmx registers
+	ret								; return
+	
+global pix_abs16x16_y2_mmx
+
+; int  pix_abs16x16_y2_mmx(unsigned char *pix1,unsigned char *pix2, int lx, int h);
+; esi = p1 (init:		blk1)
+; edi = p2 (init:		blk2)
+; ebx = p2 + lx
+; ecx = rowsleft (init:	 h)
+; edx = lx;
+
+; mm0 = distance accumulators (4 words)
+; mm1 = distance accumulators (4 words) 
+; mm2 = temp 
+; mm3 = temp
+; mm4 = temp
+; mm5 = temp 
+; mm6 = 0
+; mm7 = temp
+
+
+align 32
+pix_abs16x16_y2_mmx:
+	push				ebp					; save frame pointer
+	mov				ebp, esp
+
+	push				ebx					; Saves registers (called saves convention in
+	push				ecx					; x86 GCC it seems)
+	push				edx					; 
+	push				esi
+	push				edi
+		
+	pxor				mm0, mm0				; zero acculumators
+	pxor				mm1, mm1
+	pxor				mm6, mm6
+	mov				esi, [ebp+8]		; get pix1
+	mov				edi, [ebp+12]		; get pix2
+	mov				edx, [ebp+16]		; get lx
+	mov				ecx, [ebp+20]		; get rowsleft
+	mov				ebx, edi
+	add				ebx, edx
+	jmp				.nextrow_y2
+align 32
+
+.nextrow_y2:
+	; First 8 bytes of the row
+	
+	movq				mm4, [edi]			; load first 8 bytes of pix2 row 
+	movq				mm5, [ebx]			; load bytes 1-8 of pix2 row
+
+	movq				mm2, mm4		; copy mm4 on mm2
+	movq				mm3, mm5		; copy mm5 on mm3
+	punpcklbw		mm4, mm6		; first 4 bytes of [edi] on mm4
+	punpcklbw		mm5, mm6		; first 4 bytes of [ebx] on mm5
+	paddusw			mm4, mm5		; mm4 := first 4 bytes interpolated in words
+	psrlw				mm4, 1
+
+	punpckhbw		mm2, mm6		; last 4 bytes of [edi] on mm2
+	punpckhbw		mm3, mm6		; last 4 bytes of [edi+1] on mm3
+	paddusw			mm2, mm3		; mm2 := last 4 bytes interpolated in words
+	psrlw				mm2, 1
+	
+	packuswb			mm4, mm2 	; pack 8 bytes interpolated on mm4
+	movq				mm5,[esi]	; load first 8 bytes of pix1 row
+	
+	movq				mm3, mm4		; mm4 := abs(mm4-mm5)
+	psubusb			mm4, mm5
+	psubusb			mm5, mm3
+	por				mm4, mm5
+		
+	; Last 8 bytes of the row	
+
+	movq	mm7, [edi+8]			; load last 8 bytes of pix2 row 
+	movq	mm5, [ebx+8]			; load bytes 10-17 of pix2 row
+
+	movq				mm2, mm7		; copy mm7 on mm2
+	movq				mm3, mm5		; copy mm5 on mm3
+	punpcklbw		mm7, mm6		; first 4 bytes of [edi+8] on mm7
+	punpcklbw		mm5, mm6		; first 4 bytes of [ebx+8] on mm5
+	paddusw			mm7, mm5		; mm1 := first 4 bytes interpolated in words
+	psrlw				mm7, 1
+
+	punpckhbw		mm2, mm6		; last 4 bytes of [edi+8] on mm2
+	punpckhbw		mm3, mm6		; last 4 bytes of [ebx+8] on mm3
+	paddusw			mm2, mm3		; mm2 := last 4 bytes interpolated in words
+	psrlw				mm2, 1
+	
+	packuswb			mm7, mm2 	; pack 8 bytes interpolated on mm1
+	movq				mm5,[esi+8]	; load last 8 bytes of pix1 row
+	
+	movq				mm3, mm7		; mm7 := abs(mm1-mm5)
+	psubusb			mm7, mm5
+	psubusb			mm5, mm3
+	por				mm7, mm5
+	
+	; Now mm4 and mm7 have 16 absdiffs to add
+	
+	movq				mm3, mm4		; Make copies of these bytes
+	movq				mm2, mm7
+		
+	punpcklbw		mm4, mm6		; Unpack to words and add
+	punpcklbw		mm7, mm6
+	paddusw			mm4, mm7
+	paddusw			mm0, mm4		; Add to the acumulator (mm0)
+	
+	punpckhbw		mm3, mm6		; Unpack to words and add
+	punpckhbw		mm2, mm6
+	paddusw			mm3, mm2
+	paddusw			mm1, mm3		; Add to the acumulator (mm1)
+	
+	; Loop termination
+	
+	add				esi, edx		; update pointers to next row
+	add				edi, edx		
+	add				ebx, edx		
+	sub				ecx,1
+	test				ecx, ecx		; check rowsleft
+	jnz				near .nextrow_y2
+	
+	paddusw			mm0, mm1 
+	
+	movq				mm1, mm0		; Copy mm0 to mm1
+	psrlq				mm1, 32
+	paddusw			mm0, mm1		; Add 
+	movq				mm2, mm0
+	psrlq				mm2, 16
+	paddusw			mm0, mm2
+	movd				eax, mm0		; Store return value
+	and				eax, 0xffff
+
+	pop edi
+	pop esi	
+	pop edx			
+	pop ecx			
+	pop ebx			
+
+	pop ebp							; restore stack pointer
+
+	emms								; clear mmx registers
+	ret								; return
+	
+global pix_abs16x16_xy2_mmx
+
+; int pix_abs16x16_xy2_mmx(unsigned char *p1,unsigned char *p2,int lx,int h);
+
+; esi = p1 (init:		blk1)
+; edi = p2 (init:		blk2)
+; ebx = p1+lx
+; ecx = rowsleft (init:	 h)
+; edx = lx;
+
+; mm0 = distance accumulators (4 words)
+; mm1 = bytes p2
+; mm2 = bytes p1
+; mm3 = bytes p1+lx
+; I'd love to find someplace to stash p1+1 and p1+lx+1's bytes
+; but I don't think thats going to happen in iA32-land...
+; mm4 = temp 4 bytes in words interpolating p1, p1+1
+; mm5 = temp 4 bytes in words from p2
+; mm6 = temp comparison bit mask p1,p2
+; mm7 = temp comparison bit mask p2,p1
+
+
+align 32
+pix_abs16x16_xy2_mmx:
+	push ebp		; save stack pointer
+	mov ebp, esp	; so that we can do this
+
+	push ebx		; Saves registers (called saves convention in
+	push ecx		; x86 GCC it seems)
+	push edx		; 
+	push esi
+	push edi
+		
+	pxor mm0, mm0				; zero acculumators
+
+	mov esi, [ebp+12]			; get p1
+	mov edi, [ebp+8]			; get p2
+	mov edx, [ebp+16]			; get lx
+	mov ecx, [ebp+20]			; rowsleft := h
+	mov ebx, esi
+    add ebx, edx		
+	jmp .nextrowmm11					; snap to it
+align 32
+.nextrowmm11:
+
+		;; 
+		;; First 8 bytes of row
+		;; 
+		
+		;; First 4 bytes of 8
+
+	movq mm4, [esi]             ; mm4 := first 4 bytes p1
+	pxor mm7, mm7
+	movq mm2, mm4				;  mm2 records all 8 bytes
+	punpcklbw mm4, mm7            ;  First 4 bytes p1 in Words...
+	
+	movq mm6, [ebx]			    ;  mm6 := first 4 bytes p1+lx
+	movq mm3, mm6               ;  mm3 records all 8 bytes
+	punpcklbw mm6, mm7
+	paddw mm4, mm6              
+
+
+	movq mm5, [esi+1]			; mm5 := first 4 bytes p1+1
+	punpcklbw mm5, mm7            ;  First 4 bytes p1 in Words...
+	paddw mm4, mm5		
+	movq mm6, [ebx+1]           ;  mm6 := first 4 bytes p1+lx+1
+	punpcklbw mm6, mm7
+	paddw mm4, mm6
+
+	psrlw mm4, 2	            ; mm4 := First 4 bytes interpolated in words
+		
+	movq mm5, [edi]				; mm5:=first 4 bytes of p2 in words
+	movq mm1, mm5
+	punpcklbw mm5, mm7
+			
+	movq  mm7,mm4
+	pcmpgtw mm7,mm5		; mm7 := [i : W0..3,mm4>mm5]
+
+	movq  mm6,mm4		; mm6 := [i : W0..3, (mm4-mm5)*(mm4-mm5 > 0)]
+ 	psubw mm6,mm5
+	pand  mm6, mm7
+
+	paddw mm0, mm6				; Add to accumulator
+
+	movq  mm6,mm5       ; mm6 := [i : W0..3,mm5>mm4]
+	pcmpgtw mm6,mm4	    
+ 	psubw mm5,mm4		; mm5 := [i : B0..7, (mm5-mm4)*(mm5-mm4 > 0)]
+	pand  mm5, mm6		
+
+	paddw mm0, mm5				; Add to accumulator
+
+		;; Second 4 bytes of 8
+	
+	movq mm4, mm2		    ; mm4 := Second 4 bytes p1 in words
+	pxor  mm7, mm7
+	punpckhbw mm4, mm7			
+	movq mm6, mm3			; mm6 := Second 4 bytes p1+1 in words  
+	punpckhbw mm6, mm7
+	paddw mm4, mm6          
+
+	movq mm5, [esi+1]			; mm5 := first 4 bytes p1+1
+	punpckhbw mm5, mm7          ;  First 4 bytes p1 in Words...
+	paddw mm4, mm5
+	movq mm6, [ebx+1]           ;  mm6 := first 4 bytes p1+lx+1
+	punpckhbw mm6, mm7
+	paddw mm4, mm6
+
+	psrlw mm4, 2	            ; mm4 := First 4 bytes interpolated in words
+		
+	movq mm5, mm1			; mm5:= second 4 bytes of p2 in words
+	punpckhbw mm5, mm7
+			
+	movq  mm7,mm4
+	pcmpgtw mm7,mm5		; mm7 := [i : W0..3,mm4>mm5]
+
+	movq  mm6,mm4		; mm6 := [i : W0..3, (mm4-mm5)*(mm4-mm5 > 0)]
+ 	psubw mm6,mm5
+	pand  mm6, mm7
+
+	paddw mm0, mm6				; Add to accumulator
+
+	movq  mm6,mm5       ; mm6 := [i : W0..3,mm5>mm4]
+	pcmpgtw mm6,mm4	    
+ 	psubw mm5,mm4		; mm5 := [i : B0..7, (mm5-mm4)*(mm5-mm4 > 0)]
+	pand  mm5, mm6		
+
+	paddw mm0, mm5				; Add to accumulator
+
+
+ 		;;
+		;; Second 8 bytes of row
+		;; 
+		;; First 4 bytes of 8
+
+	movq mm4, [esi+8]             ; mm4 := first 4 bytes p1+8
+	pxor mm7, mm7
+	movq mm2, mm4				;  mm2 records all 8 bytes
+	punpcklbw mm4, mm7            ;  First 4 bytes p1 in Words...
+	
+	movq mm6, [ebx+8]			    ;  mm6 := first 4 bytes p1+lx+8
+	movq mm3, mm6               ;  mm3 records all 8 bytes
+	punpcklbw mm6, mm7
+	paddw mm4, mm6              
+
+
+	movq mm5, [esi+9]			; mm5 := first 4 bytes p1+9
+	punpcklbw mm5, mm7            ;  First 4 bytes p1 in Words...
+	paddw mm4, mm5
+	movq mm6, [ebx+9]           ;  mm6 := first 4 bytes p1+lx+9
+	punpcklbw mm6, mm7
+	paddw mm4, mm6
+
+	psrlw mm4, 2	            ; mm4 := First 4 bytes interpolated in words
+		
+	movq mm5, [edi+8]				; mm5:=first 4 bytes of p2+8 in words
+	movq mm1, mm5
+	punpcklbw mm5, mm7
+			
+	movq  mm7,mm4
+	pcmpgtw mm7,mm5		; mm7 := [i : W0..3,mm4>mm5]
+
+	movq  mm6,mm4		; mm6 := [i : W0..3, (mm4-mm5)*(mm4-mm5 > 0)]
+ 	psubw mm6,mm5
+	pand  mm6, mm7
+
+	paddw mm0, mm6				; Add to accumulator
+
+	movq  mm6,mm5       ; mm6 := [i : W0..3,mm5>mm4]
+	pcmpgtw mm6,mm4	    
+ 	psubw mm5,mm4		; mm5 := [i : B0..7, (mm5-mm4)*(mm5-mm4 > 0)]
+	pand  mm5, mm6		
+
+	paddw mm0, mm5				; Add to accumulator
+
+		;; Second 4 bytes of 8
+	
+	movq mm4, mm2		    ; mm4 := Second 4 bytes p1 in words
+	pxor  mm7, mm7
+	punpckhbw mm4, mm7			
+	movq mm6, mm3			; mm6 := Second 4 bytes p1+1 in words  
+	punpckhbw mm6, mm7
+	paddw mm4, mm6          
+
+	movq mm5, [esi+9]			; mm5 := first 4 bytes p1+1
+	punpckhbw mm5, mm7          ;  First 4 bytes p1 in Words...
+	paddw mm4, mm5	
+	movq mm6, [ebx+9]           ;  mm6 := first 4 bytes p1+lx+1
+	punpckhbw mm6, mm7
+	paddw mm4, mm6
+		
+	psrlw mm4, 2	            ; mm4 := First 4 bytes interpolated in words
+
+	movq mm5, mm1			; mm5:= second 4 bytes of p2 in words
+	punpckhbw mm5, mm7
+			
+	movq  mm7,mm4
+	pcmpgtw mm7,mm5		; mm7 := [i : W0..3,mm4>mm5]
+
+	movq  mm6,mm4		; mm6 := [i : W0..3, (mm4-mm5)*(mm4-mm5 > 0)]
+ 	psubw mm6,mm5
+	pand  mm6, mm7
+
+	paddw mm0, mm6				; Add to accumulator
+
+	movq  mm6,mm5       ; mm6 := [i : W0..3,mm5>mm4]
+	pcmpgtw mm6,mm4	    
+ 	psubw mm5,mm4		; mm5 := [i : B0..7, (mm5-mm4)*(mm5-mm4 > 0)]
+	pand  mm5, mm6		
+
+	paddw mm0, mm5				; Add to accumulator
+
+
+		;;
+		;;	Loop termination condition... and stepping
+		;;		
+
+	add esi, edx		; update pointer to next row
+	add edi, edx		; ditto
+	add ebx, edx
+
+	sub  ecx,1
+	test ecx, ecx		; check rowsleft
+	jnz near .nextrowmm11
+		
+		;; Sum the Accumulators
+	movq  mm4, mm0
+	psrlq mm4, 32
+	paddw mm0, mm4
+	movq  mm6, mm0
+	psrlq mm6, 16
+	paddw mm0, mm6
+	movd eax, mm0		; store return value
+	and  eax, 0xffff
+		
+	pop edi
+	pop esi	
+	pop edx			
+	pop ecx			
+	pop ebx			
+
+	pop ebp			; restore stack pointer
+
+	emms			; clear mmx registers
+	ret			; we now return you to your regular programming
+
+
diff --git a/libavcodec/imgconvert.c b/libavcodec/imgconvert.c
new file mode 100644
index 0000000000..01289d4d83
--- /dev/null
+++ b/libavcodec/imgconvert.c
@@ -0,0 +1,214 @@
+/*
+ * Misc image convertion routines
+ * Copyright (c) 2001 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include "avcodec.h"
+
+/* XXX: totally non optimized */
+
+static void yuv422_to_yuv420p(UINT8 *lum, UINT8 *cb, UINT8 *cr,
+                              UINT8 *src, int width, int height)
+{
+    int x, y;
+    UINT8 *p = src;
+
+    for(y=0;y<height;y+=2) {
+        for(x=0;x<width;x+=2) {
+            lum[0] = p[0];
+            cb[0] = p[1];
+            lum[1] = p[2];
+            cr[0] = p[3];
+            p += 4;
+            lum += 2;
+            cb++;
+            cr++;
+        }
+        for(x=0;x<width;x+=2) {
+            lum[0] = p[0];
+            lum[1] = p[2];
+            p += 4;
+            lum += 2;
+        }
+    }
+}
+
+#define SCALEBITS 8
+#define ONE_HALF  (1 << (SCALEBITS - 1))
+#define FIX(x)		((int) ((x) * (1L<<SCALEBITS) + 0.5))
+
+static void rgb24_to_yuv420p(UINT8 *lum, UINT8 *cb, UINT8 *cr,
+                              UINT8 *src, int width, int height)
+{
+    int wrap, wrap3, x, y;
+    int r, g, b, r1, g1, b1;
+    UINT8 *p;
+
+    wrap = width;
+    wrap3 = width * 3;
+    p = src;
+    for(y=0;y<height;y+=2) {
+        for(x=0;x<width;x+=2) {
+            r = p[0];
+            g = p[1];
+            b = p[2];
+            r1 = r;
+            g1 = g;
+            b1 = b;
+            lum[0] = (FIX(0.29900) * r + FIX(0.58700) * g + 
+                      FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
+            r = p[3];
+            g = p[4];
+            b = p[5];
+            r1 += r;
+            g1 += g;
+            b1 += b;
+            lum[1] = (FIX(0.29900) * r + FIX(0.58700) * g + 
+                      FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
+            p += wrap3;
+            lum += wrap;
+
+            r = p[0];
+            g = p[1];
+            b = p[2];
+            r1 += r;
+            g1 += g;
+            b1 += b;
+            lum[0] = (FIX(0.29900) * r + FIX(0.58700) * g + 
+                      FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
+            r = p[3];
+            g = p[4];
+            b = p[5];
+            r1 += r;
+            g1 += g;
+            b1 += b;
+            lum[1] = (FIX(0.29900) * r + FIX(0.58700) * g + 
+                      FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
+            
+            cb[0] = ((- FIX(0.16874) * r1 - FIX(0.33126) * g1 + 
+                      FIX(0.50000) * b1 + 4 * ONE_HALF - 1) >> (SCALEBITS + 2)) + 128;
+            cr[0] = ((FIX(0.50000) * r1 - FIX(0.41869) * g1 - 
+                     FIX(0.08131) * b1 + 4 * ONE_HALF - 1) >> (SCALEBITS + 2)) + 128;
+
+            cb++;
+            cr++;
+            p += -wrap3 + 2 * 3;
+            lum += -wrap + 2;
+        }
+        p += wrap3;
+        lum += wrap;
+    }
+}
+
+static void bgr24_to_yuv420p(UINT8 *lum, UINT8 *cb, UINT8 *cr,
+                              UINT8 *src, int width, int height)
+{
+    int wrap, wrap3, x, y;
+    int r, g, b, r1, g1, b1;
+    UINT8 *p;
+
+    wrap = width;
+    wrap3 = width * 3;
+    p = src;
+    for(y=0;y<height;y+=2) {
+        for(x=0;x<width;x+=2) {
+            b = p[0];
+            g = p[1];
+            r = p[2];
+            r1 = r;
+            g1 = g;
+            b1 = b;
+            lum[0] = (FIX(0.29900) * r + FIX(0.58700) * g + 
+                      FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
+            b = p[3];
+            g = p[4];
+            r = p[5];
+            r1 += r;
+            g1 += g;
+            b1 += b;
+            lum[1] = (FIX(0.29900) * r + FIX(0.58700) * g + 
+                      FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
+            p += wrap3;
+            lum += wrap;
+
+            b = p[0];
+            g = p[1];
+            r = p[2];
+            r1 += r;
+            g1 += g;
+            b1 += b;
+            lum[0] = (FIX(0.29900) * r + FIX(0.58700) * g + 
+                      FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
+            b = p[3];
+            g = p[4];
+            r = p[5];
+            r1 += r;
+            g1 += g;
+            b1 += b;
+            lum[1] = (FIX(0.29900) * r + FIX(0.58700) * g + 
+                      FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
+            
+            cb[0] = ((- FIX(0.16874) * r1 - FIX(0.33126) * g1 + 
+                      FIX(0.50000) * b1 + 4 * ONE_HALF - 1) >> (SCALEBITS + 2)) + 128;
+            cr[0] = ((FIX(0.50000) * r1 - FIX(0.41869) * g1 - 
+                     FIX(0.08131) * b1 + 4 * ONE_HALF - 1) >> (SCALEBITS + 2)) + 128;
+
+            cb++;
+            cr++;
+            p += -wrap3 + 2 * 3;
+            lum += -wrap + 2;
+        }
+        p += wrap3;
+        lum += wrap;
+    }
+}
+
+int img_convert_to_yuv420(UINT8 *img_out, UINT8 *img, 
+                          int pix_fmt, int width, int height)
+{
+    UINT8 *pict;
+    int size, size_out;
+    UINT8 *picture[3];
+
+    pict = img_out;
+    size = width * height;
+    size_out = (size * 3) / 2;
+    picture[0] = pict;
+    picture[1] = pict + size;
+    picture[2] = picture[1] + (size / 4);
+
+    switch(pix_fmt) {
+    case PIX_FMT_YUV420P:
+        memcpy(pict, img, size_out);
+        break;
+    case PIX_FMT_YUV422:
+        yuv422_to_yuv420p(picture[0], picture[1], picture[2], 
+                          img, width, height);
+        break;
+    case PIX_FMT_RGB24:
+        rgb24_to_yuv420p(picture[0], picture[1], picture[2], 
+                         img, width, height);
+        break;
+    case PIX_FMT_BGR24:
+        bgr24_to_yuv420p(picture[0], picture[1], picture[2], 
+                         img, width, height);
+        break;
+    }
+    return size_out;
+}
diff --git a/libavcodec/imgresample.c b/libavcodec/imgresample.c
new file mode 100644
index 0000000000..99153013cf
--- /dev/null
+++ b/libavcodec/imgresample.c
@@ -0,0 +1,614 @@
+/*
+ * High quality image resampling with polyphase filters 
+ * Copyright (c) 2001 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <math.h>
+#include "dsputil.h"
+#include "avcodec.h"
+
+#define NB_COMPONENTS 3
+
+#define PHASE_BITS 4
+#define NB_PHASES  (1 << PHASE_BITS)
+#define NB_TAPS    4
+#define FCENTER    1  /* index of the center of the filter */
+
+#define POS_FRAC_BITS 16
+#define POS_FRAC      (1 << POS_FRAC_BITS)
+/* 6 bits precision is needed for MMX */
+#define FILTER_BITS   8
+
+#define LINE_BUF_HEIGHT (NB_TAPS * 4)
+
+struct ImgReSampleContext {
+    int iwidth, iheight, owidth, oheight;
+    int h_incr, v_incr;
+    INT16 h_filters[NB_PHASES][NB_TAPS] __align8; /* horizontal filters */
+    INT16 v_filters[NB_PHASES][NB_TAPS] __align8; /* vertical filters */
+    UINT8 *line_buf;
+};
+
+static inline int get_phase(int pos)
+{
+    return ((pos) >> (POS_FRAC_BITS - PHASE_BITS)) & ((1 << PHASE_BITS) - 1);
+}
+
+/* This function must be optimized */
+static void h_resample_fast(UINT8 *dst, int dst_width, UINT8 *src, int src_width,
+                            int src_start, int src_incr, INT16 *filters)
+{
+    int src_pos, phase, sum, i;
+    UINT8 *s;
+    INT16 *filter;
+
+    src_pos = src_start;
+    for(i=0;i<dst_width;i++) {
+#ifdef TEST
+        /* test */
+        if ((src_pos >> POS_FRAC_BITS) < 0 ||
+            (src_pos >> POS_FRAC_BITS) > (src_width - NB_TAPS))
+            abort();
+#endif
+        s = src + (src_pos >> POS_FRAC_BITS);
+        phase = get_phase(src_pos);
+        filter = filters + phase * NB_TAPS;
+#if NB_TAPS == 4
+        sum = s[0] * filter[0] +
+            s[1] * filter[1] +
+            s[2] * filter[2] +
+            s[3] * filter[3];
+#else
+        {
+            int j;
+            sum = 0;
+            for(j=0;j<NB_TAPS;j++)
+                sum += s[j] * filter[j];
+        }
+#endif
+        sum = sum >> FILTER_BITS;
+        if (sum < 0)
+            sum = 0;
+        else if (sum > 255)
+            sum = 255;
+        dst[0] = sum;
+        src_pos += src_incr;
+        dst++;
+    }
+}
+
+/* This function must be optimized */
+static void v_resample(UINT8 *dst, int dst_width, UINT8 *src, int wrap, 
+                       INT16 *filter)
+{
+    int sum, i;
+    UINT8 *s;
+
+    s = src;
+    for(i=0;i<dst_width;i++) {
+#if NB_TAPS == 4
+        sum = s[0 * wrap] * filter[0] +
+            s[1 * wrap] * filter[1] +
+            s[2 * wrap] * filter[2] +
+            s[3 * wrap] * filter[3];
+#else
+        {
+            int j;
+            UINT8 *s1 = s;
+
+            sum = 0;
+            for(j=0;j<NB_TAPS;j++) {
+                sum += s1[0] * filter[j];
+                s1 += wrap;
+            }
+        }
+#endif
+        sum = sum >> FILTER_BITS;
+        if (sum < 0)
+            sum = 0;
+        else if (sum > 255)
+            sum = 255;
+        dst[0] = sum;
+        dst++;
+        s++;
+    }
+}
+
+#ifdef CONFIG_MMX
+
+#include "i386/mmx.h"
+
+#define FILTER4(reg) \
+{\
+        s = src + (src_pos >> POS_FRAC_BITS);\
+        phase = get_phase(src_pos);\
+        filter = filters + phase * NB_TAPS;\
+        movq_m2r(*s, reg);\
+        punpcklbw_r2r(mm7, reg);\
+        movq_m2r(*filter, mm6);\
+        pmaddwd_r2r(reg, mm6);\
+        movq_r2r(mm6, reg);\
+        psrlq_i2r(32, reg);\
+        paddd_r2r(mm6, reg);\
+        psrad_i2r(FILTER_BITS, reg);\
+        src_pos += src_incr;\
+}
+
+#define DUMP(reg) movq_r2m(reg, tmp); printf(#reg "=%016Lx\n", tmp.uq);
+
+/* XXX: do four pixels at a time */
+static void h_resample_fast4_mmx(UINT8 *dst, int dst_width, UINT8 *src, int src_width,
+                                 int src_start, int src_incr, INT16 *filters)
+{
+    int src_pos, phase;
+    UINT8 *s;
+    INT16 *filter;
+    mmx_t tmp;
+    
+    src_pos = src_start;
+    pxor_r2r(mm7, mm7);
+
+    while (dst_width >= 4) {
+
+        FILTER4(mm0);
+        FILTER4(mm1);
+        FILTER4(mm2);
+        FILTER4(mm3);
+
+        packuswb_r2r(mm7, mm0);
+        packuswb_r2r(mm7, mm1);
+        packuswb_r2r(mm7, mm3);
+        packuswb_r2r(mm7, mm2);
+        movq_r2m(mm0, tmp);
+        dst[0] = tmp.ub[0];
+        movq_r2m(mm1, tmp);
+        dst[1] = tmp.ub[0];
+        movq_r2m(mm2, tmp);
+        dst[2] = tmp.ub[0];
+        movq_r2m(mm3, tmp);
+        dst[3] = tmp.ub[0];
+        dst += 4;
+        dst_width -= 4;
+    }
+    while (dst_width > 0) {
+        FILTER4(mm0);
+        packuswb_r2r(mm7, mm0);
+        movq_r2m(mm0, tmp);
+        dst[0] = tmp.ub[0];
+        dst++;
+        dst_width--;
+    }
+    emms();
+}
+
+static void v_resample4_mmx(UINT8 *dst, int dst_width, UINT8 *src, int wrap, 
+                            INT16 *filter)
+{
+    int sum, i, v;
+    UINT8 *s;
+    mmx_t tmp;
+    mmx_t coefs[4];
+    
+    for(i=0;i<4;i++) {
+        v = filter[i];
+        coefs[i].uw[0] = v;
+        coefs[i].uw[1] = v;
+        coefs[i].uw[2] = v;
+        coefs[i].uw[3] = v;
+    }
+    
+    pxor_r2r(mm7, mm7);
+    s = src;
+    while (dst_width >= 4) {
+        movq_m2r(s[0 * wrap], mm0);
+        punpcklbw_r2r(mm7, mm0);
+        movq_m2r(s[1 * wrap], mm1);
+        punpcklbw_r2r(mm7, mm1);
+        movq_m2r(s[2 * wrap], mm2);
+        punpcklbw_r2r(mm7, mm2);
+        movq_m2r(s[3 * wrap], mm3);
+        punpcklbw_r2r(mm7, mm3);
+
+        pmullw_m2r(coefs[0], mm0);
+        pmullw_m2r(coefs[1], mm1);
+        pmullw_m2r(coefs[2], mm2);
+        pmullw_m2r(coefs[3], mm3);
+
+        paddw_r2r(mm1, mm0);
+        paddw_r2r(mm3, mm2);
+        paddw_r2r(mm2, mm0);
+        psraw_i2r(FILTER_BITS, mm0);
+        
+        packuswb_r2r(mm7, mm0);
+        movq_r2m(mm0, tmp);
+
+        *(UINT32 *)dst = tmp.ud[0];
+        dst += 4;
+        s += 4;
+        dst_width -= 4;
+    }
+    while (dst_width > 0) {
+        sum = s[0 * wrap] * filter[0] +
+            s[1 * wrap] * filter[1] +
+            s[2 * wrap] * filter[2] +
+            s[3 * wrap] * filter[3];
+        sum = sum >> FILTER_BITS;
+        if (sum < 0)
+            sum = 0;
+        else if (sum > 255)
+            sum = 255;
+        dst[0] = sum;
+        dst++;
+        s++;
+        dst_width--;
+    }
+    emms();
+}
+#endif
+
+/* slow version to handle limit cases. Does not need optimisation */
+static void h_resample_slow(UINT8 *dst, int dst_width, UINT8 *src, int src_width,
+                            int src_start, int src_incr, INT16 *filters)
+{
+    int src_pos, phase, sum, j, v, i;
+    UINT8 *s, *src_end;
+    INT16 *filter;
+
+    src_end = src + src_width;
+    src_pos = src_start;
+    for(i=0;i<dst_width;i++) {
+        s = src + (src_pos >> POS_FRAC_BITS);
+        phase = get_phase(src_pos);
+        filter = filters + phase * NB_TAPS;
+        sum = 0;
+        for(j=0;j<NB_TAPS;j++) {
+            if (s < src)
+                v = src[0];
+            else if (s >= src_end)
+                v = src_end[-1];
+            else
+                v = s[0];
+            sum += v * filter[j];
+            s++;
+        }
+        sum = sum >> FILTER_BITS;
+        if (sum < 0)
+            sum = 0;
+        else if (sum > 255)
+            sum = 255;
+        dst[0] = sum;
+        src_pos += src_incr;
+        dst++;
+    }
+}
+
+static void h_resample(UINT8 *dst, int dst_width, UINT8 *src, int src_width,
+                       int src_start, int src_incr, INT16 *filters)
+{
+    int n, src_end;
+
+    if (src_start < 0) {
+        n = (0 - src_start + src_incr - 1) / src_incr;
+        h_resample_slow(dst, n, src, src_width, src_start, src_incr, filters);
+        dst += n;
+        dst_width -= n;
+        src_start += n * src_incr;
+    }
+    src_end = src_start + dst_width * src_incr;
+    if (src_end > ((src_width - NB_TAPS) << POS_FRAC_BITS)) {
+        n = (((src_width - NB_TAPS + 1) << POS_FRAC_BITS) - 1 - src_start) / 
+            src_incr;
+    } else {
+        n = dst_width;
+    }
+#ifdef CONFIG_MMX
+    if ((mm_flags & MM_MMX) && NB_TAPS == 4)
+        h_resample_fast4_mmx(dst, n, 
+                             src, src_width, src_start, src_incr, filters);
+    else
+#endif
+        h_resample_fast(dst, n, 
+                        src, src_width, src_start, src_incr, filters);
+    if (n < dst_width) {
+        dst += n;
+        dst_width -= n;
+        src_start += n * src_incr;
+        h_resample_slow(dst, dst_width, 
+                        src, src_width, src_start, src_incr, filters);
+    }
+}
+
+static void component_resample(ImgReSampleContext *s, 
+                               UINT8 *output, int owrap, int owidth, int oheight,
+                               UINT8 *input, int iwrap, int iwidth, int iheight)
+{
+    int src_y, src_y1, last_src_y, ring_y, phase_y, y1, y;
+    UINT8 *new_line, *src_line;
+
+    last_src_y = - FCENTER - 1;
+    /* position of the bottom of the filter in the source image */
+    src_y = (last_src_y + NB_TAPS) * POS_FRAC; 
+    ring_y = NB_TAPS; /* position in ring buffer */
+    for(y=0;y<oheight;y++) {
+        /* apply horizontal filter on new lines from input if needed */
+        src_y1 = src_y >> POS_FRAC_BITS;
+        while (last_src_y < src_y1) {
+            if (++ring_y >= LINE_BUF_HEIGHT + NB_TAPS)
+                ring_y = NB_TAPS;
+            last_src_y++;
+            /* handle limit conditions : replicate line (slighly
+               inefficient because we filter multiple times */
+            y1 = last_src_y;
+            if (y1 < 0) {
+                y1 = 0;
+            } else if (y1 >= iheight) {
+                y1 = iheight - 1;
+            }
+            src_line = input + y1 * iwrap;
+            new_line = s->line_buf + ring_y * owidth;
+            /* apply filter and handle limit cases correctly */
+            h_resample(new_line, owidth, 
+                       src_line, iwidth, - FCENTER * POS_FRAC, s->h_incr, 
+                       &s->h_filters[0][0]);
+            /* handle ring buffer wraping */
+            if (ring_y >= LINE_BUF_HEIGHT) {
+                memcpy(s->line_buf + (ring_y - LINE_BUF_HEIGHT) * owidth,
+                       new_line, owidth);
+            }
+        }
+        /* apply vertical filter */
+        phase_y = get_phase(src_y);
+#ifdef CONFIG_MMX
+        /* desactivated MMX because loss of precision */
+        if ((mm_flags & MM_MMX) && NB_TAPS == 4 && 0)
+            v_resample4_mmx(output, owidth, 
+                            s->line_buf + (ring_y - NB_TAPS + 1) * owidth, owidth, 
+                            &s->v_filters[phase_y][0]);
+        else
+#endif
+            v_resample(output, owidth, 
+                       s->line_buf + (ring_y - NB_TAPS + 1) * owidth, owidth, 
+                       &s->v_filters[phase_y][0]);
+            
+        src_y += s->v_incr;
+        output += owrap;
+    }
+}
+
+/* XXX: the following filter is quite naive, but it seems to suffice
+   for 4 taps */
+static void build_filter(INT16 *filter, float factor)
+{
+    int ph, i, v;
+    float x, y, tab[NB_TAPS], norm, mult;
+
+    /* if upsampling, only need to interpolate, no filter */
+    if (factor > 1.0)
+        factor = 1.0;
+
+    for(ph=0;ph<NB_PHASES;ph++) {
+        norm = 0;
+        for(i=0;i<NB_TAPS;i++) {
+            
+            x = M_PI * ((float)(i - FCENTER) - (float)ph / NB_PHASES) * factor;
+            if (x == 0)
+                y = 1.0;
+            else
+                y = sin(x) / x;
+            tab[i] = y;
+            norm += y;
+        }
+
+        /* normalize so that an uniform color remains the same */
+        mult = (float)(1 << FILTER_BITS) / norm;
+        for(i=0;i<NB_TAPS;i++) {
+            v = (int)(tab[i] * mult);
+            filter[ph * NB_TAPS + i] = v;
+        }
+    }
+}
+
+ImgReSampleContext *img_resample_init(int owidth, int oheight,
+                                      int iwidth, int iheight)
+{
+    ImgReSampleContext *s;
+
+    s = av_mallocz(sizeof(ImgReSampleContext));
+    if (!s)
+        return NULL;
+    s->line_buf = av_mallocz(owidth * (LINE_BUF_HEIGHT + NB_TAPS));
+    if (!s->line_buf) 
+        goto fail;
+    
+    s->owidth = owidth;
+    s->oheight = oheight;
+    s->iwidth = iwidth;
+    s->iheight = iheight;
+    
+    s->h_incr = (iwidth * POS_FRAC) / owidth;
+    s->v_incr = (iheight * POS_FRAC) / oheight;
+    
+    build_filter(&s->h_filters[0][0], (float)owidth / (float)iwidth);
+    build_filter(&s->v_filters[0][0], (float)oheight / (float)iheight);
+
+    return s;
+ fail:
+    free(s);
+    return NULL;
+}
+
+void img_resample(ImgReSampleContext *s, 
+                  AVPicture *output, AVPicture *input)
+{
+    int i, shift;
+
+    for(i=0;i<3;i++) {
+        shift = (i == 0) ? 0 : 1;
+        component_resample(s, output->data[i], output->linesize[i], 
+                           s->owidth >> shift, s->oheight >> shift,
+                           input->data[i], input->linesize[i], 
+                           s->iwidth >> shift, s->iheight >> shift);
+    }
+}
+
+void img_resample_close(ImgReSampleContext *s)
+{
+    free(s->line_buf);
+    free(s);
+}
+
+#ifdef TEST
+
+void *av_mallocz(int size)
+{
+    void *ptr;
+    ptr = malloc(size);
+    memset(ptr, 0, size);
+    return ptr;
+}
+
+/* input */
+#define XSIZE 256
+#define YSIZE 256
+UINT8 img[XSIZE * YSIZE];
+
+/* output */
+#define XSIZE1 512
+#define YSIZE1 512
+UINT8 img1[XSIZE1 * YSIZE1];
+UINT8 img2[XSIZE1 * YSIZE1];
+
+void save_pgm(const char *filename, UINT8 *img, int xsize, int ysize)
+{
+    FILE *f;
+    f=fopen(filename,"w");
+    fprintf(f,"P5\n%d %d\n%d\n", xsize, ysize, 255);
+    fwrite(img,1, xsize * ysize,f);
+    fclose(f);
+}
+
+static void dump_filter(INT16 *filter)
+{
+    int i, ph;
+
+    for(ph=0;ph<NB_PHASES;ph++) {
+        printf("%2d: ", ph);
+        for(i=0;i<NB_TAPS;i++) {
+            printf(" %5.2f", filter[ph * NB_TAPS + i] / 256.0);
+        }
+        printf("\n");
+    }
+}
+
+#ifdef CONFIG_MMX
+int mm_flags;
+#endif
+
+int main(int argc, char **argv)
+{
+    int x, y, v, i, xsize, ysize;
+    ImgReSampleContext *s;
+    float fact, factors[] = { 1/2.0, 3.0/4.0, 1.0, 4.0/3.0, 16.0/9.0, 2.0 };
+    char buf[256];
+
+    /* build test image */
+    for(y=0;y<YSIZE;y++) {
+        for(x=0;x<XSIZE;x++) {
+            if (x < XSIZE/2 && y < YSIZE/2) {
+                if (x < XSIZE/4 && y < YSIZE/4) {
+                    if ((x % 10) <= 6 &&
+                        (y % 10) <= 6)
+                        v = 0xff;
+                    else
+                        v = 0x00;
+                } else if (x < XSIZE/4) {
+                    if (x & 1) 
+                        v = 0xff;
+                    else 
+                        v = 0;
+                } else if (y < XSIZE/4) {
+                    if (y & 1) 
+                        v = 0xff;
+                    else 
+                        v = 0;
+                } else {
+                    if (y < YSIZE*3/8) {
+                        if ((y+x) & 1) 
+                            v = 0xff;
+                        else 
+                            v = 0;
+                    } else {
+                        if (((x+3) % 4) <= 1 &&
+                            ((y+3) % 4) <= 1)
+                            v = 0xff;
+                        else
+                            v = 0x00;
+                    }
+                }
+            } else if (x < XSIZE/2) {
+                v = ((x - (XSIZE/2)) * 255) / (XSIZE/2);
+            } else if (y < XSIZE/2) {
+                v = ((y - (XSIZE/2)) * 255) / (XSIZE/2);
+            } else {
+                v = ((x + y - XSIZE) * 255) / XSIZE;
+            }
+            img[y * XSIZE + x] = v;
+        }
+    }
+    save_pgm("/tmp/in.pgm", img, XSIZE, YSIZE);
+    for(i=0;i<sizeof(factors)/sizeof(float);i++) {
+        fact = factors[i];
+        xsize = (int)(XSIZE * fact);
+        ysize = (int)(YSIZE * fact);
+        s = img_resample_init(xsize, ysize, XSIZE, YSIZE);
+        printf("Factor=%0.2f\n", fact);
+        dump_filter(&s->h_filters[0][0]);
+        component_resample(s, img1, xsize, xsize, ysize,
+                           img, XSIZE, XSIZE, YSIZE);
+        img_resample_close(s);
+
+        sprintf(buf, "/tmp/out%d.pgm", i);
+        save_pgm(buf, img1, xsize, ysize);
+    }
+
+    /* mmx test */
+#ifdef CONFIG_MMX
+    printf("MMX test\n");
+    fact = 0.72;
+    xsize = (int)(XSIZE * fact);
+    ysize = (int)(YSIZE * fact);
+    mm_flags = MM_MMX;
+    s = img_resample_init(xsize, ysize, XSIZE, YSIZE);
+    component_resample(s, img1, xsize, xsize, ysize,
+                       img, XSIZE, XSIZE, YSIZE);
+
+    mm_flags = 0;
+    s = img_resample_init(xsize, ysize, XSIZE, YSIZE);
+    component_resample(s, img2, xsize, xsize, ysize,
+                       img, XSIZE, XSIZE, YSIZE);
+    if (memcmp(img1, img2, xsize * ysize) != 0) {
+        fprintf(stderr, "mmx error\n");
+        exit(1);
+    }
+    printf("MMX OK\n");
+#endif
+    return 0;
+}
+
+#endif
diff --git a/libavcodec/jfdctfst.c b/libavcodec/jfdctfst.c
new file mode 100644
index 0000000000..cdc3b47f99
--- /dev/null
+++ b/libavcodec/jfdctfst.c
@@ -0,0 +1,224 @@
+/*
+ * jfdctfst.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a fast, not so accurate integer implementation of the
+ * forward DCT (Discrete Cosine Transform).
+ *
+ * A 2-D DCT can be done by 1-D DCT on each row followed by 1-D DCT
+ * on each column.  Direct algorithms are also available, but they are
+ * much more complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on Arai, Agui, and Nakajima's algorithm for
+ * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
+ * Japanese, but the algorithm is described in the Pennebaker & Mitchell
+ * JPEG textbook (see REFERENCES section in file README).  The following code
+ * is based directly on figure 4-8 in P&M.
+ * While an 8-point DCT cannot be done in less than 11 multiplies, it is
+ * possible to arrange the computation so that many of the multiplies are
+ * simple scalings of the final outputs.  These multiplies can then be
+ * folded into the multiplications or divisions by the JPEG quantization
+ * table entries.  The AA&N method leaves only 5 multiplies and 29 adds
+ * to be done in the DCT itself.
+ * The primary disadvantage of this method is that with fixed-point math,
+ * accuracy is lost due to imprecise representation of the scaled
+ * quantization values.  The smaller the quantization table entry, the less
+ * precise the scaled value, so this implementation does worse with high-
+ * quality-setting files than with low-quality ones.
+ */
+
+#include <stdlib.h>
+#include <stdio.h>
+#include "common.h"
+#include "dsputil.h"
+
+#define DCTSIZE 8
+#define GLOBAL(x) x
+#define RIGHT_SHIFT(x, n) ((x) >> (n))
+#define SHIFT_TEMPS
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/* Scaling decisions are generally the same as in the LL&M algorithm;
+ * see jfdctint.c for more details.  However, we choose to descale
+ * (right shift) multiplication products as soon as they are formed,
+ * rather than carrying additional fractional bits into subsequent additions.
+ * This compromises accuracy slightly, but it lets us save a few shifts.
+ * More importantly, 16-bit arithmetic is then adequate (for 8-bit samples)
+ * everywhere except in the multiplications proper; this saves a good deal
+ * of work on 16-bit-int machines.
+ *
+ * Again to save a few shifts, the intermediate results between pass 1 and
+ * pass 2 are not upscaled, but are represented only to integral precision.
+ *
+ * A final compromise is to represent the multiplicative constants to only
+ * 8 fractional bits, rather than 13.  This saves some shifting work on some
+ * machines, and may also reduce the cost of multiplication (since there
+ * are fewer one-bits in the constants).
+ */
+
+#define CONST_BITS  8
+
+
+/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
+ * causing a lot of useless floating-point operations at run time.
+ * To get around this we use the following pre-calculated constants.
+ * If you change CONST_BITS you may want to add appropriate values.
+ * (With a reasonable C compiler, you can just rely on the FIX() macro...)
+ */
+
+#if CONST_BITS == 8
+#define FIX_0_382683433  ((INT32)   98)		/* FIX(0.382683433) */
+#define FIX_0_541196100  ((INT32)  139)		/* FIX(0.541196100) */
+#define FIX_0_707106781  ((INT32)  181)		/* FIX(0.707106781) */
+#define FIX_1_306562965  ((INT32)  334)		/* FIX(1.306562965) */
+#else
+#define FIX_0_382683433  FIX(0.382683433)
+#define FIX_0_541196100  FIX(0.541196100)
+#define FIX_0_707106781  FIX(0.707106781)
+#define FIX_1_306562965  FIX(1.306562965)
+#endif
+
+
+/* We can gain a little more speed, with a further compromise in accuracy,
+ * by omitting the addition in a descaling shift.  This yields an incorrectly
+ * rounded result half the time...
+ */
+
+#ifndef USE_ACCURATE_ROUNDING
+#undef DESCALE
+#define DESCALE(x,n)  RIGHT_SHIFT(x, n)
+#endif
+
+
+/* Multiply a DCTELEM variable by an INT32 constant, and immediately
+ * descale to yield a DCTELEM result.
+ */
+
+#define MULTIPLY(var,const)  ((DCTELEM) DESCALE((var) * (const), CONST_BITS))
+
+
+/*
+ * Perform the forward DCT on one block of samples.
+ */
+
+GLOBAL(void)
+jpeg_fdct_ifast (DCTELEM * data)
+{
+  DCTELEM tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  DCTELEM tmp10, tmp11, tmp12, tmp13;
+  DCTELEM z1, z2, z3, z4, z5, z11, z13;
+  DCTELEM *dataptr;
+  int ctr;
+  SHIFT_TEMPS
+
+  /* Pass 1: process rows. */
+
+  dataptr = data;
+  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
+    tmp0 = dataptr[0] + dataptr[7];
+    tmp7 = dataptr[0] - dataptr[7];
+    tmp1 = dataptr[1] + dataptr[6];
+    tmp6 = dataptr[1] - dataptr[6];
+    tmp2 = dataptr[2] + dataptr[5];
+    tmp5 = dataptr[2] - dataptr[5];
+    tmp3 = dataptr[3] + dataptr[4];
+    tmp4 = dataptr[3] - dataptr[4];
+    
+    /* Even part */
+    
+    tmp10 = tmp0 + tmp3;	/* phase 2 */
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    dataptr[0] = tmp10 + tmp11; /* phase 3 */
+    dataptr[4] = tmp10 - tmp11;
+    
+    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_707106781); /* c4 */
+    dataptr[2] = tmp13 + z1;	/* phase 5 */
+    dataptr[6] = tmp13 - z1;
+    
+    /* Odd part */
+
+    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp11 = tmp5 + tmp6;
+    tmp12 = tmp6 + tmp7;
+
+    /* The rotator is modified from fig 4-8 to avoid extra negations. */
+    z5 = MULTIPLY(tmp10 - tmp12, FIX_0_382683433); /* c6 */
+    z2 = MULTIPLY(tmp10, FIX_0_541196100) + z5; /* c2-c6 */
+    z4 = MULTIPLY(tmp12, FIX_1_306562965) + z5; /* c2+c6 */
+    z3 = MULTIPLY(tmp11, FIX_0_707106781); /* c4 */
+
+    z11 = tmp7 + z3;		/* phase 5 */
+    z13 = tmp7 - z3;
+
+    dataptr[5] = z13 + z2;	/* phase 6 */
+    dataptr[3] = z13 - z2;
+    dataptr[1] = z11 + z4;
+    dataptr[7] = z11 - z4;
+
+    dataptr += DCTSIZE;		/* advance pointer to next row */
+  }
+
+  /* Pass 2: process columns. */
+
+  dataptr = data;
+  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
+    tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*7];
+    tmp7 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7];
+    tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*6];
+    tmp6 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6];
+    tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*5];
+    tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
+    tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4];
+    tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];
+    
+    /* Even part */
+    
+    tmp10 = tmp0 + tmp3;	/* phase 2 */
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    dataptr[DCTSIZE*0] = tmp10 + tmp11; /* phase 3 */
+    dataptr[DCTSIZE*4] = tmp10 - tmp11;
+    
+    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_707106781); /* c4 */
+    dataptr[DCTSIZE*2] = tmp13 + z1; /* phase 5 */
+    dataptr[DCTSIZE*6] = tmp13 - z1;
+    
+    /* Odd part */
+
+    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp11 = tmp5 + tmp6;
+    tmp12 = tmp6 + tmp7;
+
+    /* The rotator is modified from fig 4-8 to avoid extra negations. */
+    z5 = MULTIPLY(tmp10 - tmp12, FIX_0_382683433); /* c6 */
+    z2 = MULTIPLY(tmp10, FIX_0_541196100) + z5; /* c2-c6 */
+    z4 = MULTIPLY(tmp12, FIX_1_306562965) + z5; /* c2+c6 */
+    z3 = MULTIPLY(tmp11, FIX_0_707106781); /* c4 */
+
+    z11 = tmp7 + z3;		/* phase 5 */
+    z13 = tmp7 - z3;
+
+    dataptr[DCTSIZE*5] = z13 + z2; /* phase 6 */
+    dataptr[DCTSIZE*3] = z13 - z2;
+    dataptr[DCTSIZE*1] = z11 + z4;
+    dataptr[DCTSIZE*7] = z11 - z4;
+
+    dataptr++;			/* advance pointer to next column */
+  }
+}
diff --git a/libavcodec/jrevdct.c b/libavcodec/jrevdct.c
new file mode 100644
index 0000000000..2ef40f38e5
--- /dev/null
+++ b/libavcodec/jrevdct.c
@@ -0,0 +1,1167 @@
+/*
+ * jrevdct.c
+ *
+ * Copyright (C) 1991, 1992, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the basic inverse-DCT transformation subroutine.
+ *
+ * This implementation is based on an algorithm described in
+ *   C. Loeffler, A. Ligtenberg and G. Moschytz, "Practical Fast 1-D DCT
+ *   Algorithms with 11 Multiplications", Proc. Int'l. Conf. on Acoustics,
+ *   Speech, and Signal Processing 1989 (ICASSP '89), pp. 988-991.
+ * The primary algorithm described there uses 11 multiplies and 29 adds.
+ * We use their alternate method with 12 multiplies and 32 adds.
+ * The advantage of this method is that no data path contains more than one
+ * multiplication; this allows a very simple and accurate implementation in
+ * scaled fixed-point arithmetic, with a minimal number of shifts.
+ * 
+ * I've made lots of modifications to attempt to take advantage of the
+ * sparse nature of the DCT matrices we're getting.  Although the logic
+ * is cumbersome, it's straightforward and the resulting code is much
+ * faster.
+ *
+ * A better way to do this would be to pass in the DCT block as a sparse
+ * matrix, perhaps with the difference cases encoded.
+ */
+#include "common.h"
+#include "dsputil.h"
+
+#define EIGHT_BIT_SAMPLES
+
+#define DCTSIZE 8
+#define DCTSIZE2 64
+
+#define GLOBAL
+
+#define RIGHT_SHIFT(x, n) ((x) >> (n))
+
+typedef DCTELEM DCTBLOCK[DCTSIZE2];
+
+#define CONST_BITS 13
+
+/*
+ * This routine is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/*
+ * A 2-D IDCT can be done by 1-D IDCT on each row followed by 1-D IDCT
+ * on each column.  Direct algorithms are also available, but they are
+ * much more complex and seem not to be any faster when reduced to code.
+ *
+ * The poop on this scaling stuff is as follows:
+ *
+ * Each 1-D IDCT step produces outputs which are a factor of sqrt(N)
+ * larger than the true IDCT outputs.  The final outputs are therefore
+ * a factor of N larger than desired; since N=8 this can be cured by
+ * a simple right shift at the end of the algorithm.  The advantage of
+ * this arrangement is that we save two multiplications per 1-D IDCT,
+ * because the y0 and y4 inputs need not be divided by sqrt(N).
+ *
+ * We have to do addition and subtraction of the integer inputs, which
+ * is no problem, and multiplication by fractional constants, which is
+ * a problem to do in integer arithmetic.  We multiply all the constants
+ * by CONST_SCALE and convert them to integer constants (thus retaining
+ * CONST_BITS bits of precision in the constants).  After doing a
+ * multiplication we have to divide the product by CONST_SCALE, with proper
+ * rounding, to produce the correct output.  This division can be done
+ * cheaply as a right shift of CONST_BITS bits.  We postpone shifting
+ * as long as possible so that partial sums can be added together with
+ * full fractional precision.
+ *
+ * The outputs of the first pass are scaled up by PASS1_BITS bits so that
+ * they are represented to better-than-integral precision.  These outputs
+ * require BITS_IN_JSAMPLE + PASS1_BITS + 3 bits; this fits in a 16-bit word
+ * with the recommended scaling.  (To scale up 12-bit sample data further, an
+ * intermediate int32 array would be needed.)
+ *
+ * To avoid overflow of the 32-bit intermediate results in pass 2, we must
+ * have BITS_IN_JSAMPLE + CONST_BITS + PASS1_BITS <= 26.  Error analysis
+ * shows that the values given below are the most effective.
+ */
+
+#ifdef EIGHT_BIT_SAMPLES
+#define PASS1_BITS  2
+#else
+#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
+#endif
+
+#define ONE	((INT32) 1)
+
+#define CONST_SCALE (ONE << CONST_BITS)
+
+/* Convert a positive real constant to an integer scaled by CONST_SCALE.
+ * IMPORTANT: if your compiler doesn't do this arithmetic at compile time,
+ * you will pay a significant penalty in run time.  In that case, figure
+ * the correct integer constant values and insert them by hand.
+ */
+
+/* Actually FIX is no longer used, we precomputed them all */
+#define FIX(x)	((INT32) ((x) * CONST_SCALE + 0.5)) 
+
+/* Descale and correctly round an INT32 value that's scaled by N bits.
+ * We assume RIGHT_SHIFT rounds towards minus infinity, so adding
+ * the fudge factor is correct for either sign of X.
+ */
+
+#define DESCALE(x,n)  RIGHT_SHIFT((x) + (ONE << ((n)-1)), n)
+
+/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
+ * For 8-bit samples with the recommended scaling, all the variable
+ * and constant values involved are no more than 16 bits wide, so a
+ * 16x16->32 bit multiply can be used instead of a full 32x32 multiply;
+ * this provides a useful speedup on many machines.
+ * There is no way to specify a 16x16->32 multiply in portable C, but
+ * some C compilers will do the right thing if you provide the correct
+ * combination of casts.
+ * NB: for 12-bit samples, a full 32-bit multiplication will be needed.
+ */
+
+#ifdef EIGHT_BIT_SAMPLES
+#ifdef SHORTxSHORT_32		/* may work if 'int' is 32 bits */
+#define MULTIPLY(var,const)  (((INT16) (var)) * ((INT16) (const)))
+#endif
+#ifdef SHORTxLCONST_32		/* known to work with Microsoft C 6.0 */
+#define MULTIPLY(var,const)  (((INT16) (var)) * ((INT32) (const)))
+#endif
+#endif
+
+#ifndef MULTIPLY		/* default definition */
+#define MULTIPLY(var,const)  ((var) * (const))
+#endif
+
+
+/* 
+  Unlike our decoder where we approximate the FIXes, we need to use exact
+ones here or successive P-frames will drift too much with Reference frame coding 
+*/
+#define FIX_0_211164243 1730
+#define FIX_0_275899380 2260
+#define FIX_0_298631336 2446
+#define FIX_0_390180644 3196
+#define FIX_0_509795579 4176
+#define FIX_0_541196100 4433
+#define FIX_0_601344887 4926
+#define FIX_0_765366865 6270
+#define FIX_0_785694958 6436
+#define FIX_0_899976223 7373
+#define FIX_1_061594337 8697
+#define FIX_1_111140466 9102
+#define FIX_1_175875602 9633
+#define FIX_1_306562965 10703
+#define FIX_1_387039845 11363
+#define FIX_1_451774981 11893
+#define FIX_1_501321110 12299
+#define FIX_1_662939225 13623
+#define FIX_1_847759065 15137
+#define FIX_1_961570560 16069
+#define FIX_2_053119869 16819
+#define FIX_2_172734803 17799
+#define FIX_2_562915447 20995
+#define FIX_3_072711026 25172
+
+/*
+ * Perform the inverse DCT on one block of coefficients.
+ */
+
+void j_rev_dct(DCTBLOCK data)
+{
+  INT32 tmp0, tmp1, tmp2, tmp3;
+  INT32 tmp10, tmp11, tmp12, tmp13;
+  INT32 z1, z2, z3, z4, z5;
+  INT32 d0, d1, d2, d3, d4, d5, d6, d7;
+  register DCTELEM *dataptr;
+  int rowctr;
+   
+  /* Pass 1: process rows. */
+  /* Note results are scaled up by sqrt(8) compared to a true IDCT; */
+  /* furthermore, we scale the results by 2**PASS1_BITS. */
+
+  dataptr = data;
+
+  for (rowctr = DCTSIZE-1; rowctr >= 0; rowctr--) {
+    /* Due to quantization, we will usually find that many of the input
+     * coefficients are zero, especially the AC terms.  We can exploit this
+     * by short-circuiting the IDCT calculation for any row in which all
+     * the AC terms are zero.  In that case each output is equal to the
+     * DC coefficient (with scale factor as needed).
+     * With typical images and quantization tables, half or more of the
+     * row DCT calculations can be simplified this way.
+     */
+
+    register int *idataptr = (int*)dataptr;
+
+    d0 = dataptr[0];
+    d1 = dataptr[1];
+    d2 = dataptr[2];
+    d3 = dataptr[3];
+    d4 = dataptr[4];
+    d5 = dataptr[5];
+    d6 = dataptr[6];
+    d7 = dataptr[7];
+
+    if ((d1 == 0) && (idataptr[1] | idataptr[2] | idataptr[3]) == 0) {
+      /* AC terms all zero */
+      if (d0) {
+	  /* Compute a 32 bit value to assign. */
+	  DCTELEM dcval = (DCTELEM) (d0 << PASS1_BITS);
+	  register int v = (dcval & 0xffff) | ((dcval << 16) & 0xffff0000);
+	  
+	  idataptr[0] = v;
+	  idataptr[1] = v;
+	  idataptr[2] = v;
+	  idataptr[3] = v;
+      }
+      
+      dataptr += DCTSIZE;	/* advance pointer to next row */
+      continue;
+    }
+
+    /* Even part: reverse the even part of the forward DCT. */
+    /* The rotator is sqrt(2)*c(-6). */
+{
+    if (d6) {
+	if (d4) {
+	    if (d2) {
+		if (d0) {
+		    /* d0 != 0, d2 != 0, d4 != 0, d6 != 0 */
+		    z1 = MULTIPLY(d2 + d6, FIX_0_541196100);
+		    tmp2 = z1 + MULTIPLY(-d6, FIX_1_847759065);
+		    tmp3 = z1 + MULTIPLY(d2, FIX_0_765366865);
+
+		    tmp0 = (d0 + d4) << CONST_BITS;
+		    tmp1 = (d0 - d4) << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp1 + tmp2;
+		    tmp12 = tmp1 - tmp2;
+		} else {
+		    /* d0 == 0, d2 != 0, d4 != 0, d6 != 0 */
+		    z1 = MULTIPLY(d2 + d6, FIX_0_541196100);
+		    tmp2 = z1 + MULTIPLY(-d6, FIX_1_847759065);
+		    tmp3 = z1 + MULTIPLY(d2, FIX_0_765366865);
+
+		    tmp0 = d4 << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp2 - tmp0;
+		    tmp12 = -(tmp0 + tmp2);
+		}
+	    } else {
+		if (d0) {
+		    /* d0 != 0, d2 == 0, d4 != 0, d6 != 0 */
+		    tmp2 = MULTIPLY(-d6, FIX_1_306562965);
+		    tmp3 = MULTIPLY(d6, FIX_0_541196100);
+
+		    tmp0 = (d0 + d4) << CONST_BITS;
+		    tmp1 = (d0 - d4) << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp1 + tmp2;
+		    tmp12 = tmp1 - tmp2;
+		} else {
+		    /* d0 == 0, d2 == 0, d4 != 0, d6 != 0 */
+		    tmp2 = MULTIPLY(-d6, FIX_1_306562965);
+		    tmp3 = MULTIPLY(d6, FIX_0_541196100);
+
+		    tmp0 = d4 << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp2 - tmp0;
+		    tmp12 = -(tmp0 + tmp2);
+		}
+	    }
+	} else {
+	    if (d2) {
+		if (d0) {
+		    /* d0 != 0, d2 != 0, d4 == 0, d6 != 0 */
+		    z1 = MULTIPLY(d2 + d6, FIX_0_541196100);
+		    tmp2 = z1 + MULTIPLY(-d6, FIX_1_847759065);
+		    tmp3 = z1 + MULTIPLY(d2, FIX_0_765366865);
+
+		    tmp0 = d0 << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp0 + tmp2;
+		    tmp12 = tmp0 - tmp2;
+		} else {
+		    /* d0 == 0, d2 != 0, d4 == 0, d6 != 0 */
+		    z1 = MULTIPLY(d2 + d6, FIX_0_541196100);
+		    tmp2 = z1 + MULTIPLY(-d6, FIX_1_847759065);
+		    tmp3 = z1 + MULTIPLY(d2, FIX_0_765366865);
+
+		    tmp10 = tmp3;
+		    tmp13 = -tmp3;
+		    tmp11 = tmp2;
+		    tmp12 = -tmp2;
+		}
+	    } else {
+		if (d0) {
+		    /* d0 != 0, d2 == 0, d4 == 0, d6 != 0 */
+		    tmp2 = MULTIPLY(-d6, FIX_1_306562965);
+		    tmp3 = MULTIPLY(d6, FIX_0_541196100);
+
+		    tmp0 = d0 << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp0 + tmp2;
+		    tmp12 = tmp0 - tmp2;
+		} else {
+		    /* d0 == 0, d2 == 0, d4 == 0, d6 != 0 */
+		    tmp2 = MULTIPLY(-d6, FIX_1_306562965);
+		    tmp3 = MULTIPLY(d6, FIX_0_541196100);
+
+		    tmp10 = tmp3;
+		    tmp13 = -tmp3;
+		    tmp11 = tmp2;
+		    tmp12 = -tmp2;
+		}
+	    }
+	}
+    } else {
+	if (d4) {
+	    if (d2) {
+		if (d0) {
+		    /* d0 != 0, d2 != 0, d4 != 0, d6 == 0 */
+		    tmp2 = MULTIPLY(d2, FIX_0_541196100);
+		    tmp3 = MULTIPLY(d2, FIX_1_306562965);
+
+		    tmp0 = (d0 + d4) << CONST_BITS;
+		    tmp1 = (d0 - d4) << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp1 + tmp2;
+		    tmp12 = tmp1 - tmp2;
+		} else {
+		    /* d0 == 0, d2 != 0, d4 != 0, d6 == 0 */
+		    tmp2 = MULTIPLY(d2, FIX_0_541196100);
+		    tmp3 = MULTIPLY(d2, FIX_1_306562965);
+
+		    tmp0 = d4 << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp2 - tmp0;
+		    tmp12 = -(tmp0 + tmp2);
+		}
+	    } else {
+		if (d0) {
+		    /* d0 != 0, d2 == 0, d4 != 0, d6 == 0 */
+		    tmp10 = tmp13 = (d0 + d4) << CONST_BITS;
+		    tmp11 = tmp12 = (d0 - d4) << CONST_BITS;
+		} else {
+		    /* d0 == 0, d2 == 0, d4 != 0, d6 == 0 */
+		    tmp10 = tmp13 = d4 << CONST_BITS;
+		    tmp11 = tmp12 = -tmp10;
+		}
+	    }
+	} else {
+	    if (d2) {
+		if (d0) {
+		    /* d0 != 0, d2 != 0, d4 == 0, d6 == 0 */
+		    tmp2 = MULTIPLY(d2, FIX_0_541196100);
+		    tmp3 = MULTIPLY(d2, FIX_1_306562965);
+
+		    tmp0 = d0 << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp0 + tmp2;
+		    tmp12 = tmp0 - tmp2;
+		} else {
+		    /* d0 == 0, d2 != 0, d4 == 0, d6 == 0 */
+		    tmp2 = MULTIPLY(d2, FIX_0_541196100);
+		    tmp3 = MULTIPLY(d2, FIX_1_306562965);
+
+		    tmp10 = tmp3;
+		    tmp13 = -tmp3;
+		    tmp11 = tmp2;
+		    tmp12 = -tmp2;
+		}
+	    } else {
+		if (d0) {
+		    /* d0 != 0, d2 == 0, d4 == 0, d6 == 0 */
+		    tmp10 = tmp13 = tmp11 = tmp12 = d0 << CONST_BITS;
+		} else {
+		    /* d0 == 0, d2 == 0, d4 == 0, d6 == 0 */
+		    tmp10 = tmp13 = tmp11 = tmp12 = 0;
+		}
+	    }
+	}
+      }
+
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+
+    if (d7) {
+	if (d5) {
+	    if (d3) {
+		if (d1) {
+		    /* d1 != 0, d3 != 0, d5 != 0, d7 != 0 */
+		    z1 = d7 + d1;
+		    z2 = d5 + d3;
+		    z3 = d7 + d3;
+		    z4 = d5 + d1;
+		    z5 = MULTIPLY(z3 + z4, FIX_1_175875602);
+		    
+		    tmp0 = MULTIPLY(d7, FIX_0_298631336); 
+		    tmp1 = MULTIPLY(d5, FIX_2_053119869);
+		    tmp2 = MULTIPLY(d3, FIX_3_072711026);
+		    tmp3 = MULTIPLY(d1, FIX_1_501321110);
+		    z1 = MULTIPLY(-z1, FIX_0_899976223);
+		    z2 = MULTIPLY(-z2, FIX_2_562915447);
+		    z3 = MULTIPLY(-z3, FIX_1_961570560);
+		    z4 = MULTIPLY(-z4, FIX_0_390180644);
+		    
+		    z3 += z5;
+		    z4 += z5;
+		    
+		    tmp0 += z1 + z3;
+		    tmp1 += z2 + z4;
+		    tmp2 += z2 + z3;
+		    tmp3 += z1 + z4;
+		} else {
+		    /* d1 == 0, d3 != 0, d5 != 0, d7 != 0 */
+		    z2 = d5 + d3;
+		    z3 = d7 + d3;
+		    z5 = MULTIPLY(z3 + d5, FIX_1_175875602);
+		    
+		    tmp0 = MULTIPLY(d7, FIX_0_298631336); 
+		    tmp1 = MULTIPLY(d5, FIX_2_053119869);
+		    tmp2 = MULTIPLY(d3, FIX_3_072711026);
+		    z1 = MULTIPLY(-d7, FIX_0_899976223);
+		    z2 = MULTIPLY(-z2, FIX_2_562915447);
+		    z3 = MULTIPLY(-z3, FIX_1_961570560);
+		    z4 = MULTIPLY(-d5, FIX_0_390180644);
+		    
+		    z3 += z5;
+		    z4 += z5;
+		    
+		    tmp0 += z1 + z3;
+		    tmp1 += z2 + z4;
+		    tmp2 += z2 + z3;
+		    tmp3 = z1 + z4;
+		}
+	    } else {
+		if (d1) {
+		    /* d1 != 0, d3 == 0, d5 != 0, d7 != 0 */
+		    z1 = d7 + d1;
+		    z4 = d5 + d1;
+		    z5 = MULTIPLY(d7 + z4, FIX_1_175875602);
+		    
+		    tmp0 = MULTIPLY(d7, FIX_0_298631336); 
+		    tmp1 = MULTIPLY(d5, FIX_2_053119869);
+		    tmp3 = MULTIPLY(d1, FIX_1_501321110);
+		    z1 = MULTIPLY(-z1, FIX_0_899976223);
+		    z2 = MULTIPLY(-d5, FIX_2_562915447);
+		    z3 = MULTIPLY(-d7, FIX_1_961570560);
+		    z4 = MULTIPLY(-z4, FIX_0_390180644);
+		    
+		    z3 += z5;
+		    z4 += z5;
+		    
+		    tmp0 += z1 + z3;
+		    tmp1 += z2 + z4;
+		    tmp2 = z2 + z3;
+		    tmp3 += z1 + z4;
+		} else {
+		    /* d1 == 0, d3 == 0, d5 != 0, d7 != 0 */
+		    tmp0 = MULTIPLY(-d7, FIX_0_601344887); 
+		    z1 = MULTIPLY(-d7, FIX_0_899976223);
+		    z3 = MULTIPLY(-d7, FIX_1_961570560);
+		    tmp1 = MULTIPLY(-d5, FIX_0_509795579);
+		    z2 = MULTIPLY(-d5, FIX_2_562915447);
+		    z4 = MULTIPLY(-d5, FIX_0_390180644);
+		    z5 = MULTIPLY(d5 + d7, FIX_1_175875602);
+		    
+		    z3 += z5;
+		    z4 += z5;
+		    
+		    tmp0 += z3;
+		    tmp1 += z4;
+		    tmp2 = z2 + z3;
+		    tmp3 = z1 + z4;
+		}
+	    }
+	} else {
+	    if (d3) {
+		if (d1) {
+		    /* d1 != 0, d3 != 0, d5 == 0, d7 != 0 */
+		    z1 = d7 + d1;
+		    z3 = d7 + d3;
+		    z5 = MULTIPLY(z3 + d1, FIX_1_175875602);
+		    
+		    tmp0 = MULTIPLY(d7, FIX_0_298631336); 
+		    tmp2 = MULTIPLY(d3, FIX_3_072711026);
+		    tmp3 = MULTIPLY(d1, FIX_1_501321110);
+		    z1 = MULTIPLY(-z1, FIX_0_899976223);
+		    z2 = MULTIPLY(-d3, FIX_2_562915447);
+		    z3 = MULTIPLY(-z3, FIX_1_961570560);
+		    z4 = MULTIPLY(-d1, FIX_0_390180644);
+		    
+		    z3 += z5;
+		    z4 += z5;
+		    
+		    tmp0 += z1 + z3;
+		    tmp1 = z2 + z4;
+		    tmp2 += z2 + z3;
+		    tmp3 += z1 + z4;
+		} else {
+		    /* d1 == 0, d3 != 0, d5 == 0, d7 != 0 */
+		    z3 = d7 + d3;
+		    
+		    tmp0 = MULTIPLY(-d7, FIX_0_601344887); 
+		    z1 = MULTIPLY(-d7, FIX_0_899976223);
+		    tmp2 = MULTIPLY(d3, FIX_0_509795579);
+		    z2 = MULTIPLY(-d3, FIX_2_562915447);
+		    z5 = MULTIPLY(z3, FIX_1_175875602);
+		    z3 = MULTIPLY(-z3, FIX_0_785694958);
+		    
+		    tmp0 += z3;
+		    tmp1 = z2 + z5;
+		    tmp2 += z3;
+		    tmp3 = z1 + z5;
+		}
+	    } else {
+		if (d1) {
+		    /* d1 != 0, d3 == 0, d5 == 0, d7 != 0 */
+		    z1 = d7 + d1;
+		    z5 = MULTIPLY(z1, FIX_1_175875602);
+
+		    z1 = MULTIPLY(z1, FIX_0_275899380);
+		    z3 = MULTIPLY(-d7, FIX_1_961570560);
+		    tmp0 = MULTIPLY(-d7, FIX_1_662939225); 
+		    z4 = MULTIPLY(-d1, FIX_0_390180644);
+		    tmp3 = MULTIPLY(d1, FIX_1_111140466);
+
+		    tmp0 += z1;
+		    tmp1 = z4 + z5;
+		    tmp2 = z3 + z5;
+		    tmp3 += z1;
+		} else {
+		    /* d1 == 0, d3 == 0, d5 == 0, d7 != 0 */
+		    tmp0 = MULTIPLY(-d7, FIX_1_387039845);
+		    tmp1 = MULTIPLY(d7, FIX_1_175875602);
+		    tmp2 = MULTIPLY(-d7, FIX_0_785694958);
+		    tmp3 = MULTIPLY(d7, FIX_0_275899380);
+		}
+	    }
+	}
+    } else {
+	if (d5) {
+	    if (d3) {
+		if (d1) {
+		    /* d1 != 0, d3 != 0, d5 != 0, d7 == 0 */
+		    z2 = d5 + d3;
+		    z4 = d5 + d1;
+		    z5 = MULTIPLY(d3 + z4, FIX_1_175875602);
+		    
+		    tmp1 = MULTIPLY(d5, FIX_2_053119869);
+		    tmp2 = MULTIPLY(d3, FIX_3_072711026);
+		    tmp3 = MULTIPLY(d1, FIX_1_501321110);
+		    z1 = MULTIPLY(-d1, FIX_0_899976223);
+		    z2 = MULTIPLY(-z2, FIX_2_562915447);
+		    z3 = MULTIPLY(-d3, FIX_1_961570560);
+		    z4 = MULTIPLY(-z4, FIX_0_390180644);
+		    
+		    z3 += z5;
+		    z4 += z5;
+		    
+		    tmp0 = z1 + z3;
+		    tmp1 += z2 + z4;
+		    tmp2 += z2 + z3;
+		    tmp3 += z1 + z4;
+		} else {
+		    /* d1 == 0, d3 != 0, d5 != 0, d7 == 0 */
+		    z2 = d5 + d3;
+		    
+		    z5 = MULTIPLY(z2, FIX_1_175875602);
+		    tmp1 = MULTIPLY(d5, FIX_1_662939225);
+		    z4 = MULTIPLY(-d5, FIX_0_390180644);
+		    z2 = MULTIPLY(-z2, FIX_1_387039845);
+		    tmp2 = MULTIPLY(d3, FIX_1_111140466);
+		    z3 = MULTIPLY(-d3, FIX_1_961570560);
+		    
+		    tmp0 = z3 + z5;
+		    tmp1 += z2;
+		    tmp2 += z2;
+		    tmp3 = z4 + z5;
+		}
+	    } else {
+		if (d1) {
+		    /* d1 != 0, d3 == 0, d5 != 0, d7 == 0 */
+		    z4 = d5 + d1;
+		    
+		    z5 = MULTIPLY(z4, FIX_1_175875602);
+		    z1 = MULTIPLY(-d1, FIX_0_899976223);
+		    tmp3 = MULTIPLY(d1, FIX_0_601344887);
+		    tmp1 = MULTIPLY(-d5, FIX_0_509795579);
+		    z2 = MULTIPLY(-d5, FIX_2_562915447);
+		    z4 = MULTIPLY(z4, FIX_0_785694958);
+		    
+		    tmp0 = z1 + z5;
+		    tmp1 += z4;
+		    tmp2 = z2 + z5;
+		    tmp3 += z4;
+		} else {
+		    /* d1 == 0, d3 == 0, d5 != 0, d7 == 0 */
+		    tmp0 = MULTIPLY(d5, FIX_1_175875602);
+		    tmp1 = MULTIPLY(d5, FIX_0_275899380);
+		    tmp2 = MULTIPLY(-d5, FIX_1_387039845);
+		    tmp3 = MULTIPLY(d5, FIX_0_785694958);
+		}
+	    }
+	} else {
+	    if (d3) {
+		if (d1) {
+		    /* d1 != 0, d3 != 0, d5 == 0, d7 == 0 */
+		    z5 = d1 + d3;
+		    tmp3 = MULTIPLY(d1, FIX_0_211164243);
+		    tmp2 = MULTIPLY(-d3, FIX_1_451774981);
+		    z1 = MULTIPLY(d1, FIX_1_061594337);
+		    z2 = MULTIPLY(-d3, FIX_2_172734803);
+		    z4 = MULTIPLY(z5, FIX_0_785694958);
+		    z5 = MULTIPLY(z5, FIX_1_175875602);
+		    
+		    tmp0 = z1 - z4;
+		    tmp1 = z2 + z4;
+		    tmp2 += z5;
+		    tmp3 += z5;
+		} else {
+		    /* d1 == 0, d3 != 0, d5 == 0, d7 == 0 */
+		    tmp0 = MULTIPLY(-d3, FIX_0_785694958);
+		    tmp1 = MULTIPLY(-d3, FIX_1_387039845);
+		    tmp2 = MULTIPLY(-d3, FIX_0_275899380);
+		    tmp3 = MULTIPLY(d3, FIX_1_175875602);
+		}
+	    } else {
+		if (d1) {
+		    /* d1 != 0, d3 == 0, d5 == 0, d7 == 0 */
+		    tmp0 = MULTIPLY(d1, FIX_0_275899380);
+		    tmp1 = MULTIPLY(d1, FIX_0_785694958);
+		    tmp2 = MULTIPLY(d1, FIX_1_175875602);
+		    tmp3 = MULTIPLY(d1, FIX_1_387039845);
+		} else {
+		    /* d1 == 0, d3 == 0, d5 == 0, d7 == 0 */
+		    tmp0 = tmp1 = tmp2 = tmp3 = 0;
+		}
+	    }
+	}
+    }
+}
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+    dataptr[0] = (DCTELEM) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS);
+    dataptr[7] = (DCTELEM) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS);
+    dataptr[1] = (DCTELEM) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS);
+    dataptr[6] = (DCTELEM) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS);
+    dataptr[2] = (DCTELEM) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS);
+    dataptr[5] = (DCTELEM) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS);
+    dataptr[3] = (DCTELEM) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS);
+    dataptr[4] = (DCTELEM) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS);
+
+    dataptr += DCTSIZE;		/* advance pointer to next row */
+  }
+
+  /* Pass 2: process columns. */
+  /* Note that we must descale the results by a factor of 8 == 2**3, */
+  /* and also undo the PASS1_BITS scaling. */
+
+  dataptr = data;
+  for (rowctr = DCTSIZE-1; rowctr >= 0; rowctr--) {
+    /* Columns of zeroes can be exploited in the same way as we did with rows.
+     * However, the row calculation has created many nonzero AC terms, so the
+     * simplification applies less often (typically 5% to 10% of the time).
+     * On machines with very fast multiplication, it's possible that the
+     * test takes more time than it's worth.  In that case this section
+     * may be commented out.
+     */
+
+    d0 = dataptr[DCTSIZE*0];
+    d1 = dataptr[DCTSIZE*1];
+    d2 = dataptr[DCTSIZE*2];
+    d3 = dataptr[DCTSIZE*3];
+    d4 = dataptr[DCTSIZE*4];
+    d5 = dataptr[DCTSIZE*5];
+    d6 = dataptr[DCTSIZE*6];
+    d7 = dataptr[DCTSIZE*7];
+
+    /* Even part: reverse the even part of the forward DCT. */
+    /* The rotator is sqrt(2)*c(-6). */
+    if (d6) {
+	if (d4) {
+	    if (d2) {
+		if (d0) {
+		    /* d0 != 0, d2 != 0, d4 != 0, d6 != 0 */
+		    z1 = MULTIPLY(d2 + d6, FIX_0_541196100);
+		    tmp2 = z1 + MULTIPLY(-d6, FIX_1_847759065);
+		    tmp3 = z1 + MULTIPLY(d2, FIX_0_765366865);
+
+		    tmp0 = (d0 + d4) << CONST_BITS;
+		    tmp1 = (d0 - d4) << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp1 + tmp2;
+		    tmp12 = tmp1 - tmp2;
+		} else {
+		    /* d0 == 0, d2 != 0, d4 != 0, d6 != 0 */
+		    z1 = MULTIPLY(d2 + d6, FIX_0_541196100);
+		    tmp2 = z1 + MULTIPLY(-d6, FIX_1_847759065);
+		    tmp3 = z1 + MULTIPLY(d2, FIX_0_765366865);
+
+		    tmp0 = d4 << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp2 - tmp0;
+		    tmp12 = -(tmp0 + tmp2);
+		}
+	    } else {
+		if (d0) {
+		    /* d0 != 0, d2 == 0, d4 != 0, d6 != 0 */
+		    tmp2 = MULTIPLY(-d6, FIX_1_306562965);
+		    tmp3 = MULTIPLY(d6, FIX_0_541196100);
+
+		    tmp0 = (d0 + d4) << CONST_BITS;
+		    tmp1 = (d0 - d4) << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp1 + tmp2;
+		    tmp12 = tmp1 - tmp2;
+		} else {
+		    /* d0 == 0, d2 == 0, d4 != 0, d6 != 0 */
+		    tmp2 = MULTIPLY(-d6, FIX_1_306562965);
+		    tmp3 = MULTIPLY(d6, FIX_0_541196100);
+
+		    tmp0 = d4 << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp2 - tmp0;
+		    tmp12 = -(tmp0 + tmp2);
+		}
+	    }
+	} else {
+	    if (d2) {
+		if (d0) {
+		    /* d0 != 0, d2 != 0, d4 == 0, d6 != 0 */
+		    z1 = MULTIPLY(d2 + d6, FIX_0_541196100);
+		    tmp2 = z1 + MULTIPLY(-d6, FIX_1_847759065);
+		    tmp3 = z1 + MULTIPLY(d2, FIX_0_765366865);
+
+		    tmp0 = d0 << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp0 + tmp2;
+		    tmp12 = tmp0 - tmp2;
+		} else {
+		    /* d0 == 0, d2 != 0, d4 == 0, d6 != 0 */
+		    z1 = MULTIPLY(d2 + d6, FIX_0_541196100);
+		    tmp2 = z1 + MULTIPLY(-d6, FIX_1_847759065);
+		    tmp3 = z1 + MULTIPLY(d2, FIX_0_765366865);
+
+		    tmp10 = tmp3;
+		    tmp13 = -tmp3;
+		    tmp11 = tmp2;
+		    tmp12 = -tmp2;
+		}
+	    } else {
+		if (d0) {
+		    /* d0 != 0, d2 == 0, d4 == 0, d6 != 0 */
+		    tmp2 = MULTIPLY(-d6, FIX_1_306562965);
+		    tmp3 = MULTIPLY(d6, FIX_0_541196100);
+
+		    tmp0 = d0 << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp0 + tmp2;
+		    tmp12 = tmp0 - tmp2;
+		} else {
+		    /* d0 == 0, d2 == 0, d4 == 0, d6 != 0 */
+		    tmp2 = MULTIPLY(-d6, FIX_1_306562965);
+		    tmp3 = MULTIPLY(d6, FIX_0_541196100);
+
+		    tmp10 = tmp3;
+		    tmp13 = -tmp3;
+		    tmp11 = tmp2;
+		    tmp12 = -tmp2;
+		}
+	    }
+	}
+    } else {
+	if (d4) {
+	    if (d2) {
+		if (d0) {
+		    /* d0 != 0, d2 != 0, d4 != 0, d6 == 0 */
+		    tmp2 = MULTIPLY(d2, FIX_0_541196100);
+		    tmp3 = MULTIPLY(d2, FIX_1_306562965);
+
+		    tmp0 = (d0 + d4) << CONST_BITS;
+		    tmp1 = (d0 - d4) << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp1 + tmp2;
+		    tmp12 = tmp1 - tmp2;
+		} else {
+		    /* d0 == 0, d2 != 0, d4 != 0, d6 == 0 */
+		    tmp2 = MULTIPLY(d2, FIX_0_541196100);
+		    tmp3 = MULTIPLY(d2, FIX_1_306562965);
+
+		    tmp0 = d4 << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp2 - tmp0;
+		    tmp12 = -(tmp0 + tmp2);
+		}
+	    } else {
+		if (d0) {
+		    /* d0 != 0, d2 == 0, d4 != 0, d6 == 0 */
+		    tmp10 = tmp13 = (d0 + d4) << CONST_BITS;
+		    tmp11 = tmp12 = (d0 - d4) << CONST_BITS;
+		} else {
+		    /* d0 == 0, d2 == 0, d4 != 0, d6 == 0 */
+		    tmp10 = tmp13 = d4 << CONST_BITS;
+		    tmp11 = tmp12 = -tmp10;
+		}
+	    }
+	} else {
+	    if (d2) {
+		if (d0) {
+		    /* d0 != 0, d2 != 0, d4 == 0, d6 == 0 */
+		    tmp2 = MULTIPLY(d2, FIX_0_541196100);
+		    tmp3 = MULTIPLY(d2, FIX_1_306562965);
+
+		    tmp0 = d0 << CONST_BITS;
+
+		    tmp10 = tmp0 + tmp3;
+		    tmp13 = tmp0 - tmp3;
+		    tmp11 = tmp0 + tmp2;
+		    tmp12 = tmp0 - tmp2;
+		} else {
+		    /* d0 == 0, d2 != 0, d4 == 0, d6 == 0 */
+		    tmp2 = MULTIPLY(d2, FIX_0_541196100);
+		    tmp3 = MULTIPLY(d2, FIX_1_306562965);
+
+		    tmp10 = tmp3;
+		    tmp13 = -tmp3;
+		    tmp11 = tmp2;
+		    tmp12 = -tmp2;
+		}
+	    } else {
+		if (d0) {
+		    /* d0 != 0, d2 == 0, d4 == 0, d6 == 0 */
+		    tmp10 = tmp13 = tmp11 = tmp12 = d0 << CONST_BITS;
+		} else {
+		    /* d0 == 0, d2 == 0, d4 == 0, d6 == 0 */
+		    tmp10 = tmp13 = tmp11 = tmp12 = 0;
+		}
+	    }
+	}
+    }
+
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+    if (d7) {
+	if (d5) {
+	    if (d3) {
+		if (d1) {
+		    /* d1 != 0, d3 != 0, d5 != 0, d7 != 0 */
+		    z1 = d7 + d1;
+		    z2 = d5 + d3;
+		    z3 = d7 + d3;
+		    z4 = d5 + d1;
+		    z5 = MULTIPLY(z3 + z4, FIX_1_175875602);
+		    
+		    tmp0 = MULTIPLY(d7, FIX_0_298631336); 
+		    tmp1 = MULTIPLY(d5, FIX_2_053119869);
+		    tmp2 = MULTIPLY(d3, FIX_3_072711026);
+		    tmp3 = MULTIPLY(d1, FIX_1_501321110);
+		    z1 = MULTIPLY(-z1, FIX_0_899976223);
+		    z2 = MULTIPLY(-z2, FIX_2_562915447);
+		    z3 = MULTIPLY(-z3, FIX_1_961570560);
+		    z4 = MULTIPLY(-z4, FIX_0_390180644);
+		    
+		    z3 += z5;
+		    z4 += z5;
+		    
+		    tmp0 += z1 + z3;
+		    tmp1 += z2 + z4;
+		    tmp2 += z2 + z3;
+		    tmp3 += z1 + z4;
+		} else {
+		    /* d1 == 0, d3 != 0, d5 != 0, d7 != 0 */
+		    z1 = d7;
+		    z2 = d5 + d3;
+		    z3 = d7 + d3;
+		    z5 = MULTIPLY(z3 + d5, FIX_1_175875602);
+		    
+		    tmp0 = MULTIPLY(d7, FIX_0_298631336); 
+		    tmp1 = MULTIPLY(d5, FIX_2_053119869);
+		    tmp2 = MULTIPLY(d3, FIX_3_072711026);
+		    z1 = MULTIPLY(-d7, FIX_0_899976223);
+		    z2 = MULTIPLY(-z2, FIX_2_562915447);
+		    z3 = MULTIPLY(-z3, FIX_1_961570560);
+		    z4 = MULTIPLY(-d5, FIX_0_390180644);
+		    
+		    z3 += z5;
+		    z4 += z5;
+		    
+		    tmp0 += z1 + z3;
+		    tmp1 += z2 + z4;
+		    tmp2 += z2 + z3;
+		    tmp3 = z1 + z4;
+		}
+	    } else {
+		if (d1) {
+		    /* d1 != 0, d3 == 0, d5 != 0, d7 != 0 */
+		    z1 = d7 + d1;
+		    z2 = d5;
+		    z3 = d7;
+		    z4 = d5 + d1;
+		    z5 = MULTIPLY(z3 + z4, FIX_1_175875602);
+		    
+		    tmp0 = MULTIPLY(d7, FIX_0_298631336); 
+		    tmp1 = MULTIPLY(d5, FIX_2_053119869);
+		    tmp3 = MULTIPLY(d1, FIX_1_501321110);
+		    z1 = MULTIPLY(-z1, FIX_0_899976223);
+		    z2 = MULTIPLY(-d5, FIX_2_562915447);
+		    z3 = MULTIPLY(-d7, FIX_1_961570560);
+		    z4 = MULTIPLY(-z4, FIX_0_390180644);
+		    
+		    z3 += z5;
+		    z4 += z5;
+		    
+		    tmp0 += z1 + z3;
+		    tmp1 += z2 + z4;
+		    tmp2 = z2 + z3;
+		    tmp3 += z1 + z4;
+		} else {
+		    /* d1 == 0, d3 == 0, d5 != 0, d7 != 0 */
+		    tmp0 = MULTIPLY(-d7, FIX_0_601344887); 
+		    z1 = MULTIPLY(-d7, FIX_0_899976223);
+		    z3 = MULTIPLY(-d7, FIX_1_961570560);
+		    tmp1 = MULTIPLY(-d5, FIX_0_509795579);
+		    z2 = MULTIPLY(-d5, FIX_2_562915447);
+		    z4 = MULTIPLY(-d5, FIX_0_390180644);
+		    z5 = MULTIPLY(d5 + d7, FIX_1_175875602);
+		    
+		    z3 += z5;
+		    z4 += z5;
+		    
+		    tmp0 += z3;
+		    tmp1 += z4;
+		    tmp2 = z2 + z3;
+		    tmp3 = z1 + z4;
+		}
+	    }
+	} else {
+	    if (d3) {
+		if (d1) {
+		    /* d1 != 0, d3 != 0, d5 == 0, d7 != 0 */
+		    z1 = d7 + d1;
+		    z3 = d7 + d3;
+		    z5 = MULTIPLY(z3 + d1, FIX_1_175875602);
+		    
+		    tmp0 = MULTIPLY(d7, FIX_0_298631336); 
+		    tmp2 = MULTIPLY(d3, FIX_3_072711026);
+		    tmp3 = MULTIPLY(d1, FIX_1_501321110);
+		    z1 = MULTIPLY(-z1, FIX_0_899976223);
+		    z2 = MULTIPLY(-d3, FIX_2_562915447);
+		    z3 = MULTIPLY(-z3, FIX_1_961570560);
+		    z4 = MULTIPLY(-d1, FIX_0_390180644);
+		    
+		    z3 += z5;
+		    z4 += z5;
+		    
+		    tmp0 += z1 + z3;
+		    tmp1 = z2 + z4;
+		    tmp2 += z2 + z3;
+		    tmp3 += z1 + z4;
+		} else {
+		    /* d1 == 0, d3 != 0, d5 == 0, d7 != 0 */
+		    z3 = d7 + d3;
+		    
+		    tmp0 = MULTIPLY(-d7, FIX_0_601344887); 
+		    z1 = MULTIPLY(-d7, FIX_0_899976223);
+		    tmp2 = MULTIPLY(d3, FIX_0_509795579);
+		    z2 = MULTIPLY(-d3, FIX_2_562915447);
+		    z5 = MULTIPLY(z3, FIX_1_175875602);
+		    z3 = MULTIPLY(-z3, FIX_0_785694958);
+		    
+		    tmp0 += z3;
+		    tmp1 = z2 + z5;
+		    tmp2 += z3;
+		    tmp3 = z1 + z5;
+		}
+	    } else {
+		if (d1) {
+		    /* d1 != 0, d3 == 0, d5 == 0, d7 != 0 */
+		    z1 = d7 + d1;
+		    z5 = MULTIPLY(z1, FIX_1_175875602);
+
+		    z1 = MULTIPLY(z1, FIX_0_275899380);
+		    z3 = MULTIPLY(-d7, FIX_1_961570560);
+		    tmp0 = MULTIPLY(-d7, FIX_1_662939225); 
+		    z4 = MULTIPLY(-d1, FIX_0_390180644);
+		    tmp3 = MULTIPLY(d1, FIX_1_111140466);
+
+		    tmp0 += z1;
+		    tmp1 = z4 + z5;
+		    tmp2 = z3 + z5;
+		    tmp3 += z1;
+		} else {
+		    /* d1 == 0, d3 == 0, d5 == 0, d7 != 0 */
+		    tmp0 = MULTIPLY(-d7, FIX_1_387039845);
+		    tmp1 = MULTIPLY(d7, FIX_1_175875602);
+		    tmp2 = MULTIPLY(-d7, FIX_0_785694958);
+		    tmp3 = MULTIPLY(d7, FIX_0_275899380);
+		}
+	    }
+	}
+    } else {
+	if (d5) {
+	    if (d3) {
+		if (d1) {
+		    /* d1 != 0, d3 != 0, d5 != 0, d7 == 0 */
+		    z2 = d5 + d3;
+		    z4 = d5 + d1;
+		    z5 = MULTIPLY(d3 + z4, FIX_1_175875602);
+		    
+		    tmp1 = MULTIPLY(d5, FIX_2_053119869);
+		    tmp2 = MULTIPLY(d3, FIX_3_072711026);
+		    tmp3 = MULTIPLY(d1, FIX_1_501321110);
+		    z1 = MULTIPLY(-d1, FIX_0_899976223);
+		    z2 = MULTIPLY(-z2, FIX_2_562915447);
+		    z3 = MULTIPLY(-d3, FIX_1_961570560);
+		    z4 = MULTIPLY(-z4, FIX_0_390180644);
+		    
+		    z3 += z5;
+		    z4 += z5;
+		    
+		    tmp0 = z1 + z3;
+		    tmp1 += z2 + z4;
+		    tmp2 += z2 + z3;
+		    tmp3 += z1 + z4;
+		} else {
+		    /* d1 == 0, d3 != 0, d5 != 0, d7 == 0 */
+		    z2 = d5 + d3;
+		    
+		    z5 = MULTIPLY(z2, FIX_1_175875602);
+		    tmp1 = MULTIPLY(d5, FIX_1_662939225);
+		    z4 = MULTIPLY(-d5, FIX_0_390180644);
+		    z2 = MULTIPLY(-z2, FIX_1_387039845);
+		    tmp2 = MULTIPLY(d3, FIX_1_111140466);
+		    z3 = MULTIPLY(-d3, FIX_1_961570560);
+		    
+		    tmp0 = z3 + z5;
+		    tmp1 += z2;
+		    tmp2 += z2;
+		    tmp3 = z4 + z5;
+		}
+	    } else {
+		if (d1) {
+		    /* d1 != 0, d3 == 0, d5 != 0, d7 == 0 */
+		    z4 = d5 + d1;
+		    
+		    z5 = MULTIPLY(z4, FIX_1_175875602);
+		    z1 = MULTIPLY(-d1, FIX_0_899976223);
+		    tmp3 = MULTIPLY(d1, FIX_0_601344887);
+		    tmp1 = MULTIPLY(-d5, FIX_0_509795579);
+		    z2 = MULTIPLY(-d5, FIX_2_562915447);
+		    z4 = MULTIPLY(z4, FIX_0_785694958);
+		    
+		    tmp0 = z1 + z5;
+		    tmp1 += z4;
+		    tmp2 = z2 + z5;
+		    tmp3 += z4;
+		} else {
+		    /* d1 == 0, d3 == 0, d5 != 0, d7 == 0 */
+		    tmp0 = MULTIPLY(d5, FIX_1_175875602);
+		    tmp1 = MULTIPLY(d5, FIX_0_275899380);
+		    tmp2 = MULTIPLY(-d5, FIX_1_387039845);
+		    tmp3 = MULTIPLY(d5, FIX_0_785694958);
+		}
+	    }
+	} else {
+	    if (d3) {
+		if (d1) {
+		    /* d1 != 0, d3 != 0, d5 == 0, d7 == 0 */
+		    z5 = d1 + d3;
+		    tmp3 = MULTIPLY(d1, FIX_0_211164243);
+		    tmp2 = MULTIPLY(-d3, FIX_1_451774981);
+		    z1 = MULTIPLY(d1, FIX_1_061594337);
+		    z2 = MULTIPLY(-d3, FIX_2_172734803);
+		    z4 = MULTIPLY(z5, FIX_0_785694958);
+		    z5 = MULTIPLY(z5, FIX_1_175875602);
+		    
+		    tmp0 = z1 - z4;
+		    tmp1 = z2 + z4;
+		    tmp2 += z5;
+		    tmp3 += z5;
+		} else {
+		    /* d1 == 0, d3 != 0, d5 == 0, d7 == 0 */
+		    tmp0 = MULTIPLY(-d3, FIX_0_785694958);
+		    tmp1 = MULTIPLY(-d3, FIX_1_387039845);
+		    tmp2 = MULTIPLY(-d3, FIX_0_275899380);
+		    tmp3 = MULTIPLY(d3, FIX_1_175875602);
+		}
+	    } else {
+		if (d1) {
+		    /* d1 != 0, d3 == 0, d5 == 0, d7 == 0 */
+		    tmp0 = MULTIPLY(d1, FIX_0_275899380);
+		    tmp1 = MULTIPLY(d1, FIX_0_785694958);
+		    tmp2 = MULTIPLY(d1, FIX_1_175875602);
+		    tmp3 = MULTIPLY(d1, FIX_1_387039845);
+		} else {
+		    /* d1 == 0, d3 == 0, d5 == 0, d7 == 0 */
+		    tmp0 = tmp1 = tmp2 = tmp3 = 0;
+		}
+	    }
+	}
+    }
+
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+    dataptr[DCTSIZE*0] = (DCTELEM) DESCALE(tmp10 + tmp3,
+					   CONST_BITS+PASS1_BITS+3);
+    dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp10 - tmp3,
+					   CONST_BITS+PASS1_BITS+3);
+    dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp11 + tmp2,
+					   CONST_BITS+PASS1_BITS+3);
+    dataptr[DCTSIZE*6] = (DCTELEM) DESCALE(tmp11 - tmp2,
+					   CONST_BITS+PASS1_BITS+3);
+    dataptr[DCTSIZE*2] = (DCTELEM) DESCALE(tmp12 + tmp1,
+					   CONST_BITS+PASS1_BITS+3);
+    dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp12 - tmp1,
+					   CONST_BITS+PASS1_BITS+3);
+    dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp13 + tmp0,
+					   CONST_BITS+PASS1_BITS+3);
+    dataptr[DCTSIZE*4] = (DCTELEM) DESCALE(tmp13 - tmp0,
+					   CONST_BITS+PASS1_BITS+3);
+    
+    dataptr++;			/* advance pointer to next column */
+  }
+}
+
+
diff --git a/libavcodec/libac3/ac3.h b/libavcodec/libac3/ac3.h
new file mode 100644
index 0000000000..117071e7a1
--- /dev/null
+++ b/libavcodec/libac3/ac3.h
@@ -0,0 +1,108 @@
+/* 
+ *    ac3.h
+ *
+ *	Copyright (C) Aaron Holtzman - May 1999
+ *
+ *  This file is part of ac3dec, a free Dolby AC-3 stream decoder.
+ *	
+ *  ac3dec is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2, or (at your option)
+ *  any later version.
+ *   
+ *  ac3dec is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *   
+ *  You should have received a copy of the GNU General Public License
+ *  along with GNU Make; see the file COPYING.  If not, write to
+ *  the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. 
+ *
+ *
+ */
+
+typedef struct ac3_ba_s {
+    uint16_t fsnroffst;	// fine SNR offset
+    uint16_t fgaincod;	// fast gain
+    uint16_t deltbae;	// delta bit allocation exists
+    int8_t deltba[50];	// per-band delta bit allocation
+} ac3_ba_t;
+
+typedef struct ac3_state_s {
+    uint8_t fscod;		// sample rate
+    uint8_t halfrate;		// halfrate factor
+    uint8_t acmod;		// coded channels
+    float clev;			// centre channel mix level
+    float slev;			// surround channels mix level
+    uint8_t lfeon;		// coded lfe channel
+
+    int output;			// type of output
+    float level;		// output level
+    float bias;			// output bias
+
+    uint16_t cplinu;		// coupling in use
+    uint16_t chincpl[5];	// channel coupled
+    uint16_t phsflginu;		// phase flags in use (stereo only)
+    uint16_t cplbndstrc[18];	// coupling band structure
+    uint16_t cplstrtmant;	// coupling channel start mantissa
+    uint16_t cplendmant;	// coupling channel end mantissa
+    float cplco[5][18];		// coupling coordinates
+
+    // derived information
+    uint16_t cplstrtbnd;	// coupling start band (for bit allocation)
+    uint16_t ncplbnd;		// number of coupling bands
+
+    uint16_t rematflg[4];	// stereo rematrixing
+
+    uint16_t endmant[5];	// channel end mantissa
+
+    uint8_t cpl_exp[256];	// decoded coupling channel exponents
+    uint8_t fbw_exp[5][256];	// decoded channel exponents
+    uint8_t lfe_exp[7];		// decoded lfe channel exponents
+
+    uint16_t sdcycod;		// slow decay
+    uint16_t fdcycod;		// fast decay
+    uint16_t sgaincod;		// slow gain
+    uint16_t dbpbcod;		// dB per bit - encodes the dbknee value
+    uint16_t floorcod;		// masking floor
+
+    uint16_t csnroffst;		// coarse SNR offset
+    ac3_ba_t cplba;		// coupling bit allocation parameters
+    ac3_ba_t ba[5];		// channel bit allocation parameters
+    ac3_ba_t lfeba;		// lfe bit allocation parameters
+
+    uint16_t cplfleak;		// coupling fast leak init
+    uint16_t cplsleak;		// coupling slow leak init
+
+    // derived bit allocation information
+    int8_t fbw_bap[5][256];
+    int8_t cpl_bap[256];
+    int8_t lfe_bap[7];
+} ac3_state_t;
+
+/* samples work structure */
+typedef float stream_samples_t[6][256];
+
+#define AC3_CHANNEL 0
+#define AC3_MONO 1
+#define AC3_STEREO 2
+#define AC3_3F 3
+#define AC3_2F1R 4
+#define AC3_3F1R 5
+#define AC3_2F2R 6
+#define AC3_3F2R 7
+#define AC3_CHANNEL1 8
+#define AC3_CHANNEL2 9
+#define AC3_DOLBY 10
+#define AC3_CHANNEL_MASK 15
+
+#define AC3_LFE 16
+#define AC3_ADJUST_LEVEL 32
+
+void ac3_init (void);
+int ac3_syncinfo (uint8_t * buf, int * flags,
+		  int * sample_rate, int * bit_rate);
+int ac3_frame (ac3_state_t * state, uint8_t * buf, int * flags,
+	       float * level, float bias);
+int ac3_block (ac3_state_t * state);
diff --git a/libavcodec/libac3/ac3_internal.h b/libavcodec/libac3/ac3_internal.h
new file mode 100644
index 0000000000..1dce513f4c
--- /dev/null
+++ b/libavcodec/libac3/ac3_internal.h
@@ -0,0 +1,51 @@
+/* 
+ *    ac3_internal.h
+ *
+ *	Copyright (C) Aaron Holtzman - May 1999
+ *
+ *  This file is part of ac3dec, a free Dolby AC-3 stream decoder.
+ *	
+ *  ac3dec is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2, or (at your option)
+ *  any later version.
+ *   
+ *  ac3dec is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *   
+ *  You should have received a copy of the GNU General Public License
+ *  along with GNU Make; see the file COPYING.  If not, write to
+ *  the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. 
+ *
+ */
+
+#define LEVEL_PLUS3DB 1.4142135623730951
+#define LEVEL_3DB 0.7071067811865476
+#define LEVEL_45DB 0.5946035575013605
+#define LEVEL_6DB 0.5
+
+/* Exponent strategy constants */
+#define EXP_REUSE (0)
+#define EXP_D15   (1)
+#define EXP_D25   (2)
+#define EXP_D45   (3)
+
+/* Delta bit allocation constants */
+#define DELTA_BIT_REUSE (0)
+#define DELTA_BIT_NEW (1)
+#define DELTA_BIT_NONE (2)
+#define DELTA_BIT_RESERVED (3)
+
+void bit_allocate (ac3_state_t * state, ac3_ba_t * ba, int bndstart,
+		   int start, int end, int fastleak, int slowleak,
+		   uint8_t * exp, int8_t * bap);
+
+int downmix_init (int input, int flags, float * level, float clev, float slev);
+void downmix (float * samples, int acmod, int output, float level, float bias,
+	      float clev, float slev);
+
+void imdct_init (void);
+extern void (* imdct_256) (float data[], float delay[]);
+extern void (* imdct_512) (float data[], float delay[]);
diff --git a/libavcodec/libac3/bit_allocate.c b/libavcodec/libac3/bit_allocate.c
new file mode 100644
index 0000000000..307c0074c6
--- /dev/null
+++ b/libavcodec/libac3/bit_allocate.c
@@ -0,0 +1,255 @@
+/* 
+ *  bit_allocate.c
+ *
+ *	Copyright (C) Aaron Holtzman - May 1999
+ *
+ *  This file is part of ac3dec, a free Dolby AC-3 stream decoder.
+ *	
+ *  ac3dec is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2, or (at your option)
+ *  any later version.
+ *   
+ *  ac3dec is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *   
+ *  You should have received a copy of the GNU General Public License
+ *  along with GNU Make; see the file COPYING.  If not, write to
+ *  the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. 
+ *
+ */
+
+#include <inttypes.h>
+#include <stdlib.h>
+#include <string.h>
+#include "ac3.h"
+#include "ac3_internal.h"
+
+static int hthtab[3][50] = {
+    {0x730, 0x730, 0x7c0, 0x800, 0x820, 0x840, 0x850, 0x850, 0x860, 0x860,
+     0x860, 0x860, 0x860, 0x870, 0x870, 0x870, 0x880, 0x880, 0x890, 0x890,
+     0x8a0, 0x8a0, 0x8b0, 0x8b0, 0x8c0, 0x8c0, 0x8d0, 0x8e0, 0x8f0, 0x900,
+     0x910, 0x910, 0x910, 0x910, 0x900, 0x8f0, 0x8c0, 0x870, 0x820, 0x7e0,
+     0x7a0, 0x770, 0x760, 0x7a0, 0x7c0, 0x7c0, 0x6e0, 0x400, 0x3c0, 0x3c0},
+    {0x710, 0x710, 0x7a0, 0x7f0, 0x820, 0x830, 0x840, 0x850, 0x850, 0x860,
+     0x860, 0x860, 0x860, 0x860, 0x870, 0x870, 0x870, 0x880, 0x880, 0x880,
+     0x890, 0x890, 0x8a0, 0x8a0, 0x8b0, 0x8b0, 0x8c0, 0x8c0, 0x8e0, 0x8f0,
+     0x900, 0x910, 0x910, 0x910, 0x910, 0x900, 0x8e0, 0x8b0, 0x870, 0x820,
+     0x7e0, 0x7b0, 0x760, 0x770, 0x7a0, 0x7c0, 0x780, 0x5d0, 0x3c0, 0x3c0},
+    {0x680, 0x680, 0x750, 0x7b0, 0x7e0, 0x810, 0x820, 0x830, 0x840, 0x850,
+     0x850, 0x850, 0x860, 0x860, 0x860, 0x860, 0x860, 0x860, 0x860, 0x860,
+     0x870, 0x870, 0x870, 0x870, 0x880, 0x880, 0x880, 0x890, 0x8a0, 0x8b0,
+     0x8c0, 0x8d0, 0x8e0, 0x8f0, 0x900, 0x910, 0x910, 0x910, 0x900, 0x8f0,
+     0x8d0, 0x8b0, 0x840, 0x7f0, 0x790, 0x760, 0x7a0, 0x7c0, 0x7b0, 0x720}
+};
+
+static int8_t baptab[305] = {
+    16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+    16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+    16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+    16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+    16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+    16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,	// 93 padding entries
+
+    16, 16, 16, 16, 16, 16, 16, 16, 16, 14, 14, 14, 14, 14, 14, 14,
+    14, 12, 12, 12, 12, 11, 11, 11, 11, 10, 10, 10, 10,  9,  9,  9,
+     9,  8,  8,  8,  8,  7,  7,  7,  7,  6,  6,  6,  6,  5,  5,  5,
+     5,  4,  4, -3, -3,  3,  3,  3, -2, -2, -1, -1, -1, -1, -1,  0,
+
+     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
+     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
+     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
+     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
+     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
+     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
+     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
+     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
+     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
+     0,  0,  0,  0					// 148 padding entries
+};
+
+static int bndtab[30] = {21, 22,  23,  24,  25,  26,  27,  28,  31,  34,
+			 37, 40,  43,  46,  49,  55,  61,  67,  73,  79,
+			 85, 97, 109, 121, 133, 157, 181, 205, 229, 253};
+
+static int8_t latab[256] = {
+    -64, -63, -62, -61, -60, -59, -58, -57, -56, -55, -54, -53,
+    -52, -52, -51, -50, -49, -48, -47, -47, -46, -45, -44, -44,
+    -43, -42, -41, -41, -40, -39, -38, -38, -37, -36, -36, -35,
+    -35, -34, -33, -33, -32, -32, -31, -30, -30, -29, -29, -28,
+    -28, -27, -27, -26, -26, -25, -25, -24, -24, -23, -23, -22,
+    -22, -21, -21, -21, -20, -20, -19, -19, -19, -18, -18, -18,
+    -17, -17, -17, -16, -16, -16, -15, -15, -15, -14, -14, -14,
+    -13, -13, -13, -13, -12, -12, -12, -12, -11, -11, -11, -11,
+    -10, -10, -10, -10, -10,  -9,  -9,  -9,  -9,  -9,  -8,  -8,
+     -8,  -8,  -8,  -8,  -7,  -7,  -7,  -7,  -7,  -7,  -6,  -6,
+     -6,  -6,  -6,  -6,  -6,  -6,  -5,  -5,  -5,  -5,  -5,  -5,
+     -5,  -5,  -4,  -4,  -4,  -4,  -4,  -4,  -4,  -4,  -4,  -4,
+     -4,  -3,  -3,  -3,  -3,  -3,  -3,  -3,  -3,  -3,  -3,  -3,
+     -3,  -3,  -3,  -2,  -2,  -2,  -2,  -2,  -2,  -2,  -2,  -2,
+     -2,  -2,  -2,  -2,  -2,  -2,  -2,  -2,  -2,  -2,  -1,  -1,
+     -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,
+     -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,
+     -1,  -1,  -1,  -1,  -1,  -1,   0,   0,   0,   0,   0,   0,
+      0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,
+      0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,
+      0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,
+      0,   0,   0,   0
+};
+
+#define UPDATE_LEAK() 		\
+do {				\
+    fastleak += fdecay;		\
+    if (fastleak > psd + fgain)	\
+	fastleak = psd + fgain;	\
+    slowleak += sdecay;		\
+    if (slowleak > psd + sgain)	\
+	slowleak = psd + sgain;	\
+} while (0)
+
+#define COMPUTE_MASK()				\
+do {						\
+    if (psd > dbknee)				\
+	mask -= (psd - dbknee) >> 2;		\
+    if (mask > hth [i >> halfrate])		\
+	mask = hth [i >> halfrate];		\
+    mask -= snroffset + 128 * deltba[i];	\
+    mask = (mask > 0) ? 0 : ((-mask) >> 5);	\
+    mask -= floor;				\
+} while (0)
+
+void bit_allocate (ac3_state_t * state, ac3_ba_t * ba, int bndstart,
+		   int start, int end, int fastleak, int slowleak,
+		   uint8_t * exp, int8_t * bap)
+{
+    static int slowgain[4] = {0x540, 0x4d8, 0x478, 0x410};
+    static int dbpbtab[4]  = {0xc00, 0x500, 0x300, 0x100};
+    static int floortab[8] = {0x910, 0x950, 0x990, 0x9d0,
+			      0xa10, 0xa90, 0xb10, 0x1400};
+
+    int i, j;
+    int fdecay, fgain, sdecay, sgain, dbknee, floor, snroffset;
+    int psd, mask;
+    int8_t * deltba;
+    int * hth;
+    int halfrate;
+
+    halfrate = state->halfrate;
+    fdecay = (63 + 20 * state->fdcycod) >> halfrate;
+    fgain = 128 + 128 * ba->fgaincod;
+    sdecay = (15 + 2 * state->sdcycod) >> halfrate;
+    sgain = slowgain[state->sgaincod];
+    dbknee = dbpbtab[state->dbpbcod];
+    hth = hthtab[state->fscod];
+    /*
+     * if there is no delta bit allocation, make deltba point to an area
+     * known to contain zeroes. baptab+156 here.
+     */
+    deltba = (ba->deltbae == DELTA_BIT_NONE) ? baptab + 156 : ba->deltba;
+    floor = floortab[state->floorcod];
+    snroffset = 960 - 64 * state->csnroffst - 4 * ba->fsnroffst + floor;
+    floor >>= 5;
+
+    i = bndstart;
+    j = start;
+    if (start == 0) {	// not the coupling channel
+	int lowcomp;
+
+	lowcomp = 0;
+	j = end - 1;
+	do {
+	    if (i < j) {
+		if (exp[i+1] == exp[i] - 2)
+		    lowcomp = 384;
+		else if (lowcomp && (exp[i+1] > exp[i]))
+		    lowcomp -= 64;
+	    }
+	    psd = 128 * exp[i];
+	    mask = psd + fgain + lowcomp;
+	    COMPUTE_MASK ();
+	    bap[i++] = (baptab+156)[mask + 4 * exp[i]];
+	} while ((i < 3) || ((i < 7) && (exp[i] > exp[i-1])));
+	fastleak = psd + fgain;
+	slowleak = psd + sgain;
+
+	while (i < 7) {
+	    if (i < j) {
+		if (exp[i+1] == exp[i] - 2)
+		    lowcomp = 384;
+		else if (lowcomp && (exp[i+1] > exp[i]))
+		    lowcomp -= 64;
+	    }
+	    psd = 128 * exp[i];
+	    UPDATE_LEAK ();
+	    mask = ((fastleak + lowcomp < slowleak) ?
+		    fastleak + lowcomp : slowleak);
+	    COMPUTE_MASK ();
+	    bap[i++] = (baptab+156)[mask + 4 * exp[i]];
+	}
+
+	if (end == 7)	// lfe channel
+	    return;
+
+	do {
+	    if (exp[i+1] == exp[i] - 2)
+		lowcomp = 320;
+	    else if (lowcomp && (exp[i+1] > exp[i]))
+		lowcomp -= 64;
+	    psd = 128 * exp[i];
+	    UPDATE_LEAK ();
+	    mask = ((fastleak + lowcomp < slowleak) ?
+		    fastleak + lowcomp : slowleak);
+	    COMPUTE_MASK ();
+	    bap[i++] = (baptab+156)[mask + 4 * exp[i]];
+	} while (i < 20);
+
+	while (lowcomp > 128) {		// two iterations maximum
+	    lowcomp -= 128;
+	    psd = 128 * exp[i];
+	    UPDATE_LEAK ();
+	    mask = ((fastleak + lowcomp < slowleak) ?
+		    fastleak + lowcomp : slowleak);
+	    COMPUTE_MASK ();
+	    bap[i++] = (baptab+156)[mask + 4 * exp[i]];
+	}
+	j = i;
+    }
+
+    do {
+	int startband, endband;
+
+	startband = j;
+	endband = ((bndtab-20)[i] < end) ? (bndtab-20)[i] : end;
+	psd = 128 * exp[j++];
+	while (j < endband) {
+	    int next, delta;
+
+	    next = 128 * exp[j++];
+	    delta = next - psd;
+	    switch (delta >> 9) {
+	    case -6: case -5: case -4: case -3: case -2:
+		psd = next;
+		break;
+	    case -1:
+		psd = next + latab[(-delta) >> 1];
+		break;
+	    case 0:
+		psd += latab[delta >> 1];
+		break;
+	    }
+	}
+	// minpsd = -289
+	UPDATE_LEAK ();
+	mask = (fastleak < slowleak) ? fastleak : slowleak;
+	COMPUTE_MASK ();
+	i++;
+	j = startband;
+	do {
+	    // max(mask+4*exp)=147=-(minpsd+fgain-deltba-snroffset)>>5+4*exp
+	    // min(mask+4*exp)=-156=-(sgain-deltba-snroffset)>>5
+	    bap[j++] = (baptab+156)[mask + 4 * exp[j]];
+	} while (j < endband);
+    } while (j < end);
+}
diff --git a/libavcodec/libac3/bitstream.c b/libavcodec/libac3/bitstream.c
new file mode 100644
index 0000000000..7e9bd1f66a
--- /dev/null
+++ b/libavcodec/libac3/bitstream.c
@@ -0,0 +1,77 @@
+/* 
+ *  bitstream.c
+ *
+ *	Copyright (C) Aaron Holtzman - Dec 1999
+ *
+ *  This file is part of ac3dec, a free AC-3 audio decoder
+ *	
+ *  ac3dec is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2, or (at your option)
+ *  any later version.
+ *   
+ *  ac3dec is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *   
+ *  You should have received a copy of the GNU General Public License
+ *  along with GNU Make; see the file COPYING.  If not, write to
+ *  the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. 
+ *
+ */
+
+#include <inttypes.h>
+#include <stdlib.h>
+#include <stdio.h>
+
+#include "ac3.h"
+#include "ac3_internal.h"
+#include "bitstream.h"
+
+#define BUFFER_SIZE 4096
+
+static uint8_t *buffer_start;
+
+uint32_t bits_left;
+uint32_t current_word;
+
+void bitstream_set_ptr (uint8_t * buf)
+{
+    buffer_start = buf;
+    bits_left = 0;
+}
+
+static inline void
+bitstream_fill_current()
+{
+    current_word = *((uint32_t*)buffer_start)++;
+    current_word = swab32(current_word);
+}
+
+//
+// The fast paths for _get is in the
+// bitstream.h header file so it can be inlined.
+//
+// The "bottom half" of this routine is suffixed _bh
+//
+// -ah
+//
+
+uint32_t
+bitstream_get_bh(uint32_t num_bits)
+{
+    uint32_t result;
+
+    num_bits -= bits_left;
+    result = (current_word << (32 - bits_left)) >> (32 - bits_left);
+
+    bitstream_fill_current();
+
+    if(num_bits != 0)
+	result = (result << num_bits) | (current_word >> (32 - num_bits));
+	
+    bits_left = 32 - num_bits;
+
+    return result;
+}
diff --git a/libavcodec/libac3/bitstream.h b/libavcodec/libac3/bitstream.h
new file mode 100644
index 0000000000..84f3287c87
--- /dev/null
+++ b/libavcodec/libac3/bitstream.h
@@ -0,0 +1,68 @@
+/* 
+ *  bitstream.h
+ *
+ *	Copyright (C) Aaron Holtzman - Dec 1999
+ *
+ *  This file is part of ac3dec, a free AC-3 audio decoder
+ *	
+ *  ac3dec is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2, or (at your option)
+ *  any later version.
+ *   
+ *  ac3dec is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *   
+ *  You should have received a copy of the GNU General Public License
+ *  along with GNU Make; see the file COPYING.  If not, write to
+ *  the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. 
+ *
+ */
+
+//My new and improved vego-matic endian swapping routine
+//(stolen from the kernel)
+#ifdef WORDS_BIGENDIAN
+
+#	define swab32(x) (x)
+
+#else
+
+#	if defined (__i386__)
+
+#	define swab32(x) __i386_swab32(x)
+	static inline const uint32_t __i386_swab32(uint32_t x)
+	{
+		__asm__("bswap %0" : "=r" (x) : "0" (x));
+		return x;
+	}
+
+#	else
+
+#	define swab32(x)\
+((((uint8_t*)&x)[0] << 24) | (((uint8_t*)&x)[1] << 16) |  \
+ (((uint8_t*)&x)[2] << 8)  | (((uint8_t*)&x)[3]))
+
+#	endif
+#endif
+
+extern uint32_t bits_left;
+extern uint32_t current_word;
+
+void bitstream_set_ptr (uint8_t * buf);
+uint32_t bitstream_get_bh(uint32_t num_bits);
+
+static inline uint32_t 
+bitstream_get(uint32_t num_bits)
+{
+    uint32_t result;
+	
+    if(num_bits < bits_left) {
+	result = (current_word << (32 - bits_left)) >> (32 - num_bits);
+	bits_left -= num_bits;
+	return result;
+    }
+
+    return bitstream_get_bh(num_bits);
+}
diff --git a/libavcodec/libac3/downmix.c b/libavcodec/libac3/downmix.c
new file mode 100644
index 0000000000..9e7fbfb8b6
--- /dev/null
+++ b/libavcodec/libac3/downmix.c
@@ -0,0 +1,533 @@
+/*
+ *
+ *  downmix.c
+ *    
+ *	Copyright (C) Aaron Holtzman - Sept 1999
+ *
+ *	Originally based on code by Yuqing Deng.
+ *
+ *  This file is part of ac3dec, a free Dolby AC-3 stream decoder.
+ *	
+ *  ac3dec is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2, or (at your option)
+ *  any later version.
+ *   
+ *  ac3dec is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *   
+ *  You should have received a copy of the GNU General Public License
+ *  along with GNU Make; see the file COPYING.  If not, write to
+ *  the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. 
+ *
+ *
+ */
+
+#include <inttypes.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <math.h>
+#include <string.h>
+#include "ac3.h"
+#include "ac3_internal.h"
+
+
+#define CONVERT(acmod,output) (((output) << 3) + (acmod))
+
+int downmix_init (int input, int flags, float * level, float clev, float slev)
+{
+    static uint8_t table[11][8] = {
+	{AC3_CHANNEL,  AC3_DOLBY,  AC3_STEREO, AC3_STEREO,
+	 AC3_STEREO,   AC3_STEREO, AC3_STEREO, AC3_STEREO},
+	{AC3_MONO,     AC3_MONO,   AC3_MONO,   AC3_MONO,
+	 AC3_MONO,     AC3_MONO,   AC3_MONO,   AC3_MONO},
+	{AC3_CHANNEL,  AC3_DOLBY,  AC3_STEREO, AC3_STEREO,
+	 AC3_STEREO,   AC3_STEREO, AC3_STEREO, AC3_STEREO},
+	{AC3_CHANNEL,  AC3_DOLBY,  AC3_STEREO, AC3_3F,
+	 AC3_STEREO,   AC3_3F,     AC3_STEREO, AC3_3F},
+	{AC3_CHANNEL,  AC3_DOLBY,  AC3_STEREO, AC3_STEREO,
+	 AC3_2F1R,     AC3_2F1R,   AC3_2F1R,   AC3_2F1R},
+	{AC3_CHANNEL,  AC3_DOLBY,  AC3_STEREO, AC3_STEREO,
+	 AC3_2F1R,     AC3_3F1R,   AC3_2F1R,   AC3_3F1R},
+	{AC3_CHANNEL,  AC3_DOLBY,  AC3_STEREO, AC3_3F,
+	 AC3_2F2R,     AC3_2F2R,   AC3_2F2R,   AC3_2F2R},
+	{AC3_CHANNEL,  AC3_DOLBY,  AC3_STEREO, AC3_3F,
+	 AC3_2F2R,     AC3_3F2R,   AC3_2F2R,   AC3_3F2R},
+	{AC3_CHANNEL1, AC3_MONO,   AC3_MONO,   AC3_MONO,
+	 AC3_MONO,     AC3_MONO,   AC3_MONO,   AC3_MONO},
+	{AC3_CHANNEL2, AC3_MONO,   AC3_MONO,   AC3_MONO,
+	 AC3_MONO,     AC3_MONO,   AC3_MONO,   AC3_MONO},
+	{AC3_CHANNEL,  AC3_DOLBY,  AC3_STEREO, AC3_DOLBY,
+	 AC3_DOLBY,    AC3_DOLBY,  AC3_DOLBY,  AC3_DOLBY}
+    };
+    int output;
+
+    output = flags & AC3_CHANNEL_MASK;
+    if (output > AC3_DOLBY)
+	return -1;
+
+    output = table[output][input & 7];
+
+    if ((output == AC3_STEREO) &&
+	((input == AC3_DOLBY) || ((input == AC3_3F) && (clev == LEVEL_3DB))))
+	output = AC3_DOLBY;
+
+    if (flags & AC3_ADJUST_LEVEL)
+	switch (CONVERT (input & 7, output)) {
+
+	case CONVERT (AC3_3F, AC3_MONO):
+	    *level *= LEVEL_3DB / (1 + clev);
+	    break;
+
+	case CONVERT (AC3_STEREO, AC3_MONO):
+	case CONVERT (AC3_2F2R, AC3_2F1R):
+	case CONVERT (AC3_3F2R, AC3_3F1R):
+	level_3db:
+	    *level *= LEVEL_3DB;
+	    break;
+
+	case CONVERT (AC3_3F2R, AC3_2F1R):
+	    if (clev < LEVEL_PLUS3DB - 1)
+		goto level_3db;
+	    // break thru
+	case CONVERT (AC3_3F, AC3_STEREO):
+	case CONVERT (AC3_3F1R, AC3_2F1R):
+	case CONVERT (AC3_3F1R, AC3_2F2R):
+	case CONVERT (AC3_3F2R, AC3_2F2R):
+	    *level /= 1 + clev;
+	    break;
+
+	case CONVERT (AC3_2F1R, AC3_MONO):
+	    *level *= LEVEL_PLUS3DB / (2 + slev);
+	    break;
+
+	case CONVERT (AC3_2F1R, AC3_STEREO):
+	case CONVERT (AC3_3F1R, AC3_3F):
+	    *level /= 1 + slev * LEVEL_3DB;
+	    break;
+
+	case CONVERT (AC3_3F1R, AC3_MONO):
+	    *level *= LEVEL_3DB / (1 + clev + 0.5 * slev);
+	    break;
+
+	case CONVERT (AC3_3F1R, AC3_STEREO):
+	    *level /= 1 + clev + slev * LEVEL_3DB;
+	    break;
+
+	case CONVERT (AC3_2F2R, AC3_MONO):
+	    *level *= LEVEL_3DB / (1 + slev);
+	    break;
+
+	case CONVERT (AC3_2F2R, AC3_STEREO):
+	case CONVERT (AC3_3F2R, AC3_3F):
+	    *level /= (1 + slev);
+	    break;
+
+	case CONVERT (AC3_3F2R, AC3_MONO):
+	    *level *= LEVEL_3DB / (1 + clev + slev);
+	    break;
+
+	case CONVERT (AC3_3F2R, AC3_STEREO):
+	    *level /= 1 + clev + slev;
+	    break;
+
+	case CONVERT (AC3_MONO, AC3_DOLBY):
+	    *level *= LEVEL_PLUS3DB;
+	    break;
+
+	case CONVERT (AC3_3F, AC3_DOLBY):
+	case CONVERT (AC3_2F1R, AC3_DOLBY):
+	    *level *= 1 / (1 + LEVEL_3DB);
+	    break;
+
+	case CONVERT (AC3_3F1R, AC3_DOLBY):
+	case CONVERT (AC3_2F2R, AC3_DOLBY):
+	    *level *= 1 / (1 + 2 * LEVEL_3DB);
+	    break;
+
+	case CONVERT (AC3_3F2R, AC3_DOLBY):
+	    *level *= 1 / (1 + 3 * LEVEL_3DB);
+	    break;
+    }
+
+    return output;
+}
+
+static void mix1to1 (float * samples, float level, float bias)
+{
+    int i;
+
+    for (i = 0; i < 256; i++)
+	samples[i] = samples[i] * level + bias;
+}
+
+static void move1to1 (float * src, float * dest, float level, float bias)
+{
+    int i;
+
+    for (i = 0; i < 256; i++)
+	dest[i] = src[i] * level + bias;
+}
+
+static void mix2to1 (float * samples, float level, float bias)
+{
+    int i;
+
+    for (i = 0; i < 256; i++)
+	samples[i] = (samples[i] + samples[i + 256]) * level + bias;
+}
+
+static void move2to1 (float * src, float * dest, float level, float bias)
+{
+    int i;
+
+    for (i = 0; i < 256; i++)
+	dest[i] = (src[i] + src[i + 256]) * level + bias;
+}
+
+static void mix3to1 (float * samples, float level, float clev, float bias)
+{
+    int i;
+
+    for (i = 0; i < 256; i++)
+	samples[i] = ((samples[i] + samples[i + 512]) * level +
+		      samples[i + 256] * clev + bias);
+}
+
+static void mix21to1 (float * samples, float level, float slev, float bias)
+{
+    int i;
+
+    for (i = 0; i < 256; i++)
+	samples[i] = ((samples[i] + samples[i + 256]) * level +
+		      samples[i + 512] * slev + bias);
+}
+
+static void mix31to1 (float * samples, float level, float clev, float slev,
+		      float bias)
+{
+    int i;
+
+    for (i = 0; i < 256; i++)
+	samples[i] = ((samples[i] + samples[i + 512]) * level +
+		      samples[i + 256] * clev + samples[i + 768] * slev +
+		      bias);
+}
+
+static void mix22to1 (float * samples, float level, float slev, float bias)
+{
+    int i;
+
+    for (i = 0; i < 256; i++)
+	samples[i] = ((samples[i] + samples[i + 256]) * level +
+		      (samples[i + 512] + samples[i + 768]) * slev + bias);
+}
+
+static void mix32to1 (float * samples, float level, float clev, float slev,
+		      float bias)
+{
+    int i;
+
+    for (i = 0; i < 256; i++)
+	samples[i] = ((samples[i] + samples[i + 512]) * level +
+		      samples[i + 256] * clev +
+		      (samples[i + 768] + samples[i + 1024]) * slev + bias);
+}
+
+static void mix1to2 (float * src, float * dest, float level, float bias)
+{
+    int i;
+
+    for (i = 0; i < 256; i++)
+	dest[i] = src[i] = src[i] * level + bias;
+}
+
+static void mix3to2 (float * samples, float level, float clev, float bias)
+{
+    int i;
+    float common;
+
+    for (i = 0; i < 256; i++) {
+	common = samples[i + 256] * clev + bias;
+	samples[i] = samples[i] * level + common;
+	samples[i + 256] = samples[i + 512] * level + common;
+    }
+}
+
+static void mix21to2 (float * left, float * right, float level, float slev,
+		      float bias)
+{
+    int i;
+    float common;
+
+    for (i = 0; i < 256; i++) {
+	common = right[i + 256] * slev + bias;
+	left[i] = left[i] * level + common;
+	right[i] = right[i] * level + common;
+    }
+}
+
+static void mix11to1 (float * front, float * rear, float level, float slev,
+		      float bias)
+{
+    int i;
+
+    for (i = 0; i < 256; i++)
+	front[i] = front[i] * level + rear[i] * slev + bias;
+}
+
+static void mix31to2 (float * samples, float level, float clev, float slev,
+		      float bias)
+{
+    int i;
+    float common;
+
+    for (i = 0; i < 256; i++) {
+	common = samples[i + 256] * clev + samples[i + 768] * slev + bias;
+	samples[i] = samples[i] * level + common;
+	samples[i + 256] = samples[i + 512] * level + common;
+    }
+}
+
+static void mix32to2 (float * samples, float level, float clev, float slev,
+		      float bias)
+{
+    int i;
+    float common;
+
+    for (i = 0; i < 256; i++) {
+	common = samples[i + 256] * clev + bias;
+	samples[i] = samples[i] * level + common + samples[i + 768] * slev;
+	samples[i + 256] = (samples[i + 512] * level + common +
+			    samples[i + 1024] * slev);
+    }
+}
+
+static void mix21toS (float * samples, float level, float level3db, float bias)
+{
+    int i;
+    float surround;
+
+    for (i = 0; i < 256; i++) {
+	surround = samples[i + 512] * level3db;
+	samples[i] = samples[i] * level - surround + bias;
+	samples[i + 256] = samples[i + 256] * level + surround + bias;
+    }
+}
+
+static void mix22toS (float * samples, float level, float level3db, float bias)
+{
+    int i;
+    float surround;
+
+    for (i = 0; i < 256; i++) {
+	surround = (samples[i + 512] + samples[i + 768]) * level3db;
+	samples[i] = samples[i] * level - surround + bias;
+	samples[i + 256] = samples[i + 256] * level + surround + bias;
+    }
+}
+
+static void mix31toS (float * samples, float level, float level3db, float bias)
+{
+    int i;
+    float common, surround;
+
+    for (i = 0; i < 256; i++) {
+	common = samples[i + 256] * level3db + bias;
+	surround = samples[i + 768] * level3db;
+	samples[i] = samples[i] * level + common - surround;
+	samples[i + 256] = samples[i + 512] * level + common + surround;
+    }
+}
+
+static void mix32toS (float * samples, float level, float level3db, float bias)
+{
+    int i;
+    float common, surround;
+
+    for (i = 0; i < 256; i++) {
+	common = samples[i + 256] * level3db + bias;
+	surround = (samples[i + 768] + samples[i + 1024]) * level3db;
+	samples[i] = samples[i] * level + common - surround;
+	samples[i + 256] = samples[i + 512] * level + common + surround;
+    }
+}
+
+void downmix (float * samples, int acmod, int output, float level, float bias,
+	      float clev, float slev)
+{
+    switch (CONVERT (acmod, output & AC3_CHANNEL_MASK)) {
+
+    case CONVERT (AC3_3F2R, AC3_3F2R):
+	mix1to1 (samples + 1024, level, bias);
+    case CONVERT (AC3_3F1R, AC3_3F1R):
+    case CONVERT (AC3_2F2R, AC3_2F2R):
+	mix1to1 (samples + 768, level, bias);
+    case CONVERT (AC3_3F, AC3_3F):
+    case CONVERT (AC3_2F1R, AC3_2F1R):
+    mix_3to3:
+	mix1to1 (samples + 512, level, bias);
+    case CONVERT (AC3_CHANNEL, AC3_CHANNEL):
+    case CONVERT (AC3_STEREO, AC3_STEREO):
+    case CONVERT (AC3_STEREO, AC3_DOLBY):
+    mix_2to2:
+	mix1to1 (samples + 256, level, bias);
+    case CONVERT (AC3_CHANNEL, AC3_CHANNEL1):
+    case CONVERT (AC3_MONO, AC3_MONO):
+	mix1to1 (samples, level, bias);
+	break;
+
+    case CONVERT (AC3_CHANNEL, AC3_CHANNEL2):
+    mix_1to1_b:
+	mix1to1 (samples + 256, level, bias);
+	break;
+
+    case CONVERT (AC3_STEREO, AC3_MONO):
+    mix_2to1:
+	mix2to1 (samples, level * LEVEL_3DB, bias);
+	break;
+
+    case CONVERT (AC3_2F1R, AC3_MONO):
+	if (slev == 0)
+	    goto mix_2to1;
+	mix21to1 (samples, level * LEVEL_3DB, level * slev * LEVEL_3DB, bias);
+	break;
+
+    case CONVERT (AC3_2F2R, AC3_MONO):
+	if (slev == 0)
+	    goto mix_2to1;
+	mix22to1 (samples, level * LEVEL_3DB, level * slev * LEVEL_3DB, bias);
+	break;
+
+    case CONVERT (AC3_3F, AC3_MONO):
+    mix_3to1:
+	mix3to1 (samples, level * LEVEL_3DB, level * clev * LEVEL_PLUS3DB,
+		 bias);
+	break;
+
+    case CONVERT (AC3_3F1R, AC3_MONO):
+	if (slev == 0)
+	    goto mix_3to1;
+	mix31to1 (samples, level * LEVEL_3DB, level * clev * LEVEL_PLUS3DB,
+		  level * slev * LEVEL_3DB, bias);
+	break;
+
+    case CONVERT (AC3_3F2R, AC3_MONO):
+	if (slev == 0)
+	    goto mix_3to1;
+	mix32to1 (samples, level * LEVEL_3DB, level * clev * LEVEL_PLUS3DB,
+		  level * slev * LEVEL_3DB, bias);
+	break;
+
+    case CONVERT (AC3_CHANNEL, AC3_MONO):
+	mix2to1 (samples, level * LEVEL_6DB, bias);
+	break;
+
+    case CONVERT (AC3_MONO, AC3_DOLBY):
+	mix1to2 (samples, samples + 256, level * LEVEL_3DB, bias);
+	break;
+
+    case CONVERT (AC3_3F, AC3_DOLBY):
+	clev = LEVEL_3DB;
+    case CONVERT (AC3_3F, AC3_STEREO):
+    mix_3to2:
+	mix3to2 (samples, level, level * clev, bias);
+	break;
+
+    case CONVERT (AC3_2F1R, AC3_DOLBY):
+	mix21toS (samples, level, level * LEVEL_3DB, bias);
+	break;
+
+    case CONVERT (AC3_3F1R, AC3_DOLBY):
+	mix31toS (samples, level, level * LEVEL_3DB, bias);
+	break;
+
+    case CONVERT (AC3_2F2R, AC3_DOLBY):
+	mix22toS (samples, level, level * LEVEL_3DB, bias);
+	break;
+
+    case CONVERT (AC3_3F2R, AC3_DOLBY):
+	mix32toS (samples, level, level * LEVEL_3DB, bias);
+	break;
+
+    case CONVERT (AC3_2F1R, AC3_STEREO):
+	if (slev == 0)
+	    goto mix_2to2;
+	mix21to2 (samples, samples + 256, level, level * slev * LEVEL_3DB,
+		  bias);
+	break;
+
+    case CONVERT (AC3_3F1R, AC3_STEREO):
+	if (slev == 0)
+	    goto mix_3to2;
+	mix31to2 (samples, level, level * clev, level * slev * LEVEL_3DB,
+		  bias);
+	break;
+
+    case CONVERT (AC3_2F2R, AC3_STEREO):
+	if (slev == 0)
+	    goto mix_2to2;
+	mix11to1 (samples, samples + 512, level, level * slev, bias);
+	mix11to1 (samples + 256, samples + 768, level, level * slev, bias);
+	break;
+
+    case CONVERT (AC3_3F2R, AC3_STEREO):
+	if (slev == 0)
+	    goto mix_3to2;
+	mix32to2 (samples, level, level * clev, level * slev, bias);
+	break;
+
+    case CONVERT (AC3_3F1R, AC3_3F):
+	if (slev == 0)
+	    goto mix_3to3;
+	mix21to2 (samples, samples + 512, level, level * slev * LEVEL_3DB,
+		  bias);
+
+    case CONVERT (AC3_3F2R, AC3_3F):
+	if (slev == 0)
+	    goto mix_3to3;
+	mix11to1 (samples, samples + 768, level, level * slev, bias);
+	mix11to1 (samples + 512, samples + 1024, level, level * slev, bias);
+	goto mix_1to1_b;
+
+    case CONVERT (AC3_2F1R, AC3_2F2R):
+	mix1to2 (samples + 512, samples + 768, level * LEVEL_3DB, bias);
+	goto mix_2to2;
+
+    case CONVERT (AC3_3F1R, AC3_3F2R):
+	mix1to2 (samples + 768, samples + 1024, level * LEVEL_3DB, bias);
+	goto mix_3to3;
+
+    case CONVERT (AC3_2F2R, AC3_2F1R):
+	mix2to1 (samples + 512, level * LEVEL_3DB, bias);
+	goto mix_2to2;
+
+    case CONVERT (AC3_3F2R, AC3_3F1R):
+	mix2to1 (samples + 768, level * LEVEL_3DB, bias);
+	goto mix_3to3;
+
+    case CONVERT (AC3_3F1R, AC3_2F2R):
+	mix3to2 (samples, level, level * clev, bias);
+	mix1to2 (samples + 768, samples + 512, level * LEVEL_3DB, bias);
+	break;
+
+    case CONVERT (AC3_3F1R, AC3_2F1R):
+	mix3to2 (samples, level, level * clev, bias);
+	move1to1 (samples + 768, samples + 512, level, bias);
+	break;
+
+    case CONVERT (AC3_3F2R, AC3_2F1R):
+	mix3to2 (samples, level, level * clev, bias);
+	move2to1 (samples + 768, samples + 512, level * LEVEL_3DB, bias);
+	break;
+
+    case CONVERT (AC3_3F2R, AC3_2F2R):
+	mix3to2 (samples, level, level * clev, bias);
+	move1to1 (samples + 768, samples + 512, level, bias);
+	move1to1 (samples + 1024, samples + 768, level, bias);
+	break;
+
+    }
+}
diff --git a/libavcodec/libac3/imdct.c b/libavcodec/libac3/imdct.c
new file mode 100644
index 0000000000..7be33897f5
--- /dev/null
+++ b/libavcodec/libac3/imdct.c
@@ -0,0 +1,408 @@
+/* 
+ *  imdct.c
+ *
+ *	Copyright (C) Aaron Holtzman - May 1999
+ *
+ *  This file is part of ac3dec, a free Dolby AC-3 stream decoder.
+ *	
+ *  ac3dec is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2, or (at your option)
+ *  any later version.
+ *   
+ *  ac3dec is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *   
+ *  You should have received a copy of the GNU General Public License
+ *  along with GNU Make; see the file COPYING.  If not, write to
+ *  the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. 
+ *
+ *
+ */
+
+//#include "config.h"
+
+#include <inttypes.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <math.h>
+#include "ac3.h"
+#include "ac3_internal.h"
+
+void (* imdct_256) (float data[], float delay[]);
+void (* imdct_512) (float data[], float delay[]);
+
+typedef struct complex_s
+{
+    float real;
+    float imag;
+} complex_t;
+
+
+/* 128 point bit-reverse LUT */
+static uint8_t bit_reverse_512[] = {
+	0x00, 0x40, 0x20, 0x60, 0x10, 0x50, 0x30, 0x70, 
+	0x08, 0x48, 0x28, 0x68, 0x18, 0x58, 0x38, 0x78, 
+	0x04, 0x44, 0x24, 0x64, 0x14, 0x54, 0x34, 0x74, 
+	0x0c, 0x4c, 0x2c, 0x6c, 0x1c, 0x5c, 0x3c, 0x7c, 
+	0x02, 0x42, 0x22, 0x62, 0x12, 0x52, 0x32, 0x72, 
+	0x0a, 0x4a, 0x2a, 0x6a, 0x1a, 0x5a, 0x3a, 0x7a, 
+	0x06, 0x46, 0x26, 0x66, 0x16, 0x56, 0x36, 0x76, 
+	0x0e, 0x4e, 0x2e, 0x6e, 0x1e, 0x5e, 0x3e, 0x7e, 
+	0x01, 0x41, 0x21, 0x61, 0x11, 0x51, 0x31, 0x71, 
+	0x09, 0x49, 0x29, 0x69, 0x19, 0x59, 0x39, 0x79, 
+	0x05, 0x45, 0x25, 0x65, 0x15, 0x55, 0x35, 0x75, 
+	0x0d, 0x4d, 0x2d, 0x6d, 0x1d, 0x5d, 0x3d, 0x7d, 
+	0x03, 0x43, 0x23, 0x63, 0x13, 0x53, 0x33, 0x73, 
+	0x0b, 0x4b, 0x2b, 0x6b, 0x1b, 0x5b, 0x3b, 0x7b, 
+	0x07, 0x47, 0x27, 0x67, 0x17, 0x57, 0x37, 0x77, 
+	0x0f, 0x4f, 0x2f, 0x6f, 0x1f, 0x5f, 0x3f, 0x7f};
+
+static uint8_t bit_reverse_256[] = {
+	0x00, 0x20, 0x10, 0x30, 0x08, 0x28, 0x18, 0x38, 
+	0x04, 0x24, 0x14, 0x34, 0x0c, 0x2c, 0x1c, 0x3c, 
+	0x02, 0x22, 0x12, 0x32, 0x0a, 0x2a, 0x1a, 0x3a, 
+	0x06, 0x26, 0x16, 0x36, 0x0e, 0x2e, 0x1e, 0x3e, 
+	0x01, 0x21, 0x11, 0x31, 0x09, 0x29, 0x19, 0x39, 
+	0x05, 0x25, 0x15, 0x35, 0x0d, 0x2d, 0x1d, 0x3d, 
+	0x03, 0x23, 0x13, 0x33, 0x0b, 0x2b, 0x1b, 0x3b, 
+	0x07, 0x27, 0x17, 0x37, 0x0f, 0x2f, 0x1f, 0x3f};
+
+static complex_t buf[128];
+
+/* Twiddle factor LUT */
+static complex_t w_1[1];
+static complex_t w_2[2];
+static complex_t w_4[4];
+static complex_t w_8[8];
+static complex_t w_16[16];
+static complex_t w_32[32];
+static complex_t w_64[64];
+static complex_t * w[7] = {w_1, w_2, w_4, w_8, w_16, w_32, w_64};
+
+/* Twiddle factors for IMDCT */
+static float xcos1[128];
+static float xsin1[128];
+static float xcos2[64];
+static float xsin2[64];
+
+/* Windowing function for Modified DCT - Thank you acroread */
+float imdct_window[] = {
+	0.00014, 0.00024, 0.00037, 0.00051, 0.00067, 0.00086, 0.00107, 0.00130,
+	0.00157, 0.00187, 0.00220, 0.00256, 0.00297, 0.00341, 0.00390, 0.00443,
+	0.00501, 0.00564, 0.00632, 0.00706, 0.00785, 0.00871, 0.00962, 0.01061,
+	0.01166, 0.01279, 0.01399, 0.01526, 0.01662, 0.01806, 0.01959, 0.02121,
+	0.02292, 0.02472, 0.02662, 0.02863, 0.03073, 0.03294, 0.03527, 0.03770,
+	0.04025, 0.04292, 0.04571, 0.04862, 0.05165, 0.05481, 0.05810, 0.06153,
+	0.06508, 0.06878, 0.07261, 0.07658, 0.08069, 0.08495, 0.08935, 0.09389,
+	0.09859, 0.10343, 0.10842, 0.11356, 0.11885, 0.12429, 0.12988, 0.13563,
+	0.14152, 0.14757, 0.15376, 0.16011, 0.16661, 0.17325, 0.18005, 0.18699,
+	0.19407, 0.20130, 0.20867, 0.21618, 0.22382, 0.23161, 0.23952, 0.24757,
+	0.25574, 0.26404, 0.27246, 0.28100, 0.28965, 0.29841, 0.30729, 0.31626,
+	0.32533, 0.33450, 0.34376, 0.35311, 0.36253, 0.37204, 0.38161, 0.39126,
+	0.40096, 0.41072, 0.42054, 0.43040, 0.44030, 0.45023, 0.46020, 0.47019,
+	0.48020, 0.49022, 0.50025, 0.51028, 0.52031, 0.53033, 0.54033, 0.55031,
+	0.56026, 0.57019, 0.58007, 0.58991, 0.59970, 0.60944, 0.61912, 0.62873,
+	0.63827, 0.64774, 0.65713, 0.66643, 0.67564, 0.68476, 0.69377, 0.70269,
+	0.71150, 0.72019, 0.72877, 0.73723, 0.74557, 0.75378, 0.76186, 0.76981,
+	0.77762, 0.78530, 0.79283, 0.80022, 0.80747, 0.81457, 0.82151, 0.82831,
+	0.83496, 0.84145, 0.84779, 0.85398, 0.86001, 0.86588, 0.87160, 0.87716,
+	0.88257, 0.88782, 0.89291, 0.89785, 0.90264, 0.90728, 0.91176, 0.91610,
+	0.92028, 0.92432, 0.92822, 0.93197, 0.93558, 0.93906, 0.94240, 0.94560,
+	0.94867, 0.95162, 0.95444, 0.95713, 0.95971, 0.96217, 0.96451, 0.96674,
+	0.96887, 0.97089, 0.97281, 0.97463, 0.97635, 0.97799, 0.97953, 0.98099,
+	0.98236, 0.98366, 0.98488, 0.98602, 0.98710, 0.98811, 0.98905, 0.98994,
+	0.99076, 0.99153, 0.99225, 0.99291, 0.99353, 0.99411, 0.99464, 0.99513,
+	0.99558, 0.99600, 0.99639, 0.99674, 0.99706, 0.99736, 0.99763, 0.99788,
+	0.99811, 0.99831, 0.99850, 0.99867, 0.99882, 0.99895, 0.99908, 0.99919,
+	0.99929, 0.99938, 0.99946, 0.99953, 0.99959, 0.99965, 0.99969, 0.99974,
+	0.99978, 0.99981, 0.99984, 0.99986, 0.99988, 0.99990, 0.99992, 0.99993,
+	0.99994, 0.99995, 0.99996, 0.99997, 0.99998, 0.99998, 0.99998, 0.99999,
+	0.99999, 0.99999, 0.99999, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000,
+	1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000 };
+
+
+static inline void swap_cmplx(complex_t *a, complex_t *b)
+{
+    complex_t tmp;
+
+    tmp = *a;
+    *a = *b;
+    *b = tmp;
+}
+
+
+
+static inline complex_t cmplx_mult(complex_t a, complex_t b)
+{
+    complex_t ret;
+
+    ret.real = a.real * b.real - a.imag * b.imag;
+    ret.imag = a.real * b.imag + a.imag * b.real;
+
+    return ret;
+}
+
+void
+imdct_do_512(float data[],float delay[])
+{
+    int i,k;
+    int p,q;
+    int m;
+    int two_m;
+    int two_m_plus_one;
+
+    float tmp_a_i;
+    float tmp_a_r;
+    float tmp_b_i;
+    float tmp_b_r;
+
+    float *data_ptr;
+    float *delay_ptr;
+    float *window_ptr;
+	
+    //
+    // 512 IMDCT with source and dest data in 'data'
+    //
+	
+    // Pre IFFT complex multiply plus IFFT cmplx conjugate 
+    for( i=0; i < 128; i++) {
+	/* z[i] = (X[256-2*i-1] + j * X[2*i]) * (xcos1[i] + j * xsin1[i]) ; */ 
+	buf[i].real =         (data[256-2*i-1] * xcos1[i])  -  (data[2*i]       * xsin1[i]);
+	buf[i].imag = -1.0 * ((data[2*i]       * xcos1[i])  +  (data[256-2*i-1] * xsin1[i]));
+    }
+
+    //Bit reversed shuffling
+    for(i=0; i<128; i++) {
+	k = bit_reverse_512[i];
+	if (k < i)
+	    swap_cmplx(&buf[i],&buf[k]);
+    }
+
+    /* FFT Merge */
+    for (m=0; m < 7; m++) {
+	if(m)
+	    two_m = (1 << m);
+	else
+	    two_m = 1;
+
+	two_m_plus_one = (1 << (m+1));
+
+	for(k = 0; k < two_m; k++) {
+	    for(i = 0; i < 128; i += two_m_plus_one) {
+		p = k + i;
+		q = p + two_m;
+		tmp_a_r = buf[p].real;
+		tmp_a_i = buf[p].imag;
+		tmp_b_r = buf[q].real * w[m][k].real - buf[q].imag * w[m][k].imag;
+		tmp_b_i = buf[q].imag * w[m][k].real + buf[q].real * w[m][k].imag;
+		buf[p].real = tmp_a_r + tmp_b_r;
+		buf[p].imag =  tmp_a_i + tmp_b_i;
+		buf[q].real = tmp_a_r - tmp_b_r;
+		buf[q].imag =  tmp_a_i - tmp_b_i;
+	    }
+	}
+    }
+
+    /* Post IFFT complex multiply  plus IFFT complex conjugate*/
+    for( i=0; i < 128; i++) {
+	/* y[n] = z[n] * (xcos1[n] + j * xsin1[n]) ; */
+	tmp_a_r =        buf[i].real;
+	tmp_a_i = -1.0 * buf[i].imag;
+	buf[i].real =(tmp_a_r * xcos1[i])  -  (tmp_a_i  * xsin1[i]);
+	buf[i].imag =(tmp_a_r * xsin1[i])  +  (tmp_a_i  * xcos1[i]);
+    }
+	
+    data_ptr = data;
+    delay_ptr = delay;
+    window_ptr = imdct_window;
+
+    /* Window and convert to real valued signal */
+    for(i=0; i< 64; i++) { 
+	*data_ptr++   = 2.0f * (-buf[64+i].imag   * *window_ptr++ + *delay_ptr++); 
+	*data_ptr++   = 2.0f * ( buf[64-i-1].real * *window_ptr++ + *delay_ptr++); 
+    }
+
+    for(i=0; i< 64; i++) { 
+	*data_ptr++  = 2.0f * (-buf[i].real       * *window_ptr++ + *delay_ptr++); 
+	*data_ptr++  = 2.0f * ( buf[128-i-1].imag * *window_ptr++ + *delay_ptr++); 
+    }
+
+    /* The trailing edge of the window goes into the delay line */
+    delay_ptr = delay;
+
+    for(i=0; i< 64; i++) { 
+	*delay_ptr++  = -buf[64+i].real   * *--window_ptr; 
+	*delay_ptr++  =  buf[64-i-1].imag * *--window_ptr; 
+    }
+
+    for(i=0; i<64; i++) {
+	*delay_ptr++  =  buf[i].imag       * *--window_ptr; 
+	*delay_ptr++  = -buf[128-i-1].real * *--window_ptr; 
+    }
+}
+
+void
+imdct_do_256(float data[],float delay[])
+{
+    int i,k;
+    int p,q;
+    int m;
+    int two_m;
+    int two_m_plus_one;
+
+    float tmp_a_i;
+    float tmp_a_r;
+    float tmp_b_i;
+    float tmp_b_r;
+
+    float *data_ptr;
+    float *delay_ptr;
+    float *window_ptr;
+
+    complex_t *buf_1, *buf_2;
+
+    buf_1 = &buf[0];
+    buf_2 = &buf[64];
+
+    /* Pre IFFT complex multiply plus IFFT cmplx conjugate */
+    for(k=0; k<64; k++) { 
+	/* X1[k] = X[2*k]  */
+	/* X2[k] = X[2*k+1]     */
+
+	p = 2 * (128-2*k-1);
+	q = 2 * (2 * k);
+
+	/* Z1[k] = (X1[128-2*k-1] + j * X1[2*k]) * (xcos2[k] + j * xsin2[k]); */ 
+	buf_1[k].real =         data[p] * xcos2[k] - data[q] * xsin2[k];
+	buf_1[k].imag = -1.0f * (data[q] * xcos2[k] + data[p] * xsin2[k]); 
+	/* Z2[k] = (X2[128-2*k-1] + j * X2[2*k]) * (xcos2[k] + j * xsin2[k]); */ 
+	buf_2[k].real =          data[p + 1] * xcos2[k] - data[q + 1] * xsin2[k];
+	buf_2[k].imag = -1.0f * ( data[q + 1] * xcos2[k] + data[p + 1] * xsin2[k]); 
+    }
+
+    //IFFT Bit reversed shuffling
+    for(i=0; i<64; i++) { 
+	k = bit_reverse_256[i];
+	if (k < i) {
+	    swap_cmplx(&buf_1[i],&buf_1[k]);
+	    swap_cmplx(&buf_2[i],&buf_2[k]);
+	}
+    }
+
+    /* FFT Merge */
+    for (m=0; m < 6; m++) {
+	two_m = (1 << m);
+	two_m_plus_one = (1 << (m+1));
+
+	//FIXME
+	if(m)
+	    two_m = (1 << m);
+	else
+	    two_m = 1;
+
+	for(k = 0; k < two_m; k++) {
+	    for(i = 0; i < 64; i += two_m_plus_one) {
+		p = k + i;
+		q = p + two_m;
+		//Do block 1
+		tmp_a_r = buf_1[p].real;
+		tmp_a_i = buf_1[p].imag;
+		tmp_b_r = buf_1[q].real * w[m][k].real - buf_1[q].imag * w[m][k].imag;
+		tmp_b_i = buf_1[q].imag * w[m][k].real + buf_1[q].real * w[m][k].imag;
+		buf_1[p].real = tmp_a_r + tmp_b_r;
+		buf_1[p].imag =  tmp_a_i + tmp_b_i;
+		buf_1[q].real = tmp_a_r - tmp_b_r;
+		buf_1[q].imag =  tmp_a_i - tmp_b_i;
+
+		//Do block 2
+		tmp_a_r = buf_2[p].real;
+		tmp_a_i = buf_2[p].imag;
+		tmp_b_r = buf_2[q].real * w[m][k].real - buf_2[q].imag * w[m][k].imag;
+		tmp_b_i = buf_2[q].imag * w[m][k].real + buf_2[q].real * w[m][k].imag;
+		buf_2[p].real = tmp_a_r + tmp_b_r;
+		buf_2[p].imag =  tmp_a_i + tmp_b_i;
+		buf_2[q].real = tmp_a_r - tmp_b_r;
+		buf_2[q].imag =  tmp_a_i - tmp_b_i;
+	    }
+	}
+    }
+
+    /* Post IFFT complex multiply */
+    for( i=0; i < 64; i++) {
+	/* y1[n] = z1[n] * (xcos2[n] + j * xs in2[n]) ; */ 
+	tmp_a_r =  buf_1[i].real;
+	tmp_a_i = -buf_1[i].imag;
+	buf_1[i].real =(tmp_a_r * xcos2[i])  -  (tmp_a_i  * xsin2[i]);
+	buf_1[i].imag =(tmp_a_r * xsin2[i])  +  (tmp_a_i  * xcos2[i]);
+	/* y2[n] = z2[n] * (xcos2[n] + j * xsin2[n]) ; */ 
+	tmp_a_r =  buf_2[i].real;
+	tmp_a_i = -buf_2[i].imag;
+	buf_2[i].real =(tmp_a_r * xcos2[i])  -  (tmp_a_i  * xsin2[i]);
+	buf_2[i].imag =(tmp_a_r * xsin2[i])  +  (tmp_a_i  * xcos2[i]);
+    }
+	
+    data_ptr = data;
+    delay_ptr = delay;
+    window_ptr = imdct_window;
+
+    /* Window and convert to real valued signal */
+    for(i=0; i< 64; i++) { 
+	*data_ptr++  = 2.0f * (-buf_1[i].imag      * *window_ptr++ + *delay_ptr++);
+	*data_ptr++  = 2.0f * ( buf_1[64-i-1].real * *window_ptr++ + *delay_ptr++);
+    }
+
+    for(i=0; i< 64; i++) {
+	*data_ptr++  = 2.0f * (-buf_1[i].real      * *window_ptr++ + *delay_ptr++);
+	*data_ptr++  = 2.0f * ( buf_1[64-i-1].imag * *window_ptr++ + *delay_ptr++);
+    }
+	
+    delay_ptr = delay;
+
+    for(i=0; i< 64; i++) {
+	*delay_ptr++ = -buf_2[i].real      * *--window_ptr;
+	*delay_ptr++ =  buf_2[64-i-1].imag * *--window_ptr;
+    }
+
+    for(i=0; i< 64; i++) {
+	*delay_ptr++ =  buf_2[i].imag      * *--window_ptr;
+	*delay_ptr++ = -buf_2[64-i-1].real * *--window_ptr;
+    }
+}
+
+void imdct_init (void)
+{
+#ifdef LIBAC3_MLIB
+    void imdct_do_256_mlib(float data[],float delay[]);
+    void imdct_do_512_mlib(float data[],float delay[]);
+
+    imdct_512 = imdct_do_512_mlib;
+    imdct_256 = imdct_do_256_mlib;
+#else
+    int i, j, k;
+
+    /* Twiddle factors to turn IFFT into IMDCT */
+    for (i = 0; i < 128; i++) {
+	xcos1[i] = -cos ((M_PI / 2048) * (8 * i + 1));
+	xsin1[i] = -sin ((M_PI / 2048) * (8 * i + 1));
+    }
+
+    /* More twiddle factors to turn IFFT into IMDCT */
+    for (i = 0; i < 64; i++) {
+	xcos2[i] = -cos ((M_PI / 1024) * (8 * i + 1));
+	xsin2[i] = -sin ((M_PI / 1024) * (8 * i + 1));
+    }
+
+    for (i = 0; i < 7; i++) {
+	j = 1 << i;
+	for (k = 0; k < j; k++) {
+	    w[i][k].real = cos (-M_PI * k / j);
+	    w[i][k].imag = sin (-M_PI * k / j);
+	}
+    }
+    imdct_512 = imdct_do_512;
+    imdct_256 = imdct_do_256;
+#endif
+}
diff --git a/libavcodec/libac3/parse.c b/libavcodec/libac3/parse.c
new file mode 100644
index 0000000000..c6eae00ba5
--- /dev/null
+++ b/libavcodec/libac3/parse.c
@@ -0,0 +1,684 @@
+/* 
+ *    parse.c
+ *
+ *	Copyright (C) Aaron Holtzman - May 1999
+ *
+ *  This file is part of ac3dec, a free Dolby AC-3 stream decoder.
+ *	
+ *  ac3dec is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2, or (at your option)
+ *  any later version.
+ *   
+ *  ac3dec is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *   
+ *  You should have received a copy of the GNU General Public License
+ *  along with GNU Make; see the file COPYING.  If not, write to
+ *  the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. 
+ *
+ *
+ */
+
+#include <inttypes.h>
+#include <string.h>
+
+#include "ac3.h"
+#include "ac3_internal.h"
+
+#include "bitstream.h"
+#include "tables.h"
+
+extern stream_samples_t samples;	// FIXME
+static float delay[6][256];
+
+void ac3_init (void)
+{
+    imdct_init ();
+}
+
+static uint8_t halfrate[12] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 2, 3};
+
+int ac3_syncinfo (uint8_t * buf, int * flags,
+		  int * sample_rate, int * bit_rate)
+{
+    static int rate[] = { 32,  40,  48,  56,  64,  80,  96, 112,
+			 128, 160, 192, 224, 256, 320, 384, 448,
+			 512, 576, 640};
+    static uint8_t lfeon[8] = {0x10, 0x10, 0x04, 0x04, 0x04, 0x01, 0x04, 0x01};
+    int frmsizecod;
+    int bitrate;
+    int half;
+    int acmod;
+
+    if ((buf[0] != 0x0b) || (buf[1] != 0x77))	// syncword
+	return 0;
+
+    if (buf[5] >= 0x60)		// bsid >= 12
+	return 0;
+    half = halfrate[buf[5] >> 3];
+
+    // acmod, dsurmod and lfeon
+    acmod = buf[6] >> 5;
+    *flags = (((buf[6] & 0xf8) == 0x50) ? AC3_DOLBY : acmod) |
+	      ((buf[6] & lfeon[acmod]) ? AC3_LFE : 0);
+
+    frmsizecod = buf[4] & 63;
+    if (frmsizecod >= 38)
+	return 0;
+    bitrate = rate [frmsizecod >> 1];
+    *bit_rate = (bitrate * 1000) >> half;
+
+    switch (buf[4] & 0xc0) {
+    case 0:	// 48 KHz
+	*sample_rate = 48000 >> half;
+	return 4 * bitrate;
+    case 0x40:
+	*sample_rate = 44100 >> half;
+	return 2 * (320 * bitrate / 147 + (frmsizecod & 1));
+    case 0x80:
+	*sample_rate = 32000 >> half;
+	return 6 * bitrate;
+    default:
+	return 0;
+    }
+}
+
+int ac3_frame (ac3_state_t * state, uint8_t * buf, int * flags, float * level,
+	       float bias)
+{
+    static float clev[4] = {LEVEL_3DB, LEVEL_45DB, LEVEL_6DB, LEVEL_45DB};
+    static float slev[4] = {LEVEL_3DB, LEVEL_6DB, 0, LEVEL_6DB};
+    int chaninfo;
+    int acmod;
+
+    state->fscod = buf[4] >> 6;
+    state->halfrate = halfrate[buf[5] >> 3];
+    state->acmod = acmod = buf[6] >> 5;
+
+    bitstream_set_ptr (buf + 6);
+    bitstream_get (3);	// skip acmod we already parsed
+
+    if ((acmod == 2) && (bitstream_get (2) == 2))	// dsurmod
+	acmod = AC3_DOLBY;
+
+    if ((acmod & 1) && (acmod != 1))
+	state->clev = clev[bitstream_get (2)];	// cmixlev
+
+    if (acmod & 4)
+	state->slev = slev[bitstream_get (2)];	// surmixlev
+
+    state->lfeon = bitstream_get (1);
+
+    state->output = downmix_init (acmod, *flags, level,
+				  state->clev, state->slev);
+    if (state->output < 0)
+	return 1;
+    *flags = state->output;
+    state->level = *level;
+    state->bias = bias;
+
+    chaninfo = !acmod;
+    do {
+	bitstream_get (5);	// dialnorm
+	if (bitstream_get (1))	// compre
+	    bitstream_get (8);	// compr
+	if (bitstream_get (1))	// langcode
+	    bitstream_get (8);	// langcod
+	if (bitstream_get (1))	// audprodie
+	    bitstream_get (7);	// mixlevel + roomtyp
+    } while (chaninfo--);
+
+    bitstream_get (2);		// copyrightb + origbs
+
+    if (bitstream_get (1))	// timecod1e
+	bitstream_get (14);	// timecod1
+    if (bitstream_get (1))	// timecod2e
+	bitstream_get (14);	// timecod2
+
+    if (bitstream_get (1)) {	// addbsie
+	int addbsil;
+
+	addbsil = bitstream_get (6);
+	do {
+	    bitstream_get (8);	// addbsi
+	} while (addbsil--);
+    }
+
+    return 0;
+}
+
+static int parse_exponents (int expstr, int ngrps, uint8_t exponent,
+			    uint8_t * dest)
+{
+    int exps;
+
+    while (ngrps--) {
+	exps = bitstream_get (7);
+
+	exponent += exp_1[exps];
+	if (exponent > 24)
+	    return 1;
+
+	switch (expstr) {
+	case EXP_D45:
+	    *(dest++) = exponent;
+	    *(dest++) = exponent;
+	case EXP_D25:
+	    *(dest++) = exponent;
+	case EXP_D15:
+	    *(dest++) = exponent;
+	}
+
+	exponent += exp_2[exps];
+	if (exponent > 24)
+	    return 1;
+
+	switch (expstr) {
+	case EXP_D45:
+	    *(dest++) = exponent;
+	    *(dest++) = exponent;
+	case EXP_D25:
+	    *(dest++) = exponent;
+	case EXP_D15:
+	    *(dest++) = exponent;
+	}
+
+	exponent += exp_3[exps];
+	if (exponent > 24)
+	    return 1;
+
+	switch (expstr) {
+	case EXP_D45:
+	    *(dest++) = exponent;
+	    *(dest++) = exponent;
+	case EXP_D25:
+	    *(dest++) = exponent;
+	case EXP_D15:
+	    *(dest++) = exponent;
+	}
+    }	
+
+    return 0;
+}
+
+static int parse_deltba (int8_t * deltba)
+{
+    int deltnseg, deltlen, delta, j;
+
+    memset (deltba, 0, 50);
+
+    deltnseg = bitstream_get (3);
+    j = 0;
+    do {
+	j += bitstream_get (5);
+	deltlen = bitstream_get (4);
+	delta = bitstream_get (3);
+	delta -= (delta >= 4) ? 3 : 4;
+	if (!deltlen)
+	    continue;
+	if (j + deltlen >= 50)
+	    return 1;
+	while (deltlen--)
+	    deltba[j++] = delta;
+    } while (deltnseg--);
+
+    return 0;
+}
+
+static inline int zero_snr_offsets (int nfchans, ac3_state_t * state)
+{
+    int i;
+
+    if ((state->csnroffst) || (state->cplinu && state->cplba.fsnroffst) ||
+	(state->lfeon && state->lfeba.fsnroffst))
+	return 0;
+    for (i = 0; i < nfchans; i++)
+	if (state->ba[i].fsnroffst)
+	    return 0;
+    return 1;
+}
+
+static float q_1[2];
+static float q_2[2];
+static float q_4;
+static int q_1_pointer;
+static int q_2_pointer;
+static int q_4_pointer;
+
+#define GET_COEFF(COEFF,DITHER)						\
+    switch (bap[i]) {							\
+    case 0:								\
+	DITHER (scale_factor[exp[i]]);					\
+									\
+    case -1:								\
+	if (q_1_pointer >= 0) {						\
+	    COEFF (q_1[q_1_pointer--] * scale_factor[exp[i]]);		\
+	} else {							\
+	    int code;							\
+									\
+	    code = bitstream_get (5);					\
+									\
+	    q_1_pointer = 1;						\
+	    q_1[0] = q_1_2[code];					\
+	    q_1[1] = q_1_1[code];					\
+	    COEFF (q_1_0[code] * scale_factor[exp[i]]);			\
+	}								\
+									\
+    case -2:								\
+	if (q_2_pointer >= 0) {						\
+	    COEFF (q_2[q_2_pointer--] * scale_factor[exp[i]]);		\
+	} else {							\
+	    int code;							\
+									\
+	    code = bitstream_get (7);					\
+									\
+	    q_2_pointer = 1;						\
+	    q_2[0] = q_2_2[code];					\
+	    q_2[1] = q_2_1[code];					\
+	    COEFF (q_2_0[code] * scale_factor[exp[i]]);			\
+	}								\
+									\
+    case 3:								\
+	COEFF (q_3[bitstream_get (3)] * scale_factor[exp[i]]);		\
+									\
+    case -3:								\
+	if (q_4_pointer == 0) {						\
+	    q_4_pointer = -1;						\
+	    COEFF (q_4 * scale_factor[exp[i]]);				\
+	} else {							\
+	    int code;							\
+									\
+	    code = bitstream_get (7);					\
+									\
+	    q_4_pointer = 0;						\
+	    q_4 = q_4_1[code];						\
+	    COEFF (q_4_0[code] * scale_factor[exp[i]]);			\
+	}								\
+									\
+    case 4:								\
+	COEFF (q_5[bitstream_get (4)] * scale_factor[exp[i]]);		\
+									\
+    default:								\
+	COEFF (((int16_t)(bitstream_get(bap[i]) << (16 - bap[i]))) *	\
+	       scale_factor[exp[i]]);					\
+    }
+
+#define CHANNEL_COEFF(val)			\
+    coeff[i++] = val;				\
+    continue;
+
+#define CHANNEL_DITHER(val)			\
+    if (dither) {				\
+	coeff[i++] = dither_gen () * val;	\
+	continue;				\
+    } else {					\
+	coeff[i++] = 0;				\
+	continue;				\
+    }
+
+static uint16_t lfsr_state = 1;
+
+static inline int16_t dither_gen(void)
+{
+    int16_t state;
+
+    state = dither_lut[lfsr_state >> 8] ^ (lfsr_state << 8);
+	
+    lfsr_state = (uint16_t) state;
+
+    return ((state * (int) (LEVEL_3DB * 256)) >> 8);
+}
+
+static void coeff_get (float * coeff, uint8_t * exp, int8_t * bap,
+		       int dither, int end)
+{
+    int i;
+
+    i = 0;
+    while (i < end)
+	GET_COEFF (CHANNEL_COEFF, CHANNEL_DITHER);
+}
+
+#define COUPLING_COEFF(val)	\
+    cplcoeff = val;		\
+    break;
+
+#define COUPLING_DITHER(val)						\
+    cplcoeff = val;							\
+    for (ch = 0; ch < nfchans; ch++)					\
+	if (state->chincpl[ch]) {					\
+	    if (dithflag[ch])						\
+		samples[ch][i] =					\
+		    state->cplco[ch][bnd] * dither_gen () * cplcoeff;	\
+	    else							\
+		samples[ch][i] = 0;					\
+	}								\
+    i++;								\
+    continue;
+
+int ac3_block (ac3_state_t * state)
+{
+    static const uint8_t nfchans_tbl[8] = {2, 1, 2, 3, 3, 4, 4, 5};
+    static int rematrix_band[4] = {25, 37, 61, 253};
+    int i, nfchans, chaninfo;
+    uint8_t cplexpstr, chexpstr[5], lfeexpstr, do_bit_alloc, done_cpl;
+    uint8_t blksw[5], dithflag[5];
+
+    nfchans = nfchans_tbl[state->acmod];
+
+    for (i = 0; i < nfchans; i++)
+	blksw[i] = bitstream_get (1);
+
+    for (i = 0; i < nfchans; i++)
+	dithflag[i] = bitstream_get (1);
+
+    chaninfo = !(state->acmod);
+    do {
+	if (bitstream_get (1))	// dynrnge
+	    bitstream_get (8);	// dynrng
+    } while (chaninfo--);
+
+    if (bitstream_get (1)) {	// cplstre
+	state->cplinu = bitstream_get (1);
+	if (state->cplinu) {
+	    static int bndtab[16] = {31, 35, 37, 39, 41, 42, 43, 44,
+				     45, 45, 46, 46, 47, 47, 48, 48};
+	    int cplbegf;
+	    int cplendf;
+	    int ncplsubnd;
+
+	    for (i = 0; i < nfchans; i++)
+		state->chincpl[i] = bitstream_get (1);
+	    switch (state->acmod) {
+	    case 0: case 1:
+		return 1;
+	    case 2:
+		state->phsflginu = bitstream_get (1);
+	    }
+	    cplbegf = bitstream_get (4);
+	    cplendf = bitstream_get (4);
+
+	    if (cplendf + 3 - cplbegf < 0)
+		return 1;
+	    state->ncplbnd = ncplsubnd = cplendf + 3 - cplbegf;
+	    state->cplstrtbnd = bndtab[cplbegf];
+	    state->cplstrtmant = cplbegf * 12 + 37;
+	    state->cplendmant = cplendf * 12 + 73;
+
+	    for (i = 0; i < ncplsubnd - 1; i++) {
+		state->cplbndstrc[i] = bitstream_get (1);
+		state->ncplbnd -= state->cplbndstrc[i];
+	    }
+	    state->cplbndstrc[i] = 0;	// last value is a sentinel
+	}
+    }
+
+    if (state->cplinu) {
+	int j, cplcoe;
+
+	cplcoe = 0;
+	for (i = 0; i < nfchans; i++)
+	    if (state->chincpl[i])
+		if (bitstream_get (1)) {	// cplcoe
+		    int mstrcplco, cplcoexp, cplcomant;
+
+		    cplcoe = 1;
+		    mstrcplco = 3 * bitstream_get (2);
+		    for (j = 0; j < state->ncplbnd; j++) {
+			cplcoexp = bitstream_get (4);
+			cplcomant = bitstream_get (4);
+			if (cplcoexp == 15)
+			    cplcomant <<= 14;
+			else
+			    cplcomant = (cplcomant | 0x10) << 13;
+			state->cplco[i][j] =
+			    cplcomant * scale_factor[cplcoexp + mstrcplco];
+		    }
+		}
+	if ((state->acmod == 2) && state->phsflginu && cplcoe)
+	    for (j = 0; j < state->ncplbnd; j++)
+		if (bitstream_get (1))	// phsflg
+		    state->cplco[1][j] = -state->cplco[1][j];
+    }
+
+    if ((state->acmod == 2) && (bitstream_get (1))) {	// rematstr
+	int end;
+
+	end = (state->cplinu) ? state->cplstrtmant : 253;
+	i = 0;
+	do
+	    state->rematflg[i] = bitstream_get (1);
+	while (rematrix_band[i++] < end);
+    }
+
+    cplexpstr = EXP_REUSE;
+    lfeexpstr = EXP_REUSE;
+    if (state->cplinu)
+	cplexpstr = bitstream_get (2);
+    for (i = 0; i < nfchans; i++)
+	chexpstr[i] = bitstream_get (2);
+    if (state->lfeon) 
+	lfeexpstr = bitstream_get (1);
+
+    for (i = 0; i < nfchans; i++)
+	if (chexpstr[i] != EXP_REUSE) {
+	    if (state->cplinu && state->chincpl[i])
+		state->endmant[i] = state->cplstrtmant;
+	    else {
+		int chbwcod;
+
+		chbwcod = bitstream_get (6);
+		if (chbwcod > 60)
+		    return 1;
+		state->endmant[i] = chbwcod * 3 + 73;
+	    }
+	}
+
+    do_bit_alloc = 0;
+
+    if (cplexpstr != EXP_REUSE) {
+	int cplabsexp, ncplgrps;
+
+	do_bit_alloc = 1;
+	ncplgrps = ((state->cplendmant - state->cplstrtmant) /
+		    (3 << (cplexpstr - 1)));
+	cplabsexp = bitstream_get (4) << 1;
+	if (parse_exponents (cplexpstr, ncplgrps, cplabsexp,
+			     state->cpl_exp + state->cplstrtmant))
+	    return 1;
+    }
+    for (i = 0; i < nfchans; i++)
+	if (chexpstr[i] != EXP_REUSE) {
+	    int grp_size, nchgrps;
+
+	    do_bit_alloc = 1;
+	    grp_size = 3 << (chexpstr[i] - 1);
+	    nchgrps = (state->endmant[i] + grp_size - 4) / grp_size;
+	    state->fbw_exp[i][0] = bitstream_get (4);
+	    if (parse_exponents (chexpstr[i], nchgrps, state->fbw_exp[i][0],
+				 state->fbw_exp[i] + 1))
+		return 1;
+	    bitstream_get (2);	// gainrng
+	}
+    if (lfeexpstr != EXP_REUSE) {
+	do_bit_alloc = 1;
+	state->lfe_exp[0] = bitstream_get (4);
+	if (parse_exponents (lfeexpstr, 2, state->lfe_exp[0],
+			     state->lfe_exp + 1))
+	    return 1;
+    }
+
+    if (bitstream_get (1)) {	// baie
+	do_bit_alloc = 1;
+	state->sdcycod = bitstream_get (2);
+	state->fdcycod = bitstream_get (2);
+	state->sgaincod = bitstream_get (2);
+	state->dbpbcod = bitstream_get (2);
+	state->floorcod = bitstream_get (3);
+    }
+    if (bitstream_get (1)) {	//snroffste
+	do_bit_alloc = 1;
+	state->csnroffst = bitstream_get (6);
+	if (state->cplinu) {
+	    state->cplba.fsnroffst = bitstream_get (4);
+	    state->cplba.fgaincod = bitstream_get (3);
+	}
+	for (i = 0; i < nfchans; i++) {
+	    state->ba[i].fsnroffst = bitstream_get (4);
+	    state->ba[i].fgaincod = bitstream_get (3);
+	}
+	if (state->lfeon) {
+	    state->lfeba.fsnroffst = bitstream_get (4);
+	    state->lfeba.fgaincod = bitstream_get (3);
+	}
+    }
+    if ((state->cplinu) && (bitstream_get (1))) {	// cplleake
+	do_bit_alloc = 1;
+	state->cplfleak = 2304 - (bitstream_get (3) << 8);
+	state->cplsleak = 2304 - (bitstream_get (3) << 8);
+    }
+
+    if (bitstream_get (1)) {	// deltbaie
+	do_bit_alloc = 1;
+	if (state->cplinu)
+	    state->cplba.deltbae = bitstream_get (2);
+	for (i = 0; i < nfchans; i++)
+	    state->ba[i].deltbae = bitstream_get (2);
+	if (state->cplinu && (state->cplba.deltbae == DELTA_BIT_NEW) &&
+	    parse_deltba (state->cplba.deltba))
+	    return 1;
+	for (i = 0; i < nfchans; i++)
+	    if ((state->ba[i].deltbae == DELTA_BIT_NEW) &&
+		parse_deltba (state->ba[i].deltba))
+		return 1;
+    }
+
+    if (do_bit_alloc) {
+	if (zero_snr_offsets (nfchans, state)) {
+	    memset (state->cpl_bap, 0, sizeof (state->cpl_bap));
+	    memset (state->fbw_bap, 0, sizeof (state->fbw_bap));
+	    memset (state->lfe_bap, 0, sizeof (state->lfe_bap));
+	} else {
+	    if (state->cplinu)
+		bit_allocate (state, &state->cplba, state->cplstrtbnd,
+			      state->cplstrtmant, state->cplendmant,
+			      state->cplfleak, state->cplsleak,
+			      state->cpl_exp, state->cpl_bap);
+	    for (i = 0; i < nfchans; i++)
+		bit_allocate (state, state->ba + i, 0, 0, state->endmant[i],
+			      0, 0, state->fbw_exp[i], state->fbw_bap[i]);
+	    if (state->lfeon) {
+		state->lfeba.deltbae = DELTA_BIT_NONE;
+		bit_allocate (state, &state->lfeba, 0, 0, 7, 0, 0,
+			      state->lfe_exp, state->lfe_bap);
+	    }
+	}
+    }
+
+    if (bitstream_get (1)) {	// skiple
+	i = bitstream_get (9);	// skipl
+	while (i--)
+	    bitstream_get (8);
+    }
+
+    q_1_pointer = q_2_pointer = q_4_pointer = -1;
+    done_cpl = 0;
+
+    for (i = 0; i < nfchans; i++) {
+	int j;
+
+	coeff_get (samples[i], state->fbw_exp[i], state->fbw_bap[i],
+		   dithflag[i], state->endmant[i]);
+
+	if (state->cplinu && state->chincpl[i]) {
+	    if (!done_cpl) {
+		int i, i_end, bnd, sub_bnd, ch;
+		float cplcoeff;
+
+		done_cpl = 1;
+
+#define bap state->cpl_bap
+#define exp state->cpl_exp
+
+		sub_bnd = bnd = 0;
+		i = state->cplstrtmant;
+		while (i < state->cplendmant) {
+		    i_end = i + 12;
+		    while (state->cplbndstrc[sub_bnd++])
+			i_end += 12;
+
+		    while (i < i_end) {
+			GET_COEFF (COUPLING_COEFF, COUPLING_DITHER);
+			for (ch = 0; ch < nfchans; ch++)
+			    if (state->chincpl[ch])
+				samples[ch][i] =
+				    state->cplco[ch][bnd] * cplcoeff;
+			i++;
+		    }
+		    bnd++;
+		}
+
+#undef bap
+#undef exp
+	    }
+	    j = state->cplendmant;
+	} else
+	    j = state->endmant[i];
+	for (; j < 256; j++)
+	    samples[i][j] = 0;
+    }
+
+    if (state->acmod == 2) {
+	int j, end, band;
+
+	end = ((state->endmant[0] < state->endmant[1]) ?
+	       state->endmant[0] : state->endmant[1]);
+
+	i = 0;
+	j = 13;
+	do {
+	    if (!state->rematflg[i]) {
+		j = rematrix_band[i++];
+		continue;
+	    }
+	    band = rematrix_band[i++];
+	    if (band > end)
+		band = end;
+	    do {
+		float tmp0, tmp1;
+
+		tmp0 = samples[0][j];
+		tmp1 = samples[1][j];
+		samples[0][j] = tmp0 + tmp1;
+		samples[1][j] = tmp0 - tmp1;
+	    } while (++j < band);
+	} while (j < end);
+    }
+
+    if (state->lfeon) {
+	coeff_get (samples[5], state->lfe_exp, state->lfe_bap, 0, 7);
+#if 0
+	for (i = 7; i < 256; i++)
+	    samples[5][i] = 0;
+#endif
+    }
+
+    for (i = 0; i < nfchans; i++)
+	if (blksw[i])
+            imdct_256 (samples[i], delay[i]);
+        else 
+            imdct_512 (samples[i], delay[i]);
+
+#if 0
+    if (state->lfeon)
+	imdct_512 (samples[5], delay[5]);
+#endif
+
+    downmix (*samples, state->acmod, state->output, state->level, state->bias,
+	     state->clev, state->slev);
+
+    return 0;
+}
diff --git a/libavcodec/libac3/tables.h b/libavcodec/libac3/tables.h
new file mode 100644
index 0000000000..5825f9326a
--- /dev/null
+++ b/libavcodec/libac3/tables.h
@@ -0,0 +1,250 @@
+/* 
+ *    parse.c
+ *
+ *	Copyright (C) Aaron Holtzman - May 1999
+ *
+ *  This file is part of ac3dec, a free Dolby AC-3 stream decoder.
+ *	
+ *  ac3dec is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2, or (at your option)
+ *  any later version.
+ *   
+ *  ac3dec is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *   
+ *  You should have received a copy of the GNU General Public License
+ *  along with GNU Make; see the file COPYING.  If not, write to
+ *  the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. 
+ *
+ *
+ */
+
+static int8_t exp_1[128] = {
+    -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
+    -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
+     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+     1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+     2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+    25,25,25
+};
+static int8_t exp_2[128] = {
+    -2,-2,-2,-2,-2,-1,-1,-1,-1,-1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2,
+    -2,-2,-2,-2,-2,-1,-1,-1,-1,-1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2,
+    -2,-2,-2,-2,-2,-1,-1,-1,-1,-1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2,
+    -2,-2,-2,-2,-2,-1,-1,-1,-1,-1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2,
+    -2,-2,-2,-2,-2,-1,-1,-1,-1,-1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2,
+    25,25,25
+};
+static int8_t exp_3[128] = {
+    -2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,
+    -2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,
+    -2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,
+    -2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,
+    -2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,
+    25,25,25
+};
+
+#define Q0 ((-2 << 15) / 3.0)
+#define Q1 (0)
+#define Q2 ((2 << 15) / 3.0)
+
+static const float q_1_0[ 32 ] = {
+    Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,
+    Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,
+    Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,
+    0,0,0,0,0
+};
+
+static const float q_1_1[ 32 ] = {
+    Q0,Q0,Q0,Q1,Q1,Q1,Q2,Q2,Q2,
+    Q0,Q0,Q0,Q1,Q1,Q1,Q2,Q2,Q2,
+    Q0,Q0,Q0,Q1,Q1,Q1,Q2,Q2,Q2,
+    0,0,0,0,0
+};
+
+static const float q_1_2[ 32 ] = {
+    Q0,Q1,Q2,Q0,Q1,Q2,Q0,Q1,Q2,
+    Q0,Q1,Q2,Q0,Q1,Q2,Q0,Q1,Q2,
+    Q0,Q1,Q2,Q0,Q1,Q2,Q0,Q1,Q2,
+    0,0,0,0,0
+};
+
+#undef Q0
+#undef Q1
+#undef Q2
+
+#define Q0 ((-4 << 15) / 5.0)
+#define Q1 ((-2 << 15) / 5.0)
+#define Q2 (0)
+#define Q3 ((2 << 15) / 5.0)
+#define Q4 ((4 << 15) / 5.0)
+
+static const float q_2_0[ 128 ] = {
+    Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,
+    Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,
+    Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,
+    Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,
+    Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,
+    0,0,0
+};
+
+static const float q_2_1[ 128 ] = {
+    Q0,Q0,Q0,Q0,Q0,Q1,Q1,Q1,Q1,Q1,Q2,Q2,Q2,Q2,Q2,Q3,Q3,Q3,Q3,Q3,Q4,Q4,Q4,Q4,Q4,
+    Q0,Q0,Q0,Q0,Q0,Q1,Q1,Q1,Q1,Q1,Q2,Q2,Q2,Q2,Q2,Q3,Q3,Q3,Q3,Q3,Q4,Q4,Q4,Q4,Q4,
+    Q0,Q0,Q0,Q0,Q0,Q1,Q1,Q1,Q1,Q1,Q2,Q2,Q2,Q2,Q2,Q3,Q3,Q3,Q3,Q3,Q4,Q4,Q4,Q4,Q4,
+    Q0,Q0,Q0,Q0,Q0,Q1,Q1,Q1,Q1,Q1,Q2,Q2,Q2,Q2,Q2,Q3,Q3,Q3,Q3,Q3,Q4,Q4,Q4,Q4,Q4,
+    Q0,Q0,Q0,Q0,Q0,Q1,Q1,Q1,Q1,Q1,Q2,Q2,Q2,Q2,Q2,Q3,Q3,Q3,Q3,Q3,Q4,Q4,Q4,Q4,Q4,
+    0,0,0
+};
+
+static const float q_2_2[ 128 ] = {
+    Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,
+    Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,
+    Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,
+    Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,
+    Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,
+    0,0,0
+};
+
+#undef Q0
+#undef Q1
+#undef Q2
+#undef Q3
+#undef Q4
+
+static const float q_3[8] = {
+    (-6 << 15)/7.0, (-4 << 15)/7.0, (-2 << 15)/7.0, 0,
+    ( 2 << 15)/7.0, ( 4 << 15)/7.0, ( 6 << 15)/7.0, 0
+};
+
+#define Q0 ((-10 << 15) / 11.0)
+#define Q1 ((-8 << 15) / 11.0)
+#define Q2 ((-6 << 15) / 11.0)
+#define Q3 ((-4 << 15) / 11.0)
+#define Q4 ((-2 << 15) / 11.0)
+#define Q5 (0)
+#define Q6 ((2 << 15) / 11.0)
+#define Q7 ((4 << 15) / 11.0)
+#define Q8 ((6 << 15) / 11.0)
+#define Q9 ((8 << 15) / 11.0)
+#define QA ((10 << 15) / 11.0)
+
+static const float q_4_0[ 128 ] = {
+    Q0, Q0, Q0, Q0, Q0, Q0, Q0, Q0, Q0, Q0, Q0,
+    Q1, Q1, Q1, Q1, Q1, Q1, Q1, Q1, Q1, Q1, Q1,
+    Q2, Q2, Q2, Q2, Q2, Q2, Q2, Q2, Q2, Q2, Q2,
+    Q3, Q3, Q3, Q3, Q3, Q3, Q3, Q3, Q3, Q3, Q3,
+    Q4, Q4, Q4, Q4, Q4, Q4, Q4, Q4, Q4, Q4, Q4,
+    Q5, Q5, Q5, Q5, Q5, Q5, Q5, Q5, Q5, Q5, Q5,
+    Q6, Q6, Q6, Q6, Q6, Q6, Q6, Q6, Q6, Q6, Q6,
+    Q7, Q7, Q7, Q7, Q7, Q7, Q7, Q7, Q7, Q7, Q7,
+    Q8, Q8, Q8, Q8, Q8, Q8, Q8, Q8, Q8, Q8, Q8,
+    Q9, Q9, Q9, Q9, Q9, Q9, Q9, Q9, Q9, Q9, Q9,
+    QA, QA, QA, QA, QA, QA, QA, QA, QA, QA, QA,
+    0,  0,  0,  0,  0,  0,  0
+};
+
+static const float q_4_1[ 128 ] = {
+    Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
+    Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
+    Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
+    Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
+    Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
+    Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
+    Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
+    Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
+    Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
+    Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
+    Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
+    0,  0,  0,  0,  0,  0,  0
+};
+
+#undef Q0
+#undef Q1
+#undef Q2
+#undef Q3
+#undef Q4
+#undef Q5
+#undef Q6
+#undef Q7
+#undef Q8
+#undef Q9
+#undef QA
+
+static const float q_5[16] = {
+    (-14 << 15)/15.0,(-12 << 15)/15.0,(-10 << 15)/15.0,
+    ( -8 << 15)/15.0,( -6 << 15)/15.0,( -4 << 15)/15.0,
+    ( -2 << 15)/15.0,   0            ,(  2 << 15)/15.0,
+    (  4 << 15)/15.0,(  6 << 15)/15.0,(  8 << 15)/15.0,
+    ( 10 << 15)/15.0,( 12 << 15)/15.0,( 14 << 15)/15.0,
+    0
+};
+
+static const uint32_t u32_scale_factors[25] = 
+{
+	0x38000000, //2 ^ -(0 + 15)
+	0x37800000, //2 ^ -(1 + 15)
+	0x37000000, //2 ^ -(2 + 15)
+	0x36800000, //2 ^ -(3 + 15)
+	0x36000000, //2 ^ -(4 + 15)
+	0x35800000, //2 ^ -(5 + 15)
+	0x35000000, //2 ^ -(6 + 15)
+	0x34800000, //2 ^ -(7 + 15)
+	0x34000000, //2 ^ -(8 + 15)
+	0x33800000, //2 ^ -(9 + 15)
+	0x33000000, //2 ^ -(10 + 15)
+	0x32800000, //2 ^ -(11 + 15)
+	0x32000000, //2 ^ -(12 + 15)
+	0x31800000, //2 ^ -(13 + 15)
+	0x31000000, //2 ^ -(14 + 15)
+	0x30800000, //2 ^ -(15 + 15)
+	0x30000000, //2 ^ -(16 + 15)
+	0x2f800000, //2 ^ -(17 + 15)
+	0x2f000000, //2 ^ -(18 + 15)
+	0x2e800000, //2 ^ -(19 + 15)
+	0x2e000000, //2 ^ -(20 + 15)
+	0x2d800000, //2 ^ -(21 + 15)
+	0x2d000000, //2 ^ -(22 + 15)
+	0x2c800000, //2 ^ -(23 + 15)
+	0x2c000000  //2 ^ -(24 + 15)
+};
+
+static float * scale_factor = (float *) u32_scale_factors;
+
+static const uint16_t dither_lut[256] = {
+    0x0000, 0xa011, 0xe033, 0x4022, 0x6077, 0xc066, 0x8044, 0x2055,
+    0xc0ee, 0x60ff, 0x20dd, 0x80cc, 0xa099, 0x0088, 0x40aa, 0xe0bb,
+    0x21cd, 0x81dc, 0xc1fe, 0x61ef, 0x41ba, 0xe1ab, 0xa189, 0x0198,
+    0xe123, 0x4132, 0x0110, 0xa101, 0x8154, 0x2145, 0x6167, 0xc176,
+    0x439a, 0xe38b, 0xa3a9, 0x03b8, 0x23ed, 0x83fc, 0xc3de, 0x63cf,
+    0x8374, 0x2365, 0x6347, 0xc356, 0xe303, 0x4312, 0x0330, 0xa321,
+    0x6257, 0xc246, 0x8264, 0x2275, 0x0220, 0xa231, 0xe213, 0x4202,
+    0xa2b9, 0x02a8, 0x428a, 0xe29b, 0xc2ce, 0x62df, 0x22fd, 0x82ec,
+    0x8734, 0x2725, 0x6707, 0xc716, 0xe743, 0x4752, 0x0770, 0xa761,
+    0x47da, 0xe7cb, 0xa7e9, 0x07f8, 0x27ad, 0x87bc, 0xc79e, 0x678f,
+    0xa6f9, 0x06e8, 0x46ca, 0xe6db, 0xc68e, 0x669f, 0x26bd, 0x86ac,
+    0x6617, 0xc606, 0x8624, 0x2635, 0x0660, 0xa671, 0xe653, 0x4642,
+    0xc4ae, 0x64bf, 0x249d, 0x848c, 0xa4d9, 0x04c8, 0x44ea, 0xe4fb,
+    0x0440, 0xa451, 0xe473, 0x4462, 0x6437, 0xc426, 0x8404, 0x2415,
+    0xe563, 0x4572, 0x0550, 0xa541, 0x8514, 0x2505, 0x6527, 0xc536,
+    0x258d, 0x859c, 0xc5be, 0x65af, 0x45fa, 0xe5eb, 0xa5c9, 0x05d8,
+    0xae79, 0x0e68, 0x4e4a, 0xee5b, 0xce0e, 0x6e1f, 0x2e3d, 0x8e2c,
+    0x6e97, 0xce86, 0x8ea4, 0x2eb5, 0x0ee0, 0xaef1, 0xeed3, 0x4ec2,
+    0x8fb4, 0x2fa5, 0x6f87, 0xcf96, 0xefc3, 0x4fd2, 0x0ff0, 0xafe1,
+    0x4f5a, 0xef4b, 0xaf69, 0x0f78, 0x2f2d, 0x8f3c, 0xcf1e, 0x6f0f,
+    0xede3, 0x4df2, 0x0dd0, 0xadc1, 0x8d94, 0x2d85, 0x6da7, 0xcdb6,
+    0x2d0d, 0x8d1c, 0xcd3e, 0x6d2f, 0x4d7a, 0xed6b, 0xad49, 0x0d58,
+    0xcc2e, 0x6c3f, 0x2c1d, 0x8c0c, 0xac59, 0x0c48, 0x4c6a, 0xec7b,
+    0x0cc0, 0xacd1, 0xecf3, 0x4ce2, 0x6cb7, 0xcca6, 0x8c84, 0x2c95,
+    0x294d, 0x895c, 0xc97e, 0x696f, 0x493a, 0xe92b, 0xa909, 0x0918,
+    0xe9a3, 0x49b2, 0x0990, 0xa981, 0x89d4, 0x29c5, 0x69e7, 0xc9f6,
+    0x0880, 0xa891, 0xe8b3, 0x48a2, 0x68f7, 0xc8e6, 0x88c4, 0x28d5,
+    0xc86e, 0x687f, 0x285d, 0x884c, 0xa819, 0x0808, 0x482a, 0xe83b,
+    0x6ad7, 0xcac6, 0x8ae4, 0x2af5, 0x0aa0, 0xaab1, 0xea93, 0x4a82,
+    0xaa39, 0x0a28, 0x4a0a, 0xea1b, 0xca4e, 0x6a5f, 0x2a7d, 0x8a6c,
+    0x4b1a, 0xeb0b, 0xab29, 0x0b38, 0x2b6d, 0x8b7c, 0xcb5e, 0x6b4f,
+    0x8bf4, 0x2be5, 0x6bc7, 0xcbd6, 0xeb83, 0x4b92, 0x0bb0, 0xaba1
+};
diff --git a/libavcodec/mjpegenc.c b/libavcodec/mjpegenc.c
new file mode 100644
index 0000000000..ead72605e6
--- /dev/null
+++ b/libavcodec/mjpegenc.c
@@ -0,0 +1,417 @@
+/*
+ * MJPEG encoder
+ * Copyright (c) 2000 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include "avcodec.h"
+#include "dsputil.h"
+#include "mpegvideo.h"
+
+typedef struct MJpegContext {
+    UINT8 huff_size_dc_luminance[12];
+    UINT16 huff_code_dc_luminance[12];
+    UINT8 huff_size_dc_chrominance[12];
+    UINT16 huff_code_dc_chrominance[12];
+
+    UINT8 huff_size_ac_luminance[256];
+    UINT16 huff_code_ac_luminance[256];
+    UINT8 huff_size_ac_chrominance[256];
+    UINT16 huff_code_ac_chrominance[256];
+} MJpegContext;
+
+#define SOF0 0xc0
+#define SOI 0xd8
+#define EOI 0xd9
+#define DQT 0xdb
+#define DHT 0xc4
+#define SOS 0xda
+
+#if 0
+/* These are the sample quantization tables given in JPEG spec section K.1.
+ * The spec says that the values given produce "good" quality, and
+ * when divided by 2, "very good" quality.
+ */
+static const unsigned char std_luminance_quant_tbl[64] = {
+    16,  11,  10,  16,  24,  40,  51,  61,
+    12,  12,  14,  19,  26,  58,  60,  55,
+    14,  13,  16,  24,  40,  57,  69,  56,
+    14,  17,  22,  29,  51,  87,  80,  62,
+    18,  22,  37,  56,  68, 109, 103,  77,
+    24,  35,  55,  64,  81, 104, 113,  92,
+    49,  64,  78,  87, 103, 121, 120, 101,
+    72,  92,  95,  98, 112, 100, 103,  99
+};
+static const unsigned char std_chrominance_quant_tbl[64] = {
+    17,  18,  24,  47,  99,  99,  99,  99,
+    18,  21,  26,  66,  99,  99,  99,  99,
+    24,  26,  56,  99,  99,  99,  99,  99,
+    47,  66,  99,  99,  99,  99,  99,  99,
+    99,  99,  99,  99,  99,  99,  99,  99,
+    99,  99,  99,  99,  99,  99,  99,  99,
+    99,  99,  99,  99,  99,  99,  99,  99,
+    99,  99,  99,  99,  99,  99,  99,  99
+};
+#endif
+
+/* Set up the standard Huffman tables (cf. JPEG standard section K.3) */
+/* IMPORTANT: these are only valid for 8-bit data precision! */
+static const UINT8 bits_dc_luminance[17] =
+{ /* 0-base */ 0, 0, 1, 5, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0 };
+static const UINT8 val_dc_luminance[] =
+{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
+
+static const UINT8 bits_dc_chrominance[17] =
+{ /* 0-base */ 0, 0, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0 };
+static const UINT8 val_dc_chrominance[] =
+{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
+
+static const UINT8 bits_ac_luminance[17] =
+{ /* 0-base */ 0, 0, 2, 1, 3, 3, 2, 4, 3, 5, 5, 4, 4, 0, 0, 1, 0x7d };
+static const UINT8 val_ac_luminance[] =
+{ 0x01, 0x02, 0x03, 0x00, 0x04, 0x11, 0x05, 0x12,
+  0x21, 0x31, 0x41, 0x06, 0x13, 0x51, 0x61, 0x07,
+  0x22, 0x71, 0x14, 0x32, 0x81, 0x91, 0xa1, 0x08,
+  0x23, 0x42, 0xb1, 0xc1, 0x15, 0x52, 0xd1, 0xf0,
+  0x24, 0x33, 0x62, 0x72, 0x82, 0x09, 0x0a, 0x16,
+  0x17, 0x18, 0x19, 0x1a, 0x25, 0x26, 0x27, 0x28,
+  0x29, 0x2a, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39,
+  0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49,
+  0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59,
+  0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69,
+  0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79,
+  0x7a, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89,
+  0x8a, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98,
+  0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7,
+  0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6,
+  0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3, 0xc4, 0xc5,
+  0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2, 0xd3, 0xd4,
+  0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xe1, 0xe2,
+  0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea,
+  0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
+  0xf9, 0xfa 
+};
+
+static const UINT8 bits_ac_chrominance[17] =
+{ /* 0-base */ 0, 0, 2, 1, 2, 4, 4, 3, 4, 7, 5, 4, 4, 0, 1, 2, 0x77 };
+
+static const UINT8 val_ac_chrominance[] =
+{ 0x00, 0x01, 0x02, 0x03, 0x11, 0x04, 0x05, 0x21,
+  0x31, 0x06, 0x12, 0x41, 0x51, 0x07, 0x61, 0x71,
+  0x13, 0x22, 0x32, 0x81, 0x08, 0x14, 0x42, 0x91,
+  0xa1, 0xb1, 0xc1, 0x09, 0x23, 0x33, 0x52, 0xf0,
+  0x15, 0x62, 0x72, 0xd1, 0x0a, 0x16, 0x24, 0x34,
+  0xe1, 0x25, 0xf1, 0x17, 0x18, 0x19, 0x1a, 0x26,
+  0x27, 0x28, 0x29, 0x2a, 0x35, 0x36, 0x37, 0x38,
+  0x39, 0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48,
+  0x49, 0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58,
+  0x59, 0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68,
+  0x69, 0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78,
+  0x79, 0x7a, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
+  0x88, 0x89, 0x8a, 0x92, 0x93, 0x94, 0x95, 0x96,
+  0x97, 0x98, 0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5,
+  0xa6, 0xa7, 0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4,
+  0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3,
+  0xc4, 0xc5, 0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2,
+  0xd3, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda,
+  0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9,
+  0xea, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
+  0xf9, 0xfa 
+};
+
+
+/* isn't this function nicer than the one in the libjpeg ? */
+static void build_huffman_codes(UINT8 *huff_size, UINT16 *huff_code,
+                                const UINT8 *bits_table, const UINT8 *val_table)
+{
+    int i, j, k,nb, code, sym;
+
+    code = 0;
+    k = 0;
+    for(i=1;i<=16;i++) {
+        nb = bits_table[i];
+        for(j=0;j<nb;j++) {
+            sym = val_table[k++];
+            huff_size[sym] = i;
+            huff_code[sym] = code;
+            code++;
+        }
+        code <<= 1;
+    }
+}
+
+int mjpeg_init(MpegEncContext *s)
+{
+    MJpegContext *m;
+    
+    m = malloc(sizeof(MJpegContext));
+    if (!m)
+        return -1;
+
+    /* build all the huffman tables */
+    build_huffman_codes(m->huff_size_dc_luminance,
+                        m->huff_code_dc_luminance,
+                        bits_dc_luminance,
+                        val_dc_luminance);
+    build_huffman_codes(m->huff_size_dc_chrominance,
+                        m->huff_code_dc_chrominance,
+                        bits_dc_chrominance,
+                        val_dc_chrominance);
+    build_huffman_codes(m->huff_size_ac_luminance,
+                        m->huff_code_ac_luminance,
+                        bits_ac_luminance,
+                        val_ac_luminance);
+    build_huffman_codes(m->huff_size_ac_chrominance,
+                        m->huff_code_ac_chrominance,
+                        bits_ac_chrominance,
+                        val_ac_chrominance);
+    
+    s->mjpeg_ctx = m;
+    return 0;
+}
+
+void mjpeg_close(MpegEncContext *s)
+{
+    free(s->mjpeg_ctx);
+}
+
+static inline void put_marker(PutBitContext *p, int code)
+{
+    put_bits(p, 8, 0xff);
+    put_bits(p, 8, code);
+}
+
+/* table_class: 0 = DC coef, 1 = AC coefs */
+static int put_huffman_table(MpegEncContext *s, int table_class, int table_id,
+                             const UINT8 *bits_table, const UINT8 *value_table)
+{
+    PutBitContext *p = &s->pb;
+    int n, i;
+
+    put_bits(p, 4, table_class);
+    put_bits(p, 4, table_id);
+
+    n = 0;
+    for(i=1;i<=16;i++) {
+        n += bits_table[i];
+        put_bits(p, 8, bits_table[i]);
+    }
+
+    for(i=0;i<n;i++)
+        put_bits(p, 8, value_table[i]);
+
+    return n + 17;
+}
+
+static void jpeg_table_header(MpegEncContext *s)
+{
+    PutBitContext *p = &s->pb;
+    int i, size;
+    UINT8 *ptr;
+
+    /* quant matrixes */
+    put_marker(p, DQT);
+    put_bits(p, 16, 2 + 1 * (1 + 64));
+    put_bits(p, 4, 0); /* 8 bit precision */
+    put_bits(p, 4, 0); /* table 0 */
+    for(i=0;i<64;i++) {
+        put_bits(p, 8, s->intra_matrix[i]);
+    }
+#if 0
+    put_bits(p, 4, 0); /* 8 bit precision */
+    put_bits(p, 4, 1); /* table 1 */
+    for(i=0;i<64;i++) {
+        put_bits(p, 8, s->chroma_intra_matrix[i]);
+    }
+#endif
+
+    /* huffman table */
+    put_marker(p, DHT);
+    flush_put_bits(p);
+    ptr = p->buf_ptr;
+    put_bits(p, 16, 0); /* patched later */
+    size = 2;
+    size += put_huffman_table(s, 0, 0, bits_dc_luminance, val_dc_luminance);
+    size += put_huffman_table(s, 0, 1, bits_dc_chrominance, val_dc_chrominance);
+    
+    size += put_huffman_table(s, 1, 0, bits_ac_luminance, val_ac_luminance);
+    size += put_huffman_table(s, 1, 1, bits_ac_chrominance, val_ac_chrominance);
+    ptr[0] = size >> 8;
+    ptr[1] = size;
+}
+
+void mjpeg_picture_header(MpegEncContext *s)
+{
+    put_marker(&s->pb, SOI);
+
+    jpeg_table_header(s);
+
+    put_marker(&s->pb, SOF0);
+
+    put_bits(&s->pb, 16, 17);
+    put_bits(&s->pb, 8, 8); /* 8 bits/component */
+    put_bits(&s->pb, 16, s->height);
+    put_bits(&s->pb, 16, s->width);
+    put_bits(&s->pb, 8, 3); /* 3 components */
+    
+    /* Y component */
+    put_bits(&s->pb, 8, 1); /* component number */
+    put_bits(&s->pb, 4, 2); /* H factor */
+    put_bits(&s->pb, 4, 2); /* V factor */
+    put_bits(&s->pb, 8, 0); /* select matrix */
+    
+    /* Cb component */
+    put_bits(&s->pb, 8, 2); /* component number */
+    put_bits(&s->pb, 4, 1); /* H factor */
+    put_bits(&s->pb, 4, 1); /* V factor */
+    put_bits(&s->pb, 8, 0); /* select matrix */
+
+    /* Cr component */
+    put_bits(&s->pb, 8, 3); /* component number */
+    put_bits(&s->pb, 4, 1); /* H factor */
+    put_bits(&s->pb, 4, 1); /* V factor */
+    put_bits(&s->pb, 8, 0); /* select matrix */
+
+    /* scan header */
+    put_marker(&s->pb, SOS);
+    put_bits(&s->pb, 16, 12); /* length */
+    put_bits(&s->pb, 8, 3); /* 3 components */
+    
+    /* Y component */
+    put_bits(&s->pb, 8, 1); /* index */
+    put_bits(&s->pb, 4, 0); /* DC huffman table index */
+    put_bits(&s->pb, 4, 0); /* AC huffman table index */
+    
+    /* Cb component */
+    put_bits(&s->pb, 8, 2); /* index */
+    put_bits(&s->pb, 4, 1); /* DC huffman table index */
+    put_bits(&s->pb, 4, 1); /* AC huffman table index */
+    
+    /* Cr component */
+    put_bits(&s->pb, 8, 3); /* index */
+    put_bits(&s->pb, 4, 1); /* DC huffman table index */
+    put_bits(&s->pb, 4, 1); /* AC huffman table index */
+
+    put_bits(&s->pb, 8, 0); /* Ss (not used) */
+    put_bits(&s->pb, 8, 63); /* Se (not used) */
+    put_bits(&s->pb, 8, 0); /* (not used) */
+}
+
+void mjpeg_picture_trailer(MpegEncContext *s)
+{
+    jflush_put_bits(&s->pb);
+    put_marker(&s->pb, EOI);
+}
+
+static inline void encode_dc(MpegEncContext *s, int val, 
+                             UINT8 *huff_size, UINT16 *huff_code)
+{
+    int mant, nbits;
+
+    if (val == 0) {
+        jput_bits(&s->pb, huff_size[0], huff_code[0]);
+    } else {
+        mant = val;
+        if (val < 0) {
+            val = -val;
+            mant--;
+        }
+        
+        /* compute the log (XXX: optimize) */
+        nbits = 0;
+        while (val != 0) {
+            val = val >> 1;
+            nbits++;
+        }
+            
+        jput_bits(&s->pb, huff_size[nbits], huff_code[nbits]);
+        
+        jput_bits(&s->pb, nbits, mant & ((1 << nbits) - 1));
+    }
+}
+
+static void encode_block(MpegEncContext *s, DCTELEM *block, int n)
+{
+    int mant, nbits, code, i, j;
+    int component, dc, run, last_index, val;
+    MJpegContext *m = s->mjpeg_ctx;
+    UINT8 *huff_size_ac;
+    UINT16 *huff_code_ac;
+    
+    /* DC coef */
+    component = (n <= 3 ? 0 : n - 4 + 1);
+    dc = block[0]; /* overflow is impossible */
+    val = dc - s->last_dc[component];
+    if (n < 4) {
+        encode_dc(s, val, m->huff_size_dc_luminance, m->huff_code_dc_luminance);
+        huff_size_ac = m->huff_size_ac_luminance;
+        huff_code_ac = m->huff_code_ac_luminance;
+    } else {
+        encode_dc(s, val, m->huff_size_dc_chrominance, m->huff_code_dc_chrominance);
+        huff_size_ac = m->huff_size_ac_chrominance;
+        huff_code_ac = m->huff_code_ac_chrominance;
+    }
+    s->last_dc[component] = dc;
+    
+    /* AC coefs */
+    
+    run = 0;
+    last_index = s->block_last_index[n];
+    for(i=1;i<=last_index;i++) {
+        j = zigzag_direct[i];
+        val = block[j];
+        if (val == 0) {
+            run++;
+        } else {
+            while (run >= 16) {
+                jput_bits(&s->pb, huff_size_ac[0xf0], huff_code_ac[0xf0]);
+                run -= 16;
+            }
+            mant = val;
+            if (val < 0) {
+                val = -val;
+                mant--;
+            }
+            
+            /* compute the log (XXX: optimize) */
+            nbits = 0;
+            while (val != 0) {
+                val = val >> 1;
+                nbits++;
+            }
+            code = (run << 4) | nbits;
+
+            jput_bits(&s->pb, huff_size_ac[code], huff_code_ac[code]);
+        
+            jput_bits(&s->pb, nbits, mant & ((1 << nbits) - 1));
+            run = 0;
+        }
+    }
+
+    /* output EOB only if not already 64 values */
+    if (last_index < 63 || run != 0)
+        jput_bits(&s->pb, huff_size_ac[0], huff_code_ac[0]);
+}
+
+void mjpeg_encode_mb(MpegEncContext *s, 
+                     DCTELEM block[6][64])
+{
+    int i;
+    for(i=0;i<6;i++) {
+        encode_block(s, block[i], i);
+    }
+}
diff --git a/libavcodec/motion_est.c b/libavcodec/motion_est.c
new file mode 100644
index 0000000000..09fe662613
--- /dev/null
+++ b/libavcodec/motion_est.c
@@ -0,0 +1,517 @@
+/*
+ * Motion estimation 
+ * Copyright (c) 2000,2001 Gerard Lantau.
+ * 
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include "avcodec.h"
+#include "dsputil.h"
+#include "mpegvideo.h"
+
+static void halfpel_motion_search(MpegEncContext * s,
+				  int *mx_ptr, int *my_ptr, int dmin,
+				  int xmin, int ymin, int xmax, int ymax);
+
+/* config it to test motion vector encoding (send random vectors) */
+//#define CONFIG_TEST_MV_ENCODE
+
+static int pix_sum(UINT8 * pix, int line_size)
+{
+    int s, i, j;
+
+    s = 0;
+    for (i = 0; i < 16; i++) {
+	for (j = 0; j < 16; j += 8) {
+	    s += pix[0];
+	    s += pix[1];
+	    s += pix[2];
+	    s += pix[3];
+	    s += pix[4];
+	    s += pix[5];
+	    s += pix[6];
+	    s += pix[7];
+	    pix += 8;
+	}
+	pix += line_size - 16;
+    }
+    return s;
+}
+
+static int pix_norm1(UINT8 * pix, int line_size)
+{
+    int s, i, j;
+    UINT32 *sq = squareTbl + 256;
+
+    s = 0;
+    for (i = 0; i < 16; i++) {
+	for (j = 0; j < 16; j += 8) {
+	    s += sq[pix[0]];
+	    s += sq[pix[1]];
+	    s += sq[pix[2]];
+	    s += sq[pix[3]];
+	    s += sq[pix[4]];
+	    s += sq[pix[5]];
+	    s += sq[pix[6]];
+	    s += sq[pix[7]];
+	    pix += 8;
+	}
+	pix += line_size - 16;
+    }
+    return s;
+}
+
+static int pix_norm(UINT8 * pix1, UINT8 * pix2, int line_size)
+{
+    int s, i, j;
+    UINT32 *sq = squareTbl + 256;
+
+    s = 0;
+    for (i = 0; i < 16; i++) {
+	for (j = 0; j < 16; j += 8) {
+	    s += sq[pix1[0] - pix2[0]];
+	    s += sq[pix1[1] - pix2[1]];
+	    s += sq[pix1[2] - pix2[2]];
+	    s += sq[pix1[3] - pix2[3]];
+	    s += sq[pix1[4] - pix2[4]];
+	    s += sq[pix1[5] - pix2[5]];
+	    s += sq[pix1[6] - pix2[6]];
+	    s += sq[pix1[7] - pix2[7]];
+	    pix1 += 8;
+	    pix2 += 8;
+	}
+	pix1 += line_size - 16;
+	pix2 += line_size - 16;
+    }
+    return s;
+}
+
+static void no_motion_search(MpegEncContext * s,
+			     int *mx_ptr, int *my_ptr)
+{
+    *mx_ptr = 16 * s->mb_x;
+    *my_ptr = 16 * s->mb_y;
+}
+
+static int full_motion_search(MpegEncContext * s,
+                              int *mx_ptr, int *my_ptr, int range,
+                              int xmin, int ymin, int xmax, int ymax)
+{
+    int x1, y1, x2, y2, xx, yy, x, y;
+    int mx, my, dmin, d;
+    UINT8 *pix;
+
+    xx = 16 * s->mb_x;
+    yy = 16 * s->mb_y;
+    x1 = xx - range + 1;	/* we loose one pixel to avoid boundary pb with half pixel pred */
+    if (x1 < xmin)
+	x1 = xmin;
+    x2 = xx + range - 1;
+    if (x2 > xmax)
+	x2 = xmax;
+    y1 = yy - range + 1;
+    if (y1 < ymin)
+	y1 = ymin;
+    y2 = yy + range - 1;
+    if (y2 > ymax)
+	y2 = ymax;
+    pix = s->new_picture[0] + (yy * s->linesize) + xx;
+    dmin = 0x7fffffff;
+    mx = 0;
+    my = 0;
+    for (y = y1; y <= y2; y++) {
+	for (x = x1; x <= x2; x++) {
+	    d = pix_abs16x16(pix, s->last_picture[0] + (y * s->linesize) + x,
+			     s->linesize, 16);
+	    if (d < dmin ||
+		(d == dmin &&
+		 (abs(x - xx) + abs(y - yy)) <
+		 (abs(mx - xx) + abs(my - yy)))) {
+		dmin = d;
+		mx = x;
+		my = y;
+	    }
+	}
+    }
+
+    *mx_ptr = mx;
+    *my_ptr = my;
+
+#if 0
+    if (*mx_ptr < -(2 * range) || *mx_ptr >= (2 * range) ||
+	*my_ptr < -(2 * range) || *my_ptr >= (2 * range)) {
+	fprintf(stderr, "error %d %d\n", *mx_ptr, *my_ptr);
+    }
+#endif
+    return dmin;
+}
+
+
+static int log_motion_search(MpegEncContext * s,
+                             int *mx_ptr, int *my_ptr, int range,
+                             int xmin, int ymin, int xmax, int ymax)
+{
+    int x1, y1, x2, y2, xx, yy, x, y;
+    int mx, my, dmin, d;
+    UINT8 *pix;
+
+    xx = s->mb_x << 4;
+    yy = s->mb_y << 4;
+
+    /* Left limit */
+    x1 = xx - range;
+    if (x1 < xmin)
+	x1 = xmin;
+
+    /* Right limit */
+    x2 = xx + range;
+    if (x2 > xmax)
+	x2 = xmax;
+
+    /* Upper limit */
+    y1 = yy - range;
+    if (y1 < ymin)
+	y1 = ymin;
+
+    /* Lower limit */
+    y2 = yy + range;
+    if (y2 > ymax)
+	y2 = ymax;
+
+    pix = s->new_picture[0] + (yy * s->linesize) + xx;
+    dmin = 0x7fffffff;
+    mx = 0;
+    my = 0;
+
+    do {
+	for (y = y1; y <= y2; y += range) {
+	    for (x = x1; x <= x2; x += range) {
+		d = pix_abs16x16(pix, s->last_picture[0] + (y * s->linesize) + x, s->linesize, 16);
+		if (d < dmin || (d == dmin && (abs(x - xx) + abs(y - yy)) < (abs(mx - xx) + abs(my - yy)))) {
+		    dmin = d;
+		    mx = x;
+		    my = y;
+		}
+	    }
+	}
+
+	range = range >> 1;
+
+	x1 = mx - range;
+	if (x1 < xmin)
+	    x1 = xmin;
+
+	x2 = mx + range;
+	if (x2 > xmax)
+	    x2 = xmax;
+
+	y1 = my - range;
+	if (y1 < ymin)
+	    y1 = ymin;
+
+	y2 = my + range;
+	if (y2 > ymax)
+	    y2 = ymax;
+
+    } while (range >= 1);
+
+#ifdef DEBUG
+    fprintf(stderr, "log       - MX: %d\tMY: %d\n", mx, my);
+#endif
+    *mx_ptr = mx;
+    *my_ptr = my;
+    return dmin;
+}
+
+static int phods_motion_search(MpegEncContext * s,
+                               int *mx_ptr, int *my_ptr, int range,
+                               int xmin, int ymin, int xmax, int ymax)
+{
+    int x1, y1, x2, y2, xx, yy, x, y, lastx, d;
+    int mx, my, dminx, dminy;
+    UINT8 *pix;
+
+    xx = s->mb_x << 4;
+    yy = s->mb_y << 4;
+
+    /* Left limit */
+    x1 = xx - range;
+    if (x1 < xmin)
+	x1 = xmin;
+
+    /* Right limit */
+    x2 = xx + range;
+    if (x2 > xmax)
+	x2 = xmax;
+
+    /* Upper limit */
+    y1 = yy - range;
+    if (y1 < ymin)
+	y1 = ymin;
+
+    /* Lower limit */
+    y2 = yy + range;
+    if (y2 > ymax)
+	y2 = ymax;
+
+    pix = s->new_picture[0] + (yy * s->linesize) + xx;
+    mx = 0;
+    my = 0;
+
+    x = xx;
+    y = yy;
+    do {
+        dminx = 0x7fffffff;
+        dminy = 0x7fffffff;
+
+	lastx = x;
+	for (x = x1; x <= x2; x += range) {
+	    d = pix_abs16x16(pix, s->last_picture[0] + (y * s->linesize) + x, s->linesize, 16);
+	    if (d < dminx || (d == dminx && (abs(x - xx) + abs(y - yy)) < (abs(mx - xx) + abs(my - yy)))) {
+		dminx = d;
+		mx = x;
+	    }
+	}
+
+	x = lastx;
+	for (y = y1; y <= y2; y += range) {
+	    d = pix_abs16x16(pix, s->last_picture[0] + (y * s->linesize) + x, s->linesize, 16);
+	    if (d < dminy || (d == dminy && (abs(x - xx) + abs(y - yy)) < (abs(mx - xx) + abs(my - yy)))) {
+		dminy = d;
+		my = y;
+	    }
+	}
+
+	range = range >> 1;
+
+	x = mx;
+	y = my;
+	x1 = mx - range;
+	if (x1 < xmin)
+	    x1 = xmin;
+
+	x2 = mx + range;
+	if (x2 > xmax)
+	    x2 = xmax;
+
+	y1 = my - range;
+	if (y1 < ymin)
+	    y1 = ymin;
+
+	y2 = my + range;
+	if (y2 > ymax)
+	    y2 = ymax;
+
+    } while (range >= 1);
+
+#ifdef DEBUG
+    fprintf(stderr, "phods     - MX: %d\tMY: %d\n", mx, my);
+#endif
+
+    /* half pixel search */
+    *mx_ptr = mx;
+    *my_ptr = my;
+    return dminy;
+}
+
+/* The idea would be to make half pel ME after Inter/Intra decision to 
+   save time. */
+static void halfpel_motion_search(MpegEncContext * s,
+				  int *mx_ptr, int *my_ptr, int dmin,
+				  int xmin, int ymin, int xmax, int ymax)
+{
+    int mx, my, mx1, my1, d, xx, yy, dminh;
+    UINT8 *pix;
+
+    mx = *mx_ptr << 1;
+    my = *my_ptr << 1;
+
+    xx = 16 * s->mb_x;
+    yy = 16 * s->mb_y;
+
+    dminh = dmin;
+
+    /* Half pixel search */
+    mx1 = mx;
+    my1 = my;
+
+    pix = s->new_picture[0] + (yy * s->linesize) + xx;
+
+    if ((mx > (xmin << 1)) && mx < (xmax << 1) && 
+        (my > (ymin << 1)) && my < (ymax << 1)) {
+	    int dx, dy, px, py;
+	    UINT8 *ptr;
+        for (dy = -1; dy <= 1; dy++) {
+            for (dx = -1; dx <= 1; dx++) {
+                if (dx != 0 || dy != 0) {
+                    px = mx1 + dx;
+                    py = my1 + dy;
+                    ptr = s->last_picture[0] + ((py >> 1) * s->linesize) + (px >> 1);
+                    switch (((py & 1) << 1) | (px & 1)) {
+                    default:
+                    case 0:
+                        d = pix_abs16x16(pix, ptr, s->linesize, 16);
+                        break;
+                    case 1:
+                        d = pix_abs16x16_x2(pix, ptr, s->linesize, 16);
+                        break;
+                    case 2:
+                        d = pix_abs16x16_y2(pix, ptr, s->linesize, 16);
+                        break;
+                    case 3:
+                        d = pix_abs16x16_xy2(pix, ptr, s->linesize, 16);
+                        break;
+                    }
+                    if (d < dminh) {
+                        dminh = d;
+                        mx = px;
+                        my = py;
+                    }
+                }
+            }
+        }
+    }
+
+    *mx_ptr = mx - (xx << 1);
+    *my_ptr = my - (yy << 1);
+    //fprintf(stderr,"half  - MX: %d\tMY: %d\n",*mx_ptr ,*my_ptr);
+}
+
+#ifndef CONFIG_TEST_MV_ENCODE
+
+int estimate_motion(MpegEncContext * s,
+		    int mb_x, int mb_y,
+		    int *mx_ptr, int *my_ptr)
+{
+    UINT8 *pix, *ppix;
+    int sum, varc, vard, mx, my, range, dmin, xx, yy;
+    int xmin, ymin, xmax, ymax;
+
+    range = 8 * (1 << (s->f_code - 1));
+    /* XXX: temporary kludge to avoid overflow for msmpeg4 */
+    if (s->out_format == FMT_H263 && !s->h263_msmpeg4)
+	range = range * 2;
+
+    if (s->unrestricted_mv) {
+        xmin = -16;
+        ymin = -16;
+        xmax = s->width;
+        ymax = s->height;
+    } else {
+        xmin = 0;
+        ymin = 0;
+        xmax = s->width - 16;
+        ymax = s->height - 16;
+    }
+
+    switch(s->full_search) {
+    case ME_ZERO:
+    default:
+	no_motion_search(s, &mx, &my);
+        dmin = 0;
+        break;
+    case ME_FULL:
+	dmin = full_motion_search(s, &mx, &my, range, xmin, ymin, xmax, ymax);
+        break;
+    case ME_LOG:
+	dmin = log_motion_search(s, &mx, &my, range / 2, xmin, ymin, xmax, ymax);
+        break;
+    case ME_PHODS:
+	dmin = phods_motion_search(s, &mx, &my, range / 2, xmin, ymin, xmax, ymax);
+        break;
+    }
+#ifdef CONFIG_MMX
+    if (mm_flags & MM_MMX)
+        emms();
+#endif
+
+    /* intra / predictive decision */
+    xx = mb_x * 16;
+    yy = mb_y * 16;
+
+    pix = s->new_picture[0] + (yy * s->linesize) + xx;
+    /* At this point (mx,my) are full-pell and the absolute displacement */
+    ppix = s->last_picture[0] + (my * s->linesize) + mx;
+
+    sum = pix_sum(pix, s->linesize);
+    varc = pix_norm1(pix, s->linesize);
+    vard = pix_norm(pix, ppix, s->linesize);
+
+    vard = vard >> 8;
+    sum = sum >> 8;
+    varc = (varc >> 8) - (sum * sum);
+#if 0
+    printf("varc=%d (sum=%d) vard=%d mx=%d my=%d\n",
+	   varc, sum, vard, mx - xx, my - yy);
+#endif
+    if (vard <= 64 || vard < varc) {
+        if (s->full_search != ME_ZERO) {
+            halfpel_motion_search(s, &mx, &my, dmin, xmin, ymin, xmax, ymax);
+        } else {
+            mx -= 16 * s->mb_x;
+            my -= 16 * s->mb_y;
+        }
+	*mx_ptr = mx;
+	*my_ptr = my;
+	return 0;
+    } else {
+	*mx_ptr = 0;
+	*my_ptr = 0;
+	return 1;
+    }
+}
+
+#else
+
+/* test version which generates valid random vectors */
+int estimate_motion(MpegEncContext * s,
+		    int mb_x, int mb_y,
+		    int *mx_ptr, int *my_ptr)
+{
+    int xx, yy, x1, y1, x2, y2, range;
+
+    if ((random() % 10) >= 5) {
+	range = 8 * (1 << (s->f_code - 1));
+	if (s->out_format == FMT_H263 && !s->h263_msmpeg4)
+	    range = range * 2;
+
+	xx = 16 * s->mb_x;
+	yy = 16 * s->mb_y;
+	x1 = xx - range;
+	if (x1 < 0)
+	    x1 = 0;
+	x2 = xx + range - 1;
+	if (x2 > (s->width - 16))
+	    x2 = s->width - 16;
+	y1 = yy - range;
+	if (y1 < 0)
+	    y1 = 0;
+	y2 = yy + range - 1;
+	if (y2 > (s->height - 16))
+	    y2 = s->height - 16;
+
+	*mx_ptr = (random() % (2 * (x2 - x1 + 1))) + 2 * (x1 - xx);
+	*my_ptr = (random() % (2 * (y2 - y1 + 1))) + 2 * (y1 - yy);
+	return 0;
+    } else {
+	*mx_ptr = 0;
+	*my_ptr = 0;
+	return 1;
+    }
+}
+
+#endif
diff --git a/libavcodec/mpeg12.c b/libavcodec/mpeg12.c
new file mode 100644
index 0000000000..3e2e523560
--- /dev/null
+++ b/libavcodec/mpeg12.c
@@ -0,0 +1,1530 @@
+/*
+ * MPEG1 encoder / MPEG2 decoder
+ * Copyright (c) 2000,2001 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include "avcodec.h"
+#include "dsputil.h"
+#include "mpegvideo.h"
+
+#include "mpeg12data.h"
+
+//#define DEBUG
+
+#ifdef DEBUG
+#define dprintf(fmt,args...) printf(fmt, ## args)
+#else
+#define dprintf(fmt,args...)
+#endif
+
+/* Start codes. */
+#define SEQ_END_CODE		0x000001b7
+#define SEQ_START_CODE		0x000001b3
+#define GOP_START_CODE		0x000001b8
+#define PICTURE_START_CODE	0x00000100
+#define SLICE_MIN_START_CODE	0x00000101
+#define SLICE_MAX_START_CODE	0x000001af
+#define EXT_START_CODE		0x000001b5
+#define USER_START_CODE		0x000001b2
+
+static void mpeg1_encode_block(MpegEncContext *s, 
+                         DCTELEM *block, 
+                         int component);
+static void mpeg1_encode_motion(MpegEncContext *s, int val);
+static void mpeg1_skip_picture(MpegEncContext *s, int pict_num);
+static int mpeg1_decode_block(MpegEncContext *s, 
+                              DCTELEM *block, 
+                              int n);
+static int mpeg2_decode_block_non_intra(MpegEncContext *s, 
+                                        DCTELEM *block, 
+                                        int n);
+static int mpeg2_decode_block_intra(MpegEncContext *s, 
+                                    DCTELEM *block, 
+                                    int n);
+static int mpeg_decode_motion(MpegEncContext *s, int fcode, int pred);
+
+static void put_header(MpegEncContext *s, int header)
+{
+    align_put_bits(&s->pb);
+    put_bits(&s->pb, 32, header);
+}
+
+/* put sequence header if needed */
+static void mpeg1_encode_sequence_header(MpegEncContext *s)
+{
+        unsigned int vbv_buffer_size;
+        unsigned int fps, n, v;
+        UINT64 time_code;
+        
+        if ((s->picture_number % s->gop_size) == 0) {
+            /* mpeg1 header repeated every gop */
+            put_header(s, SEQ_START_CODE);
+            
+            /* search closest frame rate */
+            {
+                int i, dmin, d;
+                s->frame_rate_index = 0;
+                dmin = 0x7fffffff;
+                for(i=1;i<9;i++) {
+                    d = abs(s->frame_rate - frame_rate_tab[i]);
+                    if (d < dmin) {
+                        dmin = d;
+                        s->frame_rate_index = i;
+                    }
+                }
+            }
+ 
+            put_bits(&s->pb, 12, s->width);
+            put_bits(&s->pb, 12, s->height);
+            put_bits(&s->pb, 4, 1); /* 1/1 aspect ratio */
+            put_bits(&s->pb, 4, s->frame_rate_index);
+            v = s->bit_rate / 400;
+            if (v > 0x3ffff)
+                v = 0x3ffff;
+            put_bits(&s->pb, 18, v);
+            put_bits(&s->pb, 1, 1); /* marker */
+            /* vbv buffer size: slightly greater than an I frame. We add
+               some margin just in case */
+            vbv_buffer_size = (3 * s->I_frame_bits) / (2 * 8);
+            put_bits(&s->pb, 10, (vbv_buffer_size + 16383) / 16384); 
+            put_bits(&s->pb, 1, 1); /* constrained parameter flag */
+            put_bits(&s->pb, 1, 0); /* no custom intra matrix */
+            put_bits(&s->pb, 1, 0); /* no custom non intra matrix */
+
+            put_header(s, GOP_START_CODE);
+            put_bits(&s->pb, 1, 0); /* do drop frame */
+            /* time code : we must convert from the real frame rate to a
+               fake mpeg frame rate in case of low frame rate */
+            fps = frame_rate_tab[s->frame_rate_index];
+            time_code = s->fake_picture_number * FRAME_RATE_BASE;
+            s->gop_picture_number = s->fake_picture_number;
+            put_bits(&s->pb, 5, (time_code / (fps * 3600)) % 24);
+            put_bits(&s->pb, 6, (time_code / (fps * 60)) % 60);
+            put_bits(&s->pb, 1, 1);
+            put_bits(&s->pb, 6, (time_code / fps) % 60);
+            put_bits(&s->pb, 6, (time_code % fps) / FRAME_RATE_BASE);
+            put_bits(&s->pb, 1, 1); /* closed gop */
+            put_bits(&s->pb, 1, 0); /* broken link */
+        }
+
+        if (s->frame_rate < (24 * FRAME_RATE_BASE) && s->picture_number > 0) {
+            /* insert empty P pictures to slow down to the desired
+               frame rate. Each fake pictures takes about 20 bytes */
+            fps = frame_rate_tab[s->frame_rate_index];
+            n = ((s->picture_number * fps) / s->frame_rate) - 1;
+            while (s->fake_picture_number < n) {
+                mpeg1_skip_picture(s, s->fake_picture_number - 
+                                   s->gop_picture_number); 
+                s->fake_picture_number++;
+            }
+
+        }
+        s->fake_picture_number++;
+}
+
+
+/* insert a fake P picture */
+static void mpeg1_skip_picture(MpegEncContext *s, int pict_num)
+{
+    unsigned int mb_incr;
+
+    /* mpeg1 picture header */
+    put_header(s, PICTURE_START_CODE);
+    /* temporal reference */
+    put_bits(&s->pb, 10, pict_num & 0x3ff); 
+    
+    put_bits(&s->pb, 3, P_TYPE);
+    put_bits(&s->pb, 16, 0xffff); /* non constant bit rate */
+    
+    put_bits(&s->pb, 1, 1); /* integer coordinates */
+    put_bits(&s->pb, 3, 1); /* forward_f_code */
+    
+    put_bits(&s->pb, 1, 0); /* extra bit picture */
+    
+    /* only one slice */
+    put_header(s, SLICE_MIN_START_CODE);
+    put_bits(&s->pb, 5, 1); /* quantizer scale */
+    put_bits(&s->pb, 1, 0); /* slice extra information */
+    
+    mb_incr = 1;
+    put_bits(&s->pb, mbAddrIncrTable[mb_incr - 1][1], 
+             mbAddrIncrTable[mb_incr - 1][0]);
+    
+    /* empty macroblock */
+    put_bits(&s->pb, 3, 1); /* motion only */
+    
+    /* zero motion x & y */
+    put_bits(&s->pb, 1, 1); 
+    put_bits(&s->pb, 1, 1); 
+
+    /* output a number of empty slice */
+    mb_incr = s->mb_width * s->mb_height - 1;
+    while (mb_incr > 33) {
+        put_bits(&s->pb, 11, 0x008);
+        mb_incr -= 33;
+    }
+    put_bits(&s->pb, mbAddrIncrTable[mb_incr - 1][1], 
+             mbAddrIncrTable[mb_incr - 1][0]);
+    
+    /* empty macroblock */
+    put_bits(&s->pb, 3, 1); /* motion only */
+    
+    /* zero motion x & y */
+    put_bits(&s->pb, 1, 1); 
+    put_bits(&s->pb, 1, 1); 
+}
+
+void mpeg1_encode_picture_header(MpegEncContext *s, int picture_number)
+{
+    static int done;
+
+    if (!done) {
+        done = 1;
+        init_rl(&rl_mpeg1);
+    }
+    mpeg1_encode_sequence_header(s);
+
+    /* mpeg1 picture header */
+    put_header(s, PICTURE_START_CODE);
+    /* temporal reference */
+    put_bits(&s->pb, 10, (s->fake_picture_number - 
+                          s->gop_picture_number) & 0x3ff); 
+    
+    put_bits(&s->pb, 3, s->pict_type);
+    put_bits(&s->pb, 16, 0xffff); /* non constant bit rate */
+    
+    if (s->pict_type == P_TYPE) {
+        put_bits(&s->pb, 1, 0); /* half pel coordinates */
+        put_bits(&s->pb, 3, s->f_code); /* forward_f_code */
+    }
+    
+    put_bits(&s->pb, 1, 0); /* extra bit picture */
+    
+    /* only one slice */
+    put_header(s, SLICE_MIN_START_CODE);
+    put_bits(&s->pb, 5, s->qscale); /* quantizer scale */
+    put_bits(&s->pb, 1, 0); /* slice extra information */
+}
+
+void mpeg1_encode_mb(MpegEncContext *s,
+                     DCTELEM block[6][64],
+                     int motion_x, int motion_y)
+{
+    int mb_incr, i, cbp, mb_x, mb_y;
+
+    mb_x = s->mb_x;
+    mb_y = s->mb_y;
+
+    /* compute cbp */
+    cbp = 0;
+    for(i=0;i<6;i++) {
+        if (s->block_last_index[i] >= 0)
+            cbp |= 1 << (5 - i);
+    }
+
+    /* skip macroblock, except if first or last macroblock of a slice */
+    if ((cbp | motion_x | motion_y) == 0 &&
+        (!((mb_x | mb_y) == 0 ||
+           (mb_x == s->mb_width - 1 && mb_y == s->mb_height - 1)))) {
+        s->mb_incr++;
+    } else {
+        /* output mb incr */
+        mb_incr = s->mb_incr;
+
+        while (mb_incr > 33) {
+            put_bits(&s->pb, 11, 0x008);
+            mb_incr -= 33;
+        }
+        put_bits(&s->pb, mbAddrIncrTable[mb_incr - 1][1], 
+                 mbAddrIncrTable[mb_incr - 1][0]);
+        
+        if (s->pict_type == I_TYPE) {
+            put_bits(&s->pb, 1, 1); /* macroblock_type : macroblock_quant = 0 */
+        } else {
+            if (s->mb_intra) {
+                put_bits(&s->pb, 5, 0x03);
+            } else {
+                if (cbp != 0) {
+                    if (motion_x == 0 && motion_y == 0) {
+                        put_bits(&s->pb, 2, 1); /* macroblock_pattern only */
+                        put_bits(&s->pb, mbPatTable[cbp - 1][1], mbPatTable[cbp - 1][0]);
+                    } else {
+                        put_bits(&s->pb, 1, 1); /* motion + cbp */
+                        mpeg1_encode_motion(s, motion_x - s->last_mv[0][0][0]); 
+                        mpeg1_encode_motion(s, motion_y - s->last_mv[0][0][1]); 
+                        put_bits(&s->pb, mbPatTable[cbp - 1][1], mbPatTable[cbp - 1][0]);
+                    }
+                } else {
+                    put_bits(&s->pb, 3, 1); /* motion only */
+                    mpeg1_encode_motion(s, motion_x - s->last_mv[0][0][0]); 
+                    mpeg1_encode_motion(s, motion_y - s->last_mv[0][0][1]); 
+                }
+            }
+        }
+        for(i=0;i<6;i++) {
+            if (cbp & (1 << (5 - i))) {
+                mpeg1_encode_block(s, block[i], i);
+            }
+        }
+        s->mb_incr = 1;
+    }
+    s->last_mv[0][0][0] = motion_x;
+    s->last_mv[0][0][1] = motion_y;
+}
+
+static void mpeg1_encode_motion(MpegEncContext *s, int val)
+{
+    int code, bit_size, l, m, bits, range, sign;
+
+    if (val == 0) {
+        /* zero vector */
+        code = 0;
+        put_bits(&s->pb,
+                 mbMotionVectorTable[0][1], 
+                 mbMotionVectorTable[0][0]); 
+    } else {
+        bit_size = s->f_code - 1;
+        range = 1 << bit_size;
+        /* modulo encoding */
+        l = 16 * range;
+        m = 2 * l;
+        if (val < -l) {
+            val += m;
+        } else if (val >= l) {
+            val -= m;
+        }
+
+        if (val >= 0) {
+            val--;
+            code = (val >> bit_size) + 1;
+            bits = val & (range - 1);
+            sign = 0;
+        } else {
+            val = -val;
+            val--;
+            code = (val >> bit_size) + 1;
+            bits = val & (range - 1);
+            sign = 1;
+        }
+        put_bits(&s->pb,
+                 mbMotionVectorTable[code][1], 
+                 mbMotionVectorTable[code][0]); 
+        put_bits(&s->pb, 1, sign);
+        if (bit_size > 0) {
+            put_bits(&s->pb, bit_size, bits);
+        }
+    }
+}
+
+static inline void encode_dc(MpegEncContext *s, int diff, int component)
+{
+    int adiff, index;
+
+    adiff = abs(diff);
+    index = vlc_dc_table[adiff];
+    if (component == 0) {
+        put_bits(&s->pb, vlc_dc_lum_bits[index], vlc_dc_lum_code[index]);
+    } else {
+        put_bits(&s->pb, vlc_dc_chroma_bits[index], vlc_dc_chroma_code[index]);
+    }
+    if (diff > 0) {
+        put_bits(&s->pb, index, (diff & ((1 << index) - 1)));
+    } else if (diff < 0) {
+        put_bits(&s->pb, index, ((diff - 1) & ((1 << index) - 1)));
+    }
+}
+
+static void mpeg1_encode_block(MpegEncContext *s, 
+                               DCTELEM *block, 
+                               int n)
+{
+    int alevel, level, last_non_zero, dc, diff, i, j, run, last_index, sign;
+    int code, component;
+    RLTable *rl = &rl_mpeg1;
+
+    last_index = s->block_last_index[n];
+
+    /* DC coef */
+    if (s->mb_intra) {
+        component = (n <= 3 ? 0 : n - 4 + 1);
+        dc = block[0]; /* overflow is impossible */
+        diff = dc - s->last_dc[component];
+        encode_dc(s, diff, component);
+        s->last_dc[component] = dc;
+        i = 1;
+    } else {
+        /* encode the first coefficient : needs to be done here because
+           it is handled slightly differently */
+        level = block[0];
+        if (abs(level) == 1) {
+                code = ((UINT32)level >> 31); /* the sign bit */
+                put_bits(&s->pb, 2, code | 0x02);
+                i = 1;
+        } else {
+            i = 0;
+            last_non_zero = -1;
+            goto next_coef;
+        }
+    }
+
+    /* now quantify & encode AC coefs */
+    last_non_zero = i - 1;
+    for(;i<=last_index;i++) {
+        j = zigzag_direct[i];
+        level = block[j];
+    next_coef:
+#if 0
+        if (level != 0)
+            dprintf("level[%d]=%d\n", i, level);
+#endif            
+        /* encode using VLC */
+        if (level != 0) {
+            run = i - last_non_zero - 1;
+            sign = 0;
+            alevel = level;
+            if (alevel < 0) {
+		sign = 1;
+                alevel = -alevel;
+	    }
+            code = get_rl_index(rl, 0, run, alevel);
+            put_bits(&s->pb, rl->table_vlc[code][1], rl->table_vlc[code][0]);
+            if (code != rl->n) {
+                put_bits(&s->pb, 1, sign);
+            } else {
+                /* escape: only clip in this case */
+                put_bits(&s->pb, 6, run);
+                if (alevel < 128) {
+                    put_bits(&s->pb, 8, level & 0xff);
+                } else {
+                    if (level < 0) {
+                        put_bits(&s->pb, 16, 0x8001 + level + 255);
+                    } else {
+                        put_bits(&s->pb, 16, level & 0xffff);
+                    }
+                }
+            }
+            last_non_zero = i;
+        }
+    }
+    /* end of block */
+    put_bits(&s->pb, 2, 0x2);
+}
+
+/******************************************/
+/* decoding */
+
+static VLC dc_lum_vlc;
+static VLC dc_chroma_vlc;
+static VLC mv_vlc;
+static VLC mbincr_vlc;
+static VLC mb_ptype_vlc;
+static VLC mb_btype_vlc;
+static VLC mb_pat_vlc;
+
+void mpeg1_init_vlc(MpegEncContext *s)
+{
+    static int done = 0;
+
+    if (!done) {
+
+        init_vlc(&dc_lum_vlc, 9, 12, 
+                 vlc_dc_lum_bits, 1, 1,
+                 vlc_dc_lum_code, 2, 2);
+        init_vlc(&dc_chroma_vlc, 9, 12, 
+                 vlc_dc_chroma_bits, 1, 1,
+                 vlc_dc_chroma_code, 2, 2);
+        init_vlc(&mv_vlc, 9, 17, 
+                 &mbMotionVectorTable[0][1], 2, 1,
+                 &mbMotionVectorTable[0][0], 2, 1);
+        init_vlc(&mbincr_vlc, 9, 34, 
+                 &mbAddrIncrTable[0][1], 2, 1,
+                 &mbAddrIncrTable[0][0], 2, 1);
+        init_vlc(&mb_pat_vlc, 9, 63, 
+                 &mbPatTable[0][1], 2, 1,
+                 &mbPatTable[0][0], 2, 1);
+        
+        init_vlc(&mb_ptype_vlc, 6, 32, 
+                 &table_mb_ptype[0][1], 2, 1,
+                 &table_mb_ptype[0][0], 2, 1);
+        init_vlc(&mb_btype_vlc, 6, 32, 
+                 &table_mb_btype[0][1], 2, 1,
+                 &table_mb_btype[0][0], 2, 1);
+        init_rl(&rl_mpeg1);
+        init_rl(&rl_mpeg2);
+        /* cannot use generic init because we must add the EOB code */
+        init_vlc(&rl_mpeg1.vlc, 9, rl_mpeg1.n + 2, 
+                 &rl_mpeg1.table_vlc[0][1], 4, 2,
+                 &rl_mpeg1.table_vlc[0][0], 4, 2);
+        init_vlc(&rl_mpeg2.vlc, 9, rl_mpeg2.n + 2, 
+                 &rl_mpeg2.table_vlc[0][1], 4, 2,
+                 &rl_mpeg2.table_vlc[0][0], 4, 2);
+    }
+}
+
+static inline int get_dmv(MpegEncContext *s)
+{
+    if(get_bits(&s->gb, 1)) 
+        return 1 - (get_bits(&s->gb, 1) << 1);
+    else
+        return 0;
+}
+
+/* motion type (for mpeg2) */
+#define MT_FIELD 1
+#define MT_FRAME 2
+#define MT_16X8  2
+#define MT_DMV   3
+
+static int mpeg_decode_mb(MpegEncContext *s,
+                          DCTELEM block[6][64])
+{
+    int i, j, k, cbp, val, code, mb_type, motion_type;
+    
+    /* skip mb handling */
+    if (s->mb_incr == 0) {
+        /* read again increment */
+        s->mb_incr = 1;
+        for(;;) {
+            code = get_vlc(&s->gb, &mbincr_vlc);
+            if (code < 0)
+                return 1; /* error = end of slice */
+            if (code >= 33) {
+                if (code == 33) {
+                    s->mb_incr += 33;
+                }
+                /* otherwise, stuffing, nothing to do */
+            } else {
+                s->mb_incr += code;
+                break;
+            }
+        }
+    }
+    if (++s->mb_x >= s->mb_width) {
+        s->mb_x = 0;
+        if (s->mb_y >= (s->mb_height - 1))
+            return -1;
+        s->mb_y++;
+    }
+    dprintf("decode_mb: x=%d y=%d\n", s->mb_x, s->mb_y);
+
+    if (--s->mb_incr != 0) {
+        /* skip mb */
+        s->mb_intra = 0;
+        for(i=0;i<6;i++)
+            s->block_last_index[i] = -1;
+        s->mv_type = MV_TYPE_16X16;
+        if (s->pict_type == P_TYPE) {
+            /* if P type, zero motion vector is implied */
+            s->mv_dir = MV_DIR_FORWARD;
+            s->mv[0][0][0] = s->mv[0][0][1] = 0;
+            s->last_mv[0][0][0] = s->last_mv[0][0][1] = 0;
+        } else {
+            /* if B type, reuse previous vectors and directions */
+            s->mv[0][0][0] = s->last_mv[0][0][0];
+            s->mv[0][0][1] = s->last_mv[0][0][1];
+            s->mv[1][0][0] = s->last_mv[1][0][0];
+            s->mv[1][0][1] = s->last_mv[1][0][1];
+        }
+        return 0;
+    }
+
+    switch(s->pict_type) {
+    default:
+    case I_TYPE:
+        if (get_bits(&s->gb, 1) == 0) {
+            if (get_bits(&s->gb, 1) == 0)
+                return -1;
+            mb_type = MB_QUANT | MB_INTRA;
+        } else {
+            mb_type = MB_INTRA;
+        }
+        break;
+    case P_TYPE:
+        mb_type = get_vlc(&s->gb, &mb_ptype_vlc);
+        if (mb_type < 0)
+            return -1;
+        break;
+    case B_TYPE:
+        mb_type = get_vlc(&s->gb, &mb_btype_vlc);
+        if (mb_type < 0)
+            return -1;
+        break;
+    }
+    dprintf("mb_type=%x\n", mb_type);
+    motion_type = 0; /* avoid warning */
+    if (mb_type & (MB_FOR|MB_BACK)) {
+        /* get additionnal motion vector type */
+        if (s->picture_structure == PICT_FRAME && s->frame_pred_frame_dct) 
+            motion_type = MT_FRAME;
+        else
+            motion_type = get_bits(&s->gb, 2);
+    }
+    /* compute dct type */
+    if (s->picture_structure == PICT_FRAME && 
+        !s->frame_pred_frame_dct &&
+        (mb_type & (MB_PAT | MB_INTRA))) {
+        s->interlaced_dct = get_bits(&s->gb, 1);
+#ifdef DEBUG
+        if (s->interlaced_dct)
+            printf("interlaced_dct\n");
+#endif
+    } else {
+        s->interlaced_dct = 0; /* frame based */
+    }
+
+    if (mb_type & MB_QUANT) {
+        if (s->mpeg2) {
+            if (s->q_scale_type) {
+                s->qscale = non_linear_qscale[get_bits(&s->gb, 5)];
+            } else {
+                s->qscale = get_bits(&s->gb, 5) << 1;
+            }
+        } else {
+            /* for mpeg1, we use the generic unquant code */
+            s->qscale = get_bits(&s->gb, 5);
+        }
+    }
+    if (mb_type & MB_INTRA) {
+        if (s->concealment_motion_vectors) {
+            /* just parse them */
+            if (s->picture_structure != PICT_FRAME) 
+                get_bits(&s->gb, 1); /* field select */
+            mpeg_decode_motion(s, s->mpeg_f_code[0][0], 0);
+            mpeg_decode_motion(s, s->mpeg_f_code[0][1], 0);
+        }
+        s->mb_intra = 1;
+        cbp = 0x3f;
+        memset(s->last_mv, 0, sizeof(s->last_mv)); /* reset mv prediction */
+    } else {
+        s->mb_intra = 0;
+        cbp = 0;
+    }
+    /* special case of implicit zero motion vector */
+    if (s->pict_type == P_TYPE && !(mb_type & MB_FOR)) {
+        s->mv_dir = MV_DIR_FORWARD;
+        s->mv_type = MV_TYPE_16X16;
+        s->last_mv[0][0][0] = 0;
+        s->last_mv[0][0][1] = 0;
+        s->mv[0][0][0] = 0;
+        s->mv[0][0][1] = 0;
+    } else if (mb_type & (MB_FOR | MB_BACK)) {
+        /* motion vectors */
+        s->mv_dir = 0;
+        for(i=0;i<2;i++) {
+            if (mb_type & (MB_FOR >> i)) {
+                s->mv_dir |= (MV_DIR_FORWARD >> i);
+                dprintf("mv_type=%d\n", motion_type);
+                switch(motion_type) {
+                case MT_FRAME: /* or MT_16X8 */
+                    if (s->picture_structure == PICT_FRAME) {
+                        /* MT_FRAME */
+                        s->mv_type = MV_TYPE_16X16;
+                        for(k=0;k<2;k++) {
+                            val = mpeg_decode_motion(s, s->mpeg_f_code[i][k], 
+                                                     s->last_mv[i][0][k]);
+                            s->last_mv[i][0][k] = val;
+                            s->last_mv[i][1][k] = val;
+                            /* full_pel: only for mpeg1 */
+                            if (s->full_pel[i])
+                                val = val << 1;
+                            s->mv[i][0][k] = val;
+                            dprintf("mv%d: %d\n", k, val);
+                        }
+                    } else {
+                        /* MT_16X8 */
+                        s->mv_type = MV_TYPE_16X8;
+                        for(j=0;j<2;j++) {
+                            s->field_select[i][j] = get_bits(&s->gb, 1);
+                            for(k=0;k<2;k++) {
+                                val = mpeg_decode_motion(s, s->mpeg_f_code[i][k],
+                                                         s->last_mv[i][j][k]);
+                                s->last_mv[i][j][k] = val;
+                                s->mv[i][j][k] = val;
+                            }
+                        }
+                    }
+                    break;
+                case MT_FIELD:
+                    if (s->picture_structure == PICT_FRAME) {
+                        s->mv_type = MV_TYPE_FIELD;
+                        for(j=0;j<2;j++) {
+                            s->field_select[i][j] = get_bits(&s->gb, 1);
+                            val = mpeg_decode_motion(s, s->mpeg_f_code[i][0],
+                                                     s->last_mv[i][j][0]);
+                            s->last_mv[i][j][0] = val;
+                            s->mv[i][j][0] = val;
+                            dprintf("fmx=%d\n", val);
+                            val = mpeg_decode_motion(s, s->mpeg_f_code[i][1],
+                                                     s->last_mv[i][j][1] >> 1);
+                            s->last_mv[i][j][1] = val << 1;
+                            s->mv[i][j][1] = val;
+                            dprintf("fmy=%d\n", val);
+                        }
+                    } else {
+                        s->mv_type = MV_TYPE_16X16;
+                        s->field_select[i][0] = get_bits(&s->gb, 1);
+                        for(k=0;k<2;k++) {
+                            val = mpeg_decode_motion(s, s->mpeg_f_code[i][k],
+                                                     s->last_mv[i][0][k]);
+                            s->last_mv[i][0][k] = val;
+                            s->last_mv[i][1][k] = val;
+                            s->mv[i][0][k] = val;
+                        }
+                    }
+                    break;
+                case MT_DMV:
+                    {
+                        int dmx, dmy, mx, my, m;
+
+                        mx = mpeg_decode_motion(s, s->mpeg_f_code[i][0], 
+                                                s->last_mv[i][0][0]);
+                        s->last_mv[i][0][0] = mx;
+                        s->last_mv[i][1][0] = mx;
+                        dmx = get_dmv(s);
+                        my = mpeg_decode_motion(s, s->mpeg_f_code[i][1], 
+                                                s->last_mv[i][0][1] >> 1);
+                        dmy = get_dmv(s);
+                        s->mv_type = MV_TYPE_DMV;
+                        /* XXX: totally broken */
+                        if (s->picture_structure == PICT_FRAME) {
+                            s->last_mv[i][0][1] = my << 1;
+                            s->last_mv[i][1][1] = my << 1;
+
+                            m = s->top_field_first ? 1 : 3;
+                            /* top -> top pred */
+                            s->mv[i][0][0] = mx; 
+                            s->mv[i][0][1] = my << 1;
+                            s->mv[i][1][0] = ((mx * m + (mx > 0)) >> 1) + dmx;
+                            s->mv[i][1][1] = ((my * m + (my > 0)) >> 1) + dmy - 1;
+                            m = 4 - m;
+                            s->mv[i][2][0] = mx;
+                            s->mv[i][2][1] = my << 1;
+                            s->mv[i][3][0] = ((mx * m + (mx > 0)) >> 1) + dmx;
+                            s->mv[i][3][1] = ((my * m + (my > 0)) >> 1) + dmy + 1;
+                        } else {
+                            s->last_mv[i][0][1] = my;
+                            s->last_mv[i][1][1] = my;
+                            s->mv[i][0][0] = mx;
+                            s->mv[i][0][1] = my;
+                            s->mv[i][1][0] = ((mx + (mx > 0)) >> 1) + dmx;
+                            s->mv[i][1][1] = ((my + (my > 0)) >> 1) + dmy - 1 
+                                /* + 2 * cur_field */;
+                        }
+                    }
+                    break;
+                }
+            }
+        }
+    }
+
+    if ((mb_type & MB_INTRA) && s->concealment_motion_vectors) {
+        get_bits(&s->gb, 1); /* marker */
+    }
+    
+    if (mb_type & MB_PAT) {
+        cbp = get_vlc(&s->gb, &mb_pat_vlc);
+        if (cbp < 0)
+            return -1;
+        cbp++;
+    }
+    dprintf("cbp=%x\n", cbp);
+
+    if (s->mpeg2) {
+        if (s->mb_intra) {
+            for(i=0;i<6;i++) {
+                if (cbp & (1 << (5 - i))) {
+                    if (mpeg2_decode_block_intra(s, block[i], i) < 0)
+                        return -1;
+                }
+            }
+        } else {
+            for(i=0;i<6;i++) {
+                if (cbp & (1 << (5 - i))) {
+                    if (mpeg2_decode_block_non_intra(s, block[i], i) < 0)
+                        return -1;
+                }
+            }
+        }
+    } else {
+        for(i=0;i<6;i++) {
+            if (cbp & (1 << (5 - i))) {
+                if (mpeg1_decode_block(s, block[i], i) < 0)
+                    return -1;
+            }
+        }
+    }
+    return 0;
+}
+
+/* as h263, but only 17 codes */
+static int mpeg_decode_motion(MpegEncContext *s, int fcode, int pred)
+{
+    int code, sign, val, m, l, shift;
+
+    code = get_vlc(&s->gb, &mv_vlc);
+    if (code < 0) {
+        return 0xffff;
+    }
+    if (code == 0) {
+        return pred;
+    }
+    sign = get_bits(&s->gb, 1);
+    shift = fcode - 1;
+    val = (code - 1) << shift;
+    if (shift > 0)
+        val |= get_bits(&s->gb, shift);
+    val++;
+    if (sign)
+        val = -val;
+    val += pred;
+    
+    /* modulo decoding */
+    l = (1 << shift) * 16;
+    m = 2 * l;
+    if (val < -l) {
+        val += m;
+    } else if (val >= l) {
+        val -= m;
+    }
+    return val;
+}
+
+static inline int decode_dc(MpegEncContext *s, int component)
+{
+    int code, diff;
+
+    if (component == 0) {
+        code = get_vlc(&s->gb, &dc_lum_vlc);
+    } else {
+        code = get_vlc(&s->gb, &dc_chroma_vlc);
+    }
+    if (code < 0)
+        return 0xffff;
+    if (code == 0) {
+        diff = 0;
+    } else {
+        diff = get_bits(&s->gb, code);
+        if ((diff & (1 << (code - 1))) == 0) 
+            diff = (-1 << code) | (diff + 1);
+    }
+    return diff;
+}
+
+static int mpeg1_decode_block(MpegEncContext *s, 
+                               DCTELEM *block, 
+                               int n)
+{
+    int level, dc, diff, i, j, run;
+    int code, component;
+    RLTable *rl = &rl_mpeg1;
+
+    if (s->mb_intra) {
+        /* DC coef */
+        component = (n <= 3 ? 0 : n - 4 + 1);
+        diff = decode_dc(s, component);
+        if (diff >= 0xffff)
+            return -1;
+        dc = s->last_dc[component];
+        dc += diff;
+        s->last_dc[component] = dc;
+        block[0] = dc;
+        dprintf("dc=%d diff=%d\n", dc, diff);
+        i = 1;
+    } else {
+        int bit_cnt, v;
+        UINT32 bit_buf;
+        UINT8 *buf_ptr;
+        i = 0;
+        /* special case for the first coef. no need to add a second vlc table */
+        SAVE_BITS(&s->gb);
+        SHOW_BITS(&s->gb, v, 2);
+        if (v & 2) {
+            run = 0;
+            level = 1 - ((v & 1) << 1);
+            FLUSH_BITS(2);
+            RESTORE_BITS(&s->gb);
+            goto add_coef;
+        }
+        RESTORE_BITS(&s->gb);
+    }
+
+    /* now quantify & encode AC coefs */
+    for(;;) {
+        code = get_vlc(&s->gb, &rl->vlc);
+        if (code < 0) {
+            return -1;
+        }
+        if (code == 112) {
+            break;
+        } else if (code == 111) {
+            /* escape */
+            run = get_bits(&s->gb, 6);
+            level = get_bits(&s->gb, 8);
+            level = (level << 24) >> 24;
+            if (level == -128) {
+                level = get_bits(&s->gb, 8) - 256;
+            } else if (level == 0) {
+                level = get_bits(&s->gb, 8);
+            }
+        } else {
+            run = rl->table_run[code];
+            level = rl->table_level[code];
+            if (get_bits(&s->gb, 1))
+                level = -level;
+        }
+        i += run;
+        if (i >= 64)
+            return -1;
+    add_coef:
+        dprintf("%d: run=%d level=%d\n", n, run, level);
+	j = zigzag_direct[i];
+        block[j] = level;
+        i++;
+    }
+    s->block_last_index[n] = i;
+    return 0;
+}
+
+/* Also does unquantization here, since I will never support mpeg2
+   encoding */
+static int mpeg2_decode_block_non_intra(MpegEncContext *s, 
+                                        DCTELEM *block, 
+                                        int n)
+{
+    int level, i, j, run;
+    int code;
+    RLTable *rl = &rl_mpeg1;
+    const UINT8 *scan_table;
+    const UINT16 *matrix;
+    int mismatch;
+
+    if (s->alternate_scan)
+        scan_table = ff_alternate_vertical_scan;
+    else
+        scan_table = zigzag_direct;
+    mismatch = 1;
+
+    {
+        int bit_cnt, v;
+        UINT32 bit_buf;
+        UINT8 *buf_ptr;
+        i = 0;
+        if (n < 4) 
+            matrix = s->non_intra_matrix;
+        else
+            matrix = s->chroma_non_intra_matrix;
+            
+        /* special case for the first coef. no need to add a second vlc table */
+        SAVE_BITS(&s->gb);
+        SHOW_BITS(&s->gb, v, 2);
+        if (v & 2) {
+            run = 0;
+            level = 1 - ((v & 1) << 1);
+            FLUSH_BITS(2);
+            RESTORE_BITS(&s->gb);
+            goto add_coef;
+        }
+        RESTORE_BITS(&s->gb);
+    }
+
+    /* now quantify & encode AC coefs */
+    for(;;) {
+        code = get_vlc(&s->gb, &rl->vlc);
+        if (code < 0)
+            return -1;
+        if (code == 112) {
+            break;
+        } else if (code == 111) {
+            /* escape */
+            run = get_bits(&s->gb, 6);
+            level = get_bits(&s->gb, 12);
+            level = (level << 20) >> 20;
+        } else {
+            run = rl->table_run[code];
+            level = rl->table_level[code];
+            if (get_bits(&s->gb, 1))
+                level = -level;
+        }
+        i += run;
+        if (i >= 64)
+            return -1;
+    add_coef:
+	j = scan_table[i];
+        dprintf("%d: run=%d level=%d\n", n, run, level);
+        level = ((level * 2 + 1) * s->qscale * matrix[j]) / 32;
+        /* XXX: is it really necessary to saturate since the encoder
+           knows whats going on ? */
+        mismatch ^= level;
+        block[j] = level;
+        i++;
+    }
+    block[63] ^= (mismatch & 1);
+    s->block_last_index[n] = i;
+    return 0;
+}
+
+static int mpeg2_decode_block_intra(MpegEncContext *s, 
+                                    DCTELEM *block, 
+                                    int n)
+{
+    int level, dc, diff, i, j, run;
+    int code, component;
+    RLTable *rl;
+    const UINT8 *scan_table;
+    const UINT16 *matrix;
+    int mismatch;
+
+    if (s->alternate_scan)
+        scan_table = ff_alternate_vertical_scan;
+    else
+        scan_table = zigzag_direct;
+    mismatch = 1;
+
+    /* DC coef */
+    component = (n <= 3 ? 0 : n - 4 + 1);
+    diff = decode_dc(s, component);
+    if (diff >= 0xffff)
+        return -1;
+    dc = s->last_dc[component];
+    dc += diff;
+    s->last_dc[component] = dc;
+    block[0] = dc << (3 - s->intra_dc_precision);
+    dprintf("dc=%d\n", block[0]);
+    i = 1;
+    if (s->intra_vlc_format)
+        rl = &rl_mpeg2;
+    else
+        rl = &rl_mpeg1;
+    if (n < 4) 
+        matrix = s->intra_matrix;
+    else
+        matrix = s->chroma_intra_matrix;
+        
+    /* now quantify & encode AC coefs */
+    for(;;) {
+        code = get_vlc(&s->gb, &rl->vlc);
+        if (code < 0)
+            return -1;
+        if (code == 112) {
+            break;
+        } else if (code == 111) {
+            /* escape */
+            run = get_bits(&s->gb, 6);
+            level = get_bits(&s->gb, 12);
+            level = (level << 20) >> 20;
+        } else {
+            run = rl->table_run[code];
+            level = rl->table_level[code];
+            if (get_bits(&s->gb, 1))
+                level = -level;
+        }
+        i += run;
+        if (i >= 64)
+            return -1;
+	j = scan_table[i];
+        dprintf("%d: run=%d level=%d\n", n, run, level);
+        level = (level * s->qscale * matrix[j]) / 16;
+        /* XXX: is it really necessary to saturate since the encoder
+           knows whats going on ? */
+        mismatch ^= level;
+        block[j] = level;
+        i++;
+    }
+    block[63] ^= (mismatch & 1);
+    s->block_last_index[n] = i;
+    return 0;
+}
+
+/* compressed picture size */
+#define PICTURE_BUFFER_SIZE 100000
+
+typedef struct Mpeg1Context {
+    MpegEncContext mpeg_enc_ctx;
+    UINT32 header_state;
+    int start_code; /* current start code */
+    UINT8 buffer[PICTURE_BUFFER_SIZE]; 
+    UINT8 *buf_ptr;
+    int buffer_size;
+    int mpeg_enc_ctx_allocated; /* true if decoding context allocated */
+} Mpeg1Context;
+
+static int mpeg_decode_init(AVCodecContext *avctx)
+{
+    Mpeg1Context *s = avctx->priv_data;
+
+    s->header_state = 0xff;
+    s->mpeg_enc_ctx_allocated = 0;
+    s->buffer_size = PICTURE_BUFFER_SIZE;
+    s->start_code = -1;
+    s->buf_ptr = s->buffer;
+    s->mpeg_enc_ctx.picture_number = 0;
+    return 0;
+}
+
+/* return the 8 bit start code value and update the search
+   state. Return -1 if no start code found */
+static int find_start_code(UINT8 **pbuf_ptr, UINT8 *buf_end, 
+                           UINT32 *header_state)
+{
+    UINT8 *buf_ptr;
+    unsigned int state, v;
+    int val;
+
+    state = *header_state;
+    buf_ptr = *pbuf_ptr;
+    while (buf_ptr < buf_end) {
+        v = *buf_ptr++;
+        if (state == 0x000001) {
+            state = ((state << 8) | v) & 0xffffff;
+            val = state;
+            goto found;
+        }
+        state = ((state << 8) | v) & 0xffffff;
+    }
+    val = -1;
+ found:
+    *pbuf_ptr = buf_ptr;
+    *header_state = state;
+    return val;
+}
+
+static int mpeg1_decode_picture(AVCodecContext *avctx, 
+                                UINT8 *buf, int buf_size)
+{
+    Mpeg1Context *s1 = avctx->priv_data;
+    MpegEncContext *s = &s1->mpeg_enc_ctx;
+    int ref, f_code;
+
+    init_get_bits(&s->gb, buf, buf_size);
+
+    ref = get_bits(&s->gb, 10); /* temporal ref */
+    s->pict_type = get_bits(&s->gb, 3);
+    dprintf("pict_type=%d\n", s->pict_type);
+    get_bits(&s->gb, 16);
+    if (s->pict_type == P_TYPE || s->pict_type == B_TYPE) {
+        s->full_pel[0] = get_bits(&s->gb, 1);
+        f_code = get_bits(&s->gb, 3);
+        if (f_code == 0)
+            return -1;
+        s->mpeg_f_code[0][0] = f_code;
+        s->mpeg_f_code[0][1] = f_code;
+    }
+    if (s->pict_type == B_TYPE) {
+        s->full_pel[1] = get_bits(&s->gb, 1);
+        f_code = get_bits(&s->gb, 3);
+        if (f_code == 0)
+            return -1;
+        s->mpeg_f_code[1][0] = f_code;
+        s->mpeg_f_code[1][1] = f_code;
+    }
+    s->y_dc_scale = 8;
+    s->c_dc_scale = 8;
+    s->first_slice = 1;
+    return 0;
+}
+
+static void mpeg_decode_sequence_extension(MpegEncContext *s)
+{
+    int horiz_size_ext, vert_size_ext;
+    int bit_rate_ext, vbv_buf_ext, low_delay;
+    int frame_rate_ext_n, frame_rate_ext_d;
+
+    get_bits(&s->gb, 8); /* profil and level */
+    get_bits(&s->gb, 1); /* progressive_sequence */
+    get_bits(&s->gb, 2); /* chroma_format */
+    horiz_size_ext = get_bits(&s->gb, 2);
+    vert_size_ext = get_bits(&s->gb, 2);
+    s->width |= (horiz_size_ext << 12);
+    s->height |= (vert_size_ext << 12);
+    bit_rate_ext = get_bits(&s->gb, 12);  /* XXX: handle it */
+    s->bit_rate = ((s->bit_rate / 400) | (bit_rate_ext << 12)) * 400;
+    get_bits(&s->gb, 1); /* marker */
+    vbv_buf_ext = get_bits(&s->gb, 8);
+    low_delay = get_bits(&s->gb, 1);
+    frame_rate_ext_n = get_bits(&s->gb, 2);
+    frame_rate_ext_d = get_bits(&s->gb, 5);
+    if (frame_rate_ext_d >= 1)
+        s->frame_rate = (s->frame_rate * frame_rate_ext_n) / frame_rate_ext_d;
+    dprintf("sequence extension\n");
+    s->mpeg2 = 1;
+}
+
+static void mpeg_decode_quant_matrix_extension(MpegEncContext *s)
+{
+    int i, v;
+
+    if (get_bits(&s->gb, 1)) {
+        for(i=0;i<64;i++) {
+            v = get_bits(&s->gb, 8);
+            s->intra_matrix[i] = v;
+            s->chroma_intra_matrix[i] = v;
+        }
+    }
+    if (get_bits(&s->gb, 1)) {
+        for(i=0;i<64;i++) {
+            v = get_bits(&s->gb, 8);
+            s->non_intra_matrix[i] = v;
+            s->chroma_non_intra_matrix[i] = v;
+        }
+    }
+    if (get_bits(&s->gb, 1)) {
+        for(i=0;i<64;i++) {
+            v = get_bits(&s->gb, 8);
+            s->chroma_intra_matrix[i] = v;
+        }
+    }
+    if (get_bits(&s->gb, 1)) {
+        for(i=0;i<64;i++) {
+            v = get_bits(&s->gb, 8);
+            s->chroma_non_intra_matrix[i] = v;
+        }
+    }
+}
+
+static void mpeg_decode_picture_coding_extension(MpegEncContext *s)
+{
+    s->full_pel[0] = s->full_pel[1] = 0;
+    s->mpeg_f_code[0][0] = get_bits(&s->gb, 4);
+    s->mpeg_f_code[0][1] = get_bits(&s->gb, 4);
+    s->mpeg_f_code[1][0] = get_bits(&s->gb, 4);
+    s->mpeg_f_code[1][1] = get_bits(&s->gb, 4);
+    s->intra_dc_precision = get_bits(&s->gb, 2);
+    s->picture_structure = get_bits(&s->gb, 2);
+    s->top_field_first = get_bits(&s->gb, 1);
+    s->frame_pred_frame_dct = get_bits(&s->gb, 1);
+    s->concealment_motion_vectors = get_bits(&s->gb, 1);
+    s->q_scale_type = get_bits(&s->gb, 1);
+    s->intra_vlc_format = get_bits(&s->gb, 1);
+    s->alternate_scan = get_bits(&s->gb, 1);
+    s->repeat_first_field = get_bits(&s->gb, 1);
+    s->chroma_420_type = get_bits(&s->gb, 1);
+    s->progressive_frame = get_bits(&s->gb, 1);
+    /* composite display not parsed */
+    dprintf("dc_preci=%d\n", s->intra_dc_precision);
+    dprintf("pict_structure=%d\n", s->picture_structure);
+    dprintf("conceal=%d\n", s->concealment_motion_vectors);
+    dprintf("intrafmt=%d\n", s->intra_vlc_format);
+    dprintf("frame_pred_frame_dct=%d\n", s->frame_pred_frame_dct);
+}
+
+static void mpeg_decode_extension(AVCodecContext *avctx, 
+                                  UINT8 *buf, int buf_size)
+{
+    Mpeg1Context *s1 = avctx->priv_data;
+    MpegEncContext *s = &s1->mpeg_enc_ctx;
+    int ext_type;
+
+    init_get_bits(&s->gb, buf, buf_size);
+    
+    ext_type = get_bits(&s->gb, 4);
+    switch(ext_type) {
+    case 0x1:
+        /* sequence ext */
+        mpeg_decode_sequence_extension(s);
+        break;
+    case 0x3:
+        /* quant matrix extension */
+        mpeg_decode_quant_matrix_extension(s);
+        break;
+    case 0x8:
+        /* picture extension */
+        mpeg_decode_picture_coding_extension(s);
+        break;
+    }
+}
+
+/* return 1 if end of frame */
+static int mpeg_decode_slice(AVCodecContext *avctx, 
+                              AVPicture *pict,
+                              int start_code,
+                              UINT8 *buf, int buf_size)
+{
+    Mpeg1Context *s1 = avctx->priv_data;
+    MpegEncContext *s = &s1->mpeg_enc_ctx;
+    int ret;
+    DCTELEM block[6][64];
+
+    start_code = (start_code - 1) & 0xff;
+    if (start_code >= s->mb_height)
+        return -1;
+    s->last_dc[0] = 1 << (7 + s->intra_dc_precision);
+    s->last_dc[1] = s->last_dc[0];
+    s->last_dc[2] = s->last_dc[0];
+    memset(s->last_mv, 0, sizeof(s->last_mv));
+    s->mb_x = -1;
+    s->mb_y = start_code;
+    s->mb_incr = 0;
+
+    /* start frame decoding */
+    if (s->first_slice) {
+        s->first_slice = 0;
+        MPV_frame_start(s);
+    }
+
+    init_get_bits(&s->gb, buf, buf_size);
+
+    s->qscale = get_bits(&s->gb, 5);
+    /* extra slice info */
+    while (get_bits(&s->gb, 1) != 0) {
+        get_bits(&s->gb, 8);
+    }
+
+    for(;;) {
+        memset(block, 0, sizeof(block));
+        ret = mpeg_decode_mb(s, block);
+        dprintf("ret=%d\n", ret);
+        if (ret < 0)
+            return -1;
+        if (ret == 1)
+            break;
+        MPV_decode_mb(s, block);
+    }
+    
+    /* end of slice reached */
+    if (s->mb_x == (s->mb_width - 1) &&
+        s->mb_y == (s->mb_height - 1)) {
+        /* end of image */
+        UINT8 **picture;
+
+        MPV_frame_end(s);
+
+        /* XXX: incorrect reported qscale for mpeg2 */
+        if (s->pict_type == B_TYPE) {
+            picture = s->current_picture;
+            avctx->quality = s->qscale;
+        } else {
+            /* latency of 1 frame for I and P frames */
+            /* XXX: use another variable than picture_number */
+            if (s->picture_number == 0) {
+                picture = NULL;
+            } else {
+                picture = s->last_picture;
+                avctx->quality = s->last_qscale;
+            }
+            s->last_qscale = s->qscale;
+            s->picture_number++;
+        }
+        if (picture) {
+            pict->data[0] = picture[0];
+            pict->data[1] = picture[1];
+            pict->data[2] = picture[2];
+            pict->linesize[0] = s->linesize;
+            pict->linesize[1] = s->linesize / 2;
+            pict->linesize[2] = s->linesize / 2;
+            return 1;
+        } else {
+            return 0;
+        }
+    } else {
+        return 0;
+    }
+}
+
+static int mpeg1_decode_sequence(AVCodecContext *avctx, 
+                                 UINT8 *buf, int buf_size)
+{
+    Mpeg1Context *s1 = avctx->priv_data;
+    MpegEncContext *s = &s1->mpeg_enc_ctx;
+    int width, height, i, v;
+    
+    init_get_bits(&s->gb, buf, buf_size);
+
+    width = get_bits(&s->gb, 12);
+    height = get_bits(&s->gb, 12);
+    get_bits(&s->gb, 4);
+    s->frame_rate_index = get_bits(&s->gb, 4);
+    if (s->frame_rate_index == 0)
+        return -1;
+    s->bit_rate = get_bits(&s->gb, 18) * 400;
+    if (get_bits(&s->gb, 1) == 0) /* marker */
+        return -1;
+    if (width <= 0 || height <= 0 ||
+        (width % 2) != 0 || (height % 2) != 0)
+        return -1;
+    if (width != s->width ||
+        height != s->height) {
+        /* start new mpeg1 context decoding */
+        s->out_format = FMT_MPEG1;
+        if (s1->mpeg_enc_ctx_allocated) {
+            MPV_common_end(s);
+        }
+        s->width = width;
+        s->height = height;
+        s->has_b_frames = 1;
+        avctx->width = width;
+        avctx->height = height;
+        avctx->frame_rate = frame_rate_tab[s->frame_rate_index];
+        avctx->bit_rate = s->bit_rate;
+        
+        if (MPV_common_init(s) < 0)
+            return -1;
+        mpeg1_init_vlc(s);
+        s1->mpeg_enc_ctx_allocated = 1;
+    }
+
+    get_bits(&s->gb, 10); /* vbv_buffer_size */
+    get_bits(&s->gb, 1);
+
+    /* get matrix */
+    if (get_bits(&s->gb, 1)) {
+        for(i=0;i<64;i++) {
+            v = get_bits(&s->gb, 8);
+            s->intra_matrix[i] = v;
+            s->chroma_intra_matrix[i] = v;
+        }
+    } else {
+        for(i=0;i<64;i++) {
+            v = default_intra_matrix[i];
+            s->intra_matrix[i] = v;
+            s->chroma_intra_matrix[i] = v;
+        }
+    }
+    if (get_bits(&s->gb, 1)) {
+        for(i=0;i<64;i++) {
+            v = get_bits(&s->gb, 8);
+            s->non_intra_matrix[i] = v;
+            s->chroma_non_intra_matrix[i] = v;
+        }
+    } else {
+        for(i=0;i<64;i++) {
+            v = default_non_intra_matrix[i];
+            s->non_intra_matrix[i] = v;
+            s->chroma_non_intra_matrix[i] = v;
+        }
+    }
+
+    /* we set mpeg2 parameters so that it emulates mpeg1 */
+    s->progressive_sequence = 1;
+    s->progressive_frame = 1;
+    s->picture_structure = PICT_FRAME;
+    s->frame_pred_frame_dct = 1;
+    s->mpeg2 = 0;
+    return 0;
+}
+
+/* handle buffering and image synchronisation */
+static int mpeg_decode_frame(AVCodecContext *avctx, 
+                             void *data, int *data_size,
+                             UINT8 *buf, int buf_size)
+{
+    Mpeg1Context *s = avctx->priv_data;
+    UINT8 *buf_end, *buf_ptr, *buf_start;
+    int len, start_code_found, ret, code, start_code, input_size;
+    AVPicture *picture = data;
+
+    dprintf("fill_buffer\n");
+
+    *data_size = 0;
+    /* special case for last picture */
+    if (buf_size == 0) {
+        MpegEncContext *s2 = &s->mpeg_enc_ctx;
+        if (s2->picture_number > 0) {
+            picture->data[0] = s2->next_picture[0];
+            picture->data[1] = s2->next_picture[1];
+            picture->data[2] = s2->next_picture[2];
+            picture->linesize[0] = s2->linesize;
+            picture->linesize[1] = s2->linesize / 2;
+            picture->linesize[2] = s2->linesize / 2;
+            *data_size = sizeof(AVPicture);
+        }
+        return 0;
+    }
+
+    buf_ptr = buf;
+    buf_end = buf + buf_size;
+    while (buf_ptr < buf_end) {
+        buf_start = buf_ptr;
+        /* find start next code */
+        code = find_start_code(&buf_ptr, buf_end, &s->header_state);
+        if (code >= 0) {
+            start_code_found = 1;
+        } else {
+            start_code_found = 0;
+        }
+        /* copy to buffer */
+        len = buf_ptr - buf_start;
+        if (len + (s->buf_ptr - s->buffer) > s->buffer_size) {
+            /* data too big : flush */
+            s->buf_ptr = s->buffer;
+            if (start_code_found)
+                s->start_code = code;
+        } else {
+            memcpy(s->buf_ptr, buf_start, len);
+            s->buf_ptr += len;
+            
+            if (start_code_found) {
+                /* prepare data for next start code */
+                input_size = s->buf_ptr - s->buffer;
+                start_code = s->start_code;
+                s->buf_ptr = s->buffer;
+                s->start_code = code;
+                switch(start_code) {
+                case SEQ_START_CODE:
+                    mpeg1_decode_sequence(avctx, s->buffer, 
+                                          input_size);
+                    break;
+                            
+                case PICTURE_START_CODE:
+                    /* we have a complete image : we try to decompress it */
+                    mpeg1_decode_picture(avctx, 
+                                         s->buffer, input_size);
+                    break;
+                case EXT_START_CODE:
+                    mpeg_decode_extension(avctx,
+                                          s->buffer, input_size);
+                    break;
+                default:
+                    if (start_code >= SLICE_MIN_START_CODE &&
+                        start_code <= SLICE_MAX_START_CODE) {
+                        ret = mpeg_decode_slice(avctx, picture,
+                                                start_code, s->buffer, input_size);
+                        if (ret == 1) {
+                            /* got a picture: exit */
+                            *data_size = sizeof(AVPicture);
+                            goto the_end;
+                        }
+                    }
+                    break;
+                }
+            }
+        }
+    }
+ the_end:
+    return buf_ptr - buf;
+}
+
+static int mpeg_decode_end(AVCodecContext *avctx)
+{
+    Mpeg1Context *s = avctx->priv_data;
+
+    if (s->mpeg_enc_ctx_allocated)
+        MPV_common_end(&s->mpeg_enc_ctx);
+    return 0;
+}
+
+AVCodec mpeg_decoder = {
+    "mpegvideo",
+    CODEC_TYPE_VIDEO,
+    CODEC_ID_MPEG1VIDEO,
+    sizeof(Mpeg1Context),
+    mpeg_decode_init,
+    NULL,
+    mpeg_decode_end,
+    mpeg_decode_frame,
+};
diff --git a/libavcodec/mpeg12data.h b/libavcodec/mpeg12data.h
new file mode 100644
index 0000000000..2b5a9c2e22
--- /dev/null
+++ b/libavcodec/mpeg12data.h
@@ -0,0 +1,362 @@
+/*
+ * MPEG1/2 tables
+ */
+
+const UINT8 default_intra_matrix[64] = {
+	8, 16, 19, 22, 26, 27, 29, 34,
+	16, 16, 22, 24, 27, 29, 34, 37,
+	19, 22, 26, 27, 29, 34, 34, 38,
+	22, 22, 26, 27, 29, 34, 37, 40,
+	22, 26, 27, 29, 32, 35, 40, 48,
+	26, 27, 29, 32, 35, 40, 48, 58,
+	26, 27, 29, 34, 38, 46, 56, 69,
+	27, 29, 35, 38, 46, 56, 69, 83
+};
+
+const UINT8 default_non_intra_matrix[64] = {
+    16, 16, 16, 16, 16, 16, 16, 16,
+    16, 16, 16, 16, 16, 16, 16, 16,
+    16, 16, 16, 16, 16, 16, 16, 16,
+    16, 16, 16, 16, 16, 16, 16, 16,
+    16, 16, 16, 16, 16, 16, 16, 16,
+    16, 16, 16, 16, 16, 16, 16, 16,
+    16, 16, 16, 16, 16, 16, 16, 16,
+    16, 16, 16, 16, 16, 16, 16, 16,
+};
+
+const unsigned char vlc_dc_table[256] = {
+    0, 1, 2, 2,
+    3, 3, 3, 3,
+    4, 4, 4, 4, 4, 4, 4, 4,
+    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+    6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+    6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+
+    7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+    7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+    7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+    7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+
+    8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+    8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+    8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+    8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+    8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+    8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+    8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+    8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+};
+
+const UINT16 vlc_dc_lum_code[12] = {
+    0x4, 0x0, 0x1, 0x5, 0x6, 0xe, 0x1e, 0x3e, 0x7e, 0xfe, 0x1fe, 0x1ff,
+};
+const unsigned char vlc_dc_lum_bits[12] = {
+    3, 2, 2, 3, 3, 4, 5, 6, 7, 8, 9, 9,
+};
+
+const UINT16 vlc_dc_chroma_code[12] = {
+    0x0, 0x1, 0x2, 0x6, 0xe, 0x1e, 0x3e, 0x7e, 0xfe, 0x1fe, 0x3fe, 0x3ff,
+};
+const unsigned char vlc_dc_chroma_bits[12] = {
+    2, 2, 2, 3, 4, 5, 6, 7, 8, 9, 10, 10,
+};
+
+static const UINT16 mpeg1_vlc[113][2] = {
+ { 0x3, 2 }, { 0x4, 4 }, { 0x5, 5 }, { 0x6, 7 },
+ { 0x26, 8 }, { 0x21, 8 }, { 0xa, 10 }, { 0x1d, 12 },
+ { 0x18, 12 }, { 0x13, 12 }, { 0x10, 12 }, { 0x1a, 13 },
+ { 0x19, 13 }, { 0x18, 13 }, { 0x17, 13 }, { 0x1f, 14 },
+ { 0x1e, 14 }, { 0x1d, 14 }, { 0x1c, 14 }, { 0x1b, 14 },
+ { 0x1a, 14 }, { 0x19, 14 }, { 0x18, 14 }, { 0x17, 14 },
+ { 0x16, 14 }, { 0x15, 14 }, { 0x14, 14 }, { 0x13, 14 },
+ { 0x12, 14 }, { 0x11, 14 }, { 0x10, 14 }, { 0x18, 15 },
+ { 0x17, 15 }, { 0x16, 15 }, { 0x15, 15 }, { 0x14, 15 },
+ { 0x13, 15 }, { 0x12, 15 }, { 0x11, 15 }, { 0x10, 15 },
+ { 0x3, 3 }, { 0x6, 6 }, { 0x25, 8 }, { 0xc, 10 },
+ { 0x1b, 12 }, { 0x16, 13 }, { 0x15, 13 }, { 0x1f, 15 },
+ { 0x1e, 15 }, { 0x1d, 15 }, { 0x1c, 15 }, { 0x1b, 15 },
+ { 0x1a, 15 }, { 0x19, 15 }, { 0x13, 16 }, { 0x12, 16 },
+ { 0x11, 16 }, { 0x10, 16 }, { 0x5, 4 }, { 0x4, 7 },
+ { 0xb, 10 }, { 0x14, 12 }, { 0x14, 13 }, { 0x7, 5 },
+ { 0x24, 8 }, { 0x1c, 12 }, { 0x13, 13 }, { 0x6, 5 },
+ { 0xf, 10 }, { 0x12, 12 }, { 0x7, 6 }, { 0x9, 10 },
+ { 0x12, 13 }, { 0x5, 6 }, { 0x1e, 12 }, { 0x14, 16 },
+ { 0x4, 6 }, { 0x15, 12 }, { 0x7, 7 }, { 0x11, 12 },
+ { 0x5, 7 }, { 0x11, 13 }, { 0x27, 8 }, { 0x10, 13 },
+ { 0x23, 8 }, { 0x1a, 16 }, { 0x22, 8 }, { 0x19, 16 },
+ { 0x20, 8 }, { 0x18, 16 }, { 0xe, 10 }, { 0x17, 16 },
+ { 0xd, 10 }, { 0x16, 16 }, { 0x8, 10 }, { 0x15, 16 },
+ { 0x1f, 12 }, { 0x1a, 12 }, { 0x19, 12 }, { 0x17, 12 },
+ { 0x16, 12 }, { 0x1f, 13 }, { 0x1e, 13 }, { 0x1d, 13 },
+ { 0x1c, 13 }, { 0x1b, 13 }, { 0x1f, 16 }, { 0x1e, 16 },
+ { 0x1d, 16 }, { 0x1c, 16 }, { 0x1b, 16 },
+ { 0x1, 6 }, /* escape */
+ { 0x2, 2 }, /* EOB */
+};
+
+static const UINT16 mpeg2_vlc[113][2] = {
+  {0x02, 2}, {0x06, 3}, {0x07, 4}, {0x1c, 5},
+  {0x1d, 5}, {0x05, 6}, {0x04, 6}, {0x7b, 7},
+  {0x7c, 7}, {0x23, 8}, {0x22, 8}, {0xfa, 8},
+  {0xfb, 8}, {0xfe, 8}, {0xff, 8}, {0x1f,14},
+  {0x1e,14}, {0x1d,14}, {0x1c,14}, {0x1b,14},
+  {0x1a,14}, {0x19,14}, {0x18,14}, {0x17,14},
+  {0x16,14}, {0x15,14}, {0x14,14}, {0x13,14},
+  {0x12,14}, {0x11,14}, {0x10,14}, {0x18,15},
+  {0x17,15}, {0x16,15}, {0x15,15}, {0x14,15},
+  {0x13,15}, {0x12,15}, {0x11,15}, {0x10,15},
+  {0x02, 3}, {0x06, 5}, {0x79, 7}, {0x27, 8},
+  {0x20, 8}, {0x16,13}, {0x15,13}, {0x1f,15},
+  {0x1e,15}, {0x1d,15}, {0x1c,15}, {0x1b,15},
+  {0x1a,15}, {0x19,15}, {0x13,16}, {0x12,16},
+  {0x11,16}, {0x10,16}, {0x05, 5}, {0x07, 7}, 
+  {0xfc, 8}, {0x0c,10}, {0x14,13}, {0x07, 5}, 
+  {0x26, 8}, {0x1c,12}, {0x13,13}, {0x06, 6}, 
+  {0xfd, 8}, {0x12,12}, {0x07, 6}, {0x04, 9}, 
+  {0x12,13}, {0x06, 7}, {0x1e,12}, {0x14,16}, 
+  {0x04, 7}, {0x15,12}, {0x05, 7}, {0x11,12}, 
+  {0x78, 7}, {0x11,13}, {0x7a, 7}, {0x10,13}, 
+  {0x21, 8}, {0x1a,16}, {0x25, 8}, {0x19,16}, 
+  {0x24, 8}, {0x18,16}, {0x05, 9}, {0x17,16}, 
+  {0x07, 9}, {0x16,16}, {0x0d,10}, {0x15,16}, 
+  {0x1f,12}, {0x1a,12}, {0x19,12}, {0x17,12}, 
+  {0x16,12}, {0x1f,13}, {0x1e,13}, {0x1d,13}, 
+  {0x1c,13}, {0x1b,13}, {0x1f,16}, {0x1e,16}, 
+  {0x1d,16}, {0x1c,16}, {0x1b,16}, 
+  {0x01,6}, /* escape */
+  {0x06,4}, /* EOB */
+};
+
+static const UINT8 mpeg1_level[111] = {
+  1,  2,  3,  4,  5,  6,  7,  8,
+  9, 10, 11, 12, 13, 14, 15, 16,
+ 17, 18, 19, 20, 21, 22, 23, 24,
+ 25, 26, 27, 28, 29, 30, 31, 32,
+ 33, 34, 35, 36, 37, 38, 39, 40,
+  1,  2,  3,  4,  5,  6,  7,  8,
+  9, 10, 11, 12, 13, 14, 15, 16,
+ 17, 18,  1,  2,  3,  4,  5,  1,
+  2,  3,  4,  1,  2,  3,  1,  2,
+  3,  1,  2,  3,  1,  2,  1,  2,
+  1,  2,  1,  2,  1,  2,  1,  2,
+  1,  2,  1,  2,  1,  2,  1,  2,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  1,  1,
+};
+
+static const UINT8 mpeg1_run[111] = {
+  0,  0,  0,  0,  0,  0,  0,  0,
+  0,  0,  0,  0,  0,  0,  0,  0,
+  0,  0,  0,  0,  0,  0,  0,  0,
+  0,  0,  0,  0,  0,  0,  0,  0,
+  0,  0,  0,  0,  0,  0,  0,  0,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  2,  2,  2,  2,  2,  3,
+  3,  3,  3,  4,  4,  4,  5,  5,
+  5,  6,  6,  6,  7,  7,  8,  8,
+  9,  9, 10, 10, 11, 11, 12, 12,
+ 13, 13, 14, 14, 15, 15, 16, 16,
+ 17, 18, 19, 20, 21, 22, 23, 24,
+ 25, 26, 27, 28, 29, 30, 31,
+};
+
+static RLTable rl_mpeg1 = {
+    111,
+    111,
+    mpeg1_vlc,
+    mpeg1_run,
+    mpeg1_level,
+};
+
+static RLTable rl_mpeg2 = {
+    111,
+    111,
+    mpeg2_vlc,
+    mpeg1_run,
+    mpeg1_level,
+};
+
+static const UINT8 mbAddrIncrTable[35][2] = {
+    {0x1, 1},
+    {0x3, 3},
+    {0x2, 3},
+    {0x3, 4},
+    {0x2, 4},
+    {0x3, 5},
+    {0x2, 5},
+    {0x7, 7},
+    {0x6, 7},
+    {0xb, 8},
+    {0xa, 8},
+    {0x9, 8},
+    {0x8, 8},
+    {0x7, 8},
+    {0x6, 8},
+    {0x17, 10},
+    {0x16, 10},
+    {0x15, 10},
+    {0x14, 10},
+    {0x13, 10},
+    {0x12, 10},
+    {0x23, 11},
+    {0x22, 11},
+    {0x21, 11},
+    {0x20, 11},
+    {0x1f, 11},
+    {0x1e, 11},
+    {0x1d, 11},
+    {0x1c, 11},
+    {0x1b, 11},
+    {0x1a, 11},
+    {0x19, 11},
+    {0x18, 11},
+    {0x8, 11}, /* escape */
+    {0xf, 11}, /* stuffing */
+};
+
+static const UINT8 mbPatTable[63][2] = {
+    {0xb, 5},
+    {0x9, 5},
+    {0xd, 6},
+    {0xd, 4},
+    {0x17, 7},
+    {0x13, 7},
+    {0x1f, 8},
+    {0xc, 4},
+    {0x16, 7},
+    {0x12, 7},
+    {0x1e, 8},
+    {0x13, 5},
+    {0x1b, 8},
+    {0x17, 8},
+    {0x13, 8},
+    {0xb, 4},
+    {0x15, 7},
+    {0x11, 7},
+    {0x1d, 8},
+    {0x11, 5},
+    {0x19, 8},
+    {0x15, 8},
+    {0x11, 8},
+    {0xf, 6},
+    {0xf, 8},
+    {0xd, 8},
+    {0x3, 9},
+    {0xf, 5},
+    {0xb, 8},
+    {0x7, 8},
+    {0x7, 9},
+    {0xa, 4},
+    {0x14, 7},
+    {0x10, 7},
+    {0x1c, 8},
+    {0xe, 6},
+    {0xe, 8},
+    {0xc, 8},
+    {0x2, 9},
+    {0x10, 5},
+    {0x18, 8},
+    {0x14, 8},
+    {0x10, 8},
+    {0xe, 5},
+    {0xa, 8},
+    {0x6, 8},
+    {0x6, 9},
+    {0x12, 5},
+    {0x1a, 8},
+    {0x16, 8},
+    {0x12, 8},
+    {0xd, 5},
+    {0x9, 8},
+    {0x5, 8},
+    {0x5, 9},
+    {0xc, 5},
+    {0x8, 8},
+    {0x4, 8},
+    {0x4, 9},
+    {0x7, 3},
+    {0xa, 5},
+    {0x8, 5},
+    {0xc, 6}
+};
+
+#define MB_INTRA 0x01
+#define MB_PAT   0x02
+#define MB_BACK  0x04
+#define MB_FOR   0x08
+#define MB_QUANT 0x10  
+
+static const UINT8 table_mb_ptype[32][2] = {
+    [ MB_FOR|MB_PAT ] { 1, 1 },
+    [ MB_PAT ] { 1, 2 },
+    [ MB_FOR ] { 1, 3 },
+    [ MB_INTRA ] { 3, 5 },
+    [ MB_QUANT|MB_FOR|MB_PAT ] { 2, 5 },
+    [ MB_QUANT|MB_PAT ] { 1, 5 },
+    [ MB_QUANT|MB_INTRA ] { 1, 6 },
+};
+
+static const UINT8 table_mb_btype[32][2] = {
+    [ MB_FOR|MB_BACK ] { 2, 2 },
+    [ MB_FOR|MB_BACK|MB_PAT ] { 3, 2 },
+    [ MB_BACK ] { 2, 3 },
+    [ MB_BACK|MB_PAT ] { 3, 3 },
+    [ MB_FOR ] { 2, 4 },
+    [ MB_FOR|MB_PAT ] { 3, 4 },
+    [ MB_INTRA ] { 3, 5 },
+    [ MB_QUANT|MB_FOR|MB_BACK|MB_PAT ] { 2, 5 },
+    [ MB_QUANT|MB_FOR|MB_PAT ] { 3, 6 },
+    [ MB_QUANT|MB_BACK|MB_PAT ] { 2, 6 },
+    [ MB_QUANT|MB_INTRA ] { 1, 6 },
+};
+
+static const UINT8 mbMotionVectorTable[17][2] = {
+{ 0x1, 1 },
+{ 0x1, 2 },
+{ 0x1, 3 },
+{ 0x1, 4 },
+{ 0x3, 6 },
+{ 0x5, 7 },
+{ 0x4, 7 },
+{ 0x3, 7 },
+{ 0xb, 9 },
+{ 0xa, 9 },
+{ 0x9, 9 },
+{ 0x11, 10 },
+{ 0x10, 10 },
+{ 0xf, 10 },
+{ 0xe, 10 },
+{ 0xd, 10 },
+{ 0xc, 10 },
+};
+
+const UINT8 zigzag_direct[64] = {
+    0, 1, 8, 16, 9, 2, 3, 10,
+    17, 24, 32, 25, 18, 11, 4, 5,
+    12, 19, 26, 33, 40, 48, 41, 34,
+    27, 20, 13, 6, 7, 14, 21, 28,
+    35, 42, 49, 56, 57, 50, 43, 36,
+    29, 22, 15, 23, 30, 37, 44, 51,
+    58, 59, 52, 45, 38, 31, 39, 46,
+    53, 60, 61, 54, 47, 55, 62, 63
+};
+
+static const int frame_rate_tab[9] = {
+    0, 
+    (int)(23.976 * FRAME_RATE_BASE), 
+    (int)(24 * FRAME_RATE_BASE), 
+    (int)(25 * FRAME_RATE_BASE), 
+    (int)(29.97 * FRAME_RATE_BASE), 
+    (int)(30 * FRAME_RATE_BASE), 
+    (int)(50 * FRAME_RATE_BASE), 
+    (int)(59.94 * FRAME_RATE_BASE), 
+    (int)(60 * FRAME_RATE_BASE), 
+};
+
+static const UINT8 non_linear_qscale[32] = {
+    0, 1, 2, 3, 4, 5, 6, 7,
+    8,10,12,14,16,18,20,22,
+    24,28,32,36,40,44,48,52,
+    56,64,72,80,88,96,104,112,
+};
diff --git a/libavcodec/mpeg4data.h b/libavcodec/mpeg4data.h
new file mode 100644
index 0000000000..54b93d97ed
--- /dev/null
+++ b/libavcodec/mpeg4data.h
@@ -0,0 +1,106 @@
+
+/* dc encoding for mpeg4 */
+static const UINT8 DCtab_lum[13][2] =
+{
+    {3,3}, {3,2}, {2,2}, {2,3}, {1,3}, {1,4}, {1,5}, {1,6}, {1,7},
+    {1,8}, {1,9}, {1,10}, {1,11},
+}; 
+
+static const UINT8 DCtab_chrom[13][2] =
+{
+    {3,2}, {2,2}, {1,2}, {1,3}, {1,4}, {1,5}, {1,6}, {1,7}, {1,8},
+    {1,9}, {1,10}, {1,11}, {1,12},
+}; 
+
+const UINT16 intra_vlc[103][2] = {
+{ 0x2, 2 },
+{ 0x6, 3 },{ 0xf, 4 },{ 0xd, 5 },{ 0xc, 5 },
+{ 0x15, 6 },{ 0x13, 6 },{ 0x12, 6 },{ 0x17, 7 },
+{ 0x1f, 8 },{ 0x1e, 8 },{ 0x1d, 8 },{ 0x25, 9 },
+{ 0x24, 9 },{ 0x23, 9 },{ 0x21, 9 },{ 0x21, 10 },
+{ 0x20, 10 },{ 0xf, 10 },{ 0xe, 10 },{ 0x7, 11 },
+{ 0x6, 11 },{ 0x20, 11 },{ 0x21, 11 },{ 0x50, 12 },
+{ 0x51, 12 },{ 0x52, 12 },{ 0xe, 4 },{ 0x14, 6 },
+{ 0x16, 7 },{ 0x1c, 8 },{ 0x20, 9 },{ 0x1f, 9 },
+{ 0xd, 10 },{ 0x22, 11 },{ 0x53, 12 },{ 0x55, 12 },
+{ 0xb, 5 },{ 0x15, 7 },{ 0x1e, 9 },{ 0xc, 10 },
+{ 0x56, 12 },{ 0x11, 6 },{ 0x1b, 8 },{ 0x1d, 9 },
+{ 0xb, 10 },{ 0x10, 6 },{ 0x22, 9 },{ 0xa, 10 },
+{ 0xd, 6 },{ 0x1c, 9 },{ 0x8, 10 },{ 0x12, 7 },
+{ 0x1b, 9 },{ 0x54, 12 },{ 0x14, 7 },{ 0x1a, 9 },
+{ 0x57, 12 },{ 0x19, 8 },{ 0x9, 10 },{ 0x18, 8 },
+{ 0x23, 11 },{ 0x17, 8 },{ 0x19, 9 },{ 0x18, 9 },
+{ 0x7, 10 },{ 0x58, 12 },{ 0x7, 4 },{ 0xc, 6 },
+{ 0x16, 8 },{ 0x17, 9 },{ 0x6, 10 },{ 0x5, 11 },
+{ 0x4, 11 },{ 0x59, 12 },{ 0xf, 6 },{ 0x16, 9 },
+{ 0x5, 10 },{ 0xe, 6 },{ 0x4, 10 },{ 0x11, 7 },
+{ 0x24, 11 },{ 0x10, 7 },{ 0x25, 11 },{ 0x13, 7 },
+{ 0x5a, 12 },{ 0x15, 8 },{ 0x5b, 12 },{ 0x14, 8 },
+{ 0x13, 8 },{ 0x1a, 8 },{ 0x15, 9 },{ 0x14, 9 },
+{ 0x13, 9 },{ 0x12, 9 },{ 0x11, 9 },{ 0x26, 11 },
+{ 0x27, 11 },{ 0x5c, 12 },{ 0x5d, 12 },{ 0x5e, 12 },
+{ 0x5f, 12 },{ 0x3, 7 },
+};
+
+const INT8 intra_level[102] = {
+  1,  2,  3,  4,  5,  6,  7,  8,
+  9, 10, 11, 12, 13, 14, 15, 16,
+ 17, 18, 19, 20, 21, 22, 23, 24,
+ 25, 26, 27,  1,  2,  3,  4,  5,
+  6,  7,  8,  9, 10,  1,  2,  3,
+  4,  5,  1,  2,  3,  4,  1,  2,
+  3,  1,  2,  3,  1,  2,  3,  1,
+  2,  3,  1,  2,  1,  2,  1,  1,
+  1,  1,  1,  1,  2,  3,  4,  5,
+  6,  7,  8,  1,  2,  3,  1,  2,
+  1,  2,  1,  2,  1,  2,  1,  2,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  1,
+};
+
+const INT8 intra_run[102] = {
+  0,  0,  0,  0,  0,  0,  0,  0,
+  0,  0,  0,  0,  0,  0,  0,  0,
+  0,  0,  0,  0,  0,  0,  0,  0,
+  0,  0,  0,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  2,  2,  2,
+  2,  2,  3,  3,  3,  3,  4,  4,
+  4,  5,  5,  5,  6,  6,  6,  7,
+  7,  7,  8,  8,  9,  9, 10, 11,
+ 12, 13, 14,  0,  0,  0,  0,  0,
+  0,  0,  0,  1,  1,  1,  2,  2,
+  3,  3,  4,  4,  5,  5,  6,  6,
+  7,  8,  9, 10, 11, 12, 13, 14,
+ 15, 16, 17, 18, 19, 20,
+};
+
+static RLTable rl_intra = {
+    102,
+    67,
+    intra_vlc,
+    intra_run,
+    intra_level,
+};
+
+/* alternate scan orders used when doing AC prediction */
+UINT8 ff_alternate_horizontal_scan[64] = {
+    0,  1,  2,  3,  8,  9, 16, 17, 
+    10, 11,  4,  5,  6,  7, 15, 14,
+    13, 12, 19, 18, 24, 25, 32, 33, 
+    26, 27, 20, 21, 22, 23, 28, 29,
+    30, 31, 34, 35, 40, 41, 48, 49, 
+    42, 43, 36, 37, 38, 39, 44, 45,
+    46, 47, 50, 51, 56, 57, 58, 59, 
+    52, 53, 54, 55, 60, 61, 62, 63,
+};
+
+UINT8 ff_alternate_vertical_scan[64] = {
+    0,  8, 16, 24,  1,  9,  2, 10, 
+    17, 25, 32, 40, 48, 56, 57, 49,
+    41, 33, 26, 18,  3, 11,  4, 12, 
+    19, 27, 34, 42, 50, 58, 35, 43,
+    51, 59, 20, 28,  5, 13,  6, 14, 
+    21, 29, 36, 44, 52, 60, 37, 45,
+    53, 61, 22, 30,  7, 15, 23, 31, 
+    38, 46, 54, 62, 39, 47, 55, 63,
+};
diff --git a/libavcodec/mpegaudio.c b/libavcodec/mpegaudio.c
new file mode 100644
index 0000000000..af05e29279
--- /dev/null
+++ b/libavcodec/mpegaudio.c
@@ -0,0 +1,774 @@
+/*
+ * The simplest mpeg audio layer 2 encoder
+ * Copyright (c) 2000 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <math.h>
+#include "avcodec.h"
+#include "mpegaudio.h"
+
+#define NDEBUG
+#include <assert.h>
+
+/* define it to use floats in quantization (I don't like floats !) */
+//#define USE_FLOATS
+
+#define MPA_STEREO  0
+#define MPA_JSTEREO 1
+#define MPA_DUAL    2
+#define MPA_MONO    3
+
+#include "mpegaudiotab.h"
+
+int MPA_encode_init(AVCodecContext *avctx)
+{
+    MpegAudioContext *s = avctx->priv_data;
+    int freq = avctx->sample_rate;
+    int bitrate = avctx->bit_rate;
+    int channels = avctx->channels;
+    int i, v, table, ch_bitrate;
+    float a;
+
+    if (channels > 2)
+        return -1;
+    bitrate = bitrate / 1000;
+    s->nb_channels = channels;
+    s->freq = freq;
+    s->bit_rate = bitrate * 1000;
+    avctx->frame_size = MPA_FRAME_SIZE;
+    avctx->key_frame = 1; /* always key frame */
+
+    /* encoding freq */
+    s->lsf = 0;
+    for(i=0;i<3;i++) {
+        if (freq_tab[i] == freq) 
+            break;
+        if ((freq_tab[i] / 2) == freq) {
+            s->lsf = 1;
+            break;
+        }
+    }
+    if (i == 3)
+        return -1;
+    s->freq_index = i;
+
+    /* encoding bitrate & frequency */
+    for(i=0;i<15;i++) {
+        if (bitrate_tab[1-s->lsf][i] == bitrate) 
+            break;
+    }
+    if (i == 15)
+        return -1;
+    s->bitrate_index = i;
+
+    /* compute total header size & pad bit */
+    
+    a = (float)(bitrate * 1000 * MPA_FRAME_SIZE) / (freq * 8.0);
+    s->frame_size = ((int)a) * 8;
+
+    /* frame fractional size to compute padding */
+    s->frame_frac = 0;
+    s->frame_frac_incr = (int)((a - floor(a)) * 65536.0);
+    
+    /* select the right allocation table */
+    ch_bitrate = bitrate / s->nb_channels;
+    if (!s->lsf) {
+        if ((freq == 48000 && ch_bitrate >= 56) ||
+            (ch_bitrate >= 56 && ch_bitrate <= 80)) 
+            table = 0;
+        else if (freq != 48000 && ch_bitrate >= 96) 
+            table = 1;
+        else if (freq != 32000 && ch_bitrate <= 48) 
+            table = 2;
+        else 
+            table = 3;
+    } else {
+        table = 4;
+    }
+    /* number of used subbands */
+    s->sblimit = sblimit_table[table];
+    s->alloc_table = alloc_tables[table];
+
+#ifdef DEBUG
+    printf("%d kb/s, %d Hz, frame_size=%d bits, table=%d, padincr=%x\n", 
+           bitrate, freq, s->frame_size, table, s->frame_frac_incr);
+#endif
+
+    for(i=0;i<s->nb_channels;i++)
+        s->samples_offset[i] = 0;
+
+    for(i=0;i<512;i++) {
+        float a = enwindow[i] * 32768.0 * 16.0;
+        filter_bank[i] = (int)(a);
+    }
+    for(i=0;i<64;i++) {
+        v = (int)(pow(2.0, (3 - i) / 3.0) * (1 << 20));
+        if (v <= 0)
+            v = 1;
+        scale_factor_table[i] = v;
+#ifdef USE_FLOATS
+        scale_factor_inv_table[i] = pow(2.0, -(3 - i) / 3.0) / (float)(1 << 20);
+#else
+#define P 15
+        scale_factor_shift[i] = 21 - P - (i / 3);
+        scale_factor_mult[i] = (1 << P) * pow(2.0, (i % 3) / 3.0);
+#endif
+    }
+    for(i=0;i<128;i++) {
+        v = i - 64;
+        if (v <= -3)
+            v = 0;
+        else if (v < 0)
+            v = 1;
+        else if (v == 0)
+            v = 2;
+        else if (v < 3)
+            v = 3;
+        else 
+            v = 4;
+        scale_diff_table[i] = v;
+    }
+
+    for(i=0;i<17;i++) {
+        v = quant_bits[i];
+        if (v < 0) 
+            v = -v;
+        else
+            v = v * 3;
+        total_quant_bits[i] = 12 * v;
+    }
+
+    return 0;
+}
+
+/* 32 point floating point IDCT */
+static void idct32(int *out, int *tab, int sblimit, int left_shift)
+{
+    int i, j;
+    int *t, *t1, xr;
+    const int *xp = costab32;
+
+    for(j=31;j>=3;j-=2) tab[j] += tab[j - 2];
+    
+    t = tab + 30;
+    t1 = tab + 2;
+    do {
+        t[0] += t[-4];
+        t[1] += t[1 - 4];
+        t -= 4;
+    } while (t != t1);
+
+    t = tab + 28;
+    t1 = tab + 4;
+    do {
+        t[0] += t[-8];
+        t[1] += t[1-8];
+        t[2] += t[2-8];
+        t[3] += t[3-8];
+        t -= 8;
+    } while (t != t1);
+    
+    t = tab;
+    t1 = tab + 32;
+    do {
+        t[ 3] = -t[ 3];    
+        t[ 6] = -t[ 6];    
+        
+        t[11] = -t[11];    
+        t[12] = -t[12];    
+        t[13] = -t[13];    
+        t[15] = -t[15]; 
+        t += 16;
+    } while (t != t1);
+
+    
+    t = tab;
+    t1 = tab + 8;
+    do {
+        int x1, x2, x3, x4;
+        
+        x3 = MUL(t[16], FIX(SQRT2*0.5));
+        x4 = t[0] - x3;
+        x3 = t[0] + x3;
+        
+        x2 = MUL(-(t[24] + t[8]), FIX(SQRT2*0.5));
+        x1 = MUL((t[8] - x2), xp[0]);
+        x2 = MUL((t[8] + x2), xp[1]);
+
+        t[ 0] = x3 + x1;
+        t[ 8] = x4 - x2;
+        t[16] = x4 + x2;
+        t[24] = x3 - x1;
+        t++;
+    } while (t != t1);
+
+    xp += 2;
+    t = tab;
+    t1 = tab + 4;
+    do {
+        xr = MUL(t[28],xp[0]);
+        t[28] = (t[0] - xr);
+        t[0] = (t[0] + xr);
+
+        xr = MUL(t[4],xp[1]);
+        t[ 4] = (t[24] - xr);
+        t[24] = (t[24] + xr);
+        
+        xr = MUL(t[20],xp[2]);
+        t[20] = (t[8] - xr);
+        t[ 8] = (t[8] + xr);
+            
+        xr = MUL(t[12],xp[3]);
+        t[12] = (t[16] - xr);
+        t[16] = (t[16] + xr);
+        t++;
+    } while (t != t1);
+    xp += 4;
+
+    for (i = 0; i < 4; i++) {
+        xr = MUL(tab[30-i*4],xp[0]);
+        tab[30-i*4] = (tab[i*4] - xr);
+        tab[   i*4] = (tab[i*4] + xr);
+        
+        xr = MUL(tab[ 2+i*4],xp[1]);
+        tab[ 2+i*4] = (tab[28-i*4] - xr);
+        tab[28-i*4] = (tab[28-i*4] + xr);
+        
+        xr = MUL(tab[31-i*4],xp[0]);
+        tab[31-i*4] = (tab[1+i*4] - xr);
+        tab[ 1+i*4] = (tab[1+i*4] + xr);
+        
+        xr = MUL(tab[ 3+i*4],xp[1]);
+        tab[ 3+i*4] = (tab[29-i*4] - xr);
+        tab[29-i*4] = (tab[29-i*4] + xr);
+        
+        xp += 2;
+    }
+
+    t = tab + 30;
+    t1 = tab + 1;
+    do {
+        xr = MUL(t1[0], *xp);
+        t1[0] = (t[0] - xr);
+        t[0] = (t[0] + xr);
+        t -= 2;
+        t1 += 2;
+        xp++;
+    } while (t >= tab);
+
+    for(i=0;i<32;i++) {
+        out[i] = tab[bitinv32[i]] << left_shift;
+    }
+}
+
+static void filter(MpegAudioContext *s, int ch, short *samples, int incr)
+{
+    short *p, *q;
+    int sum, offset, i, j, norm, n;
+    short tmp[64];
+    int tmp1[32];
+    int *out;
+
+    //    print_pow1(samples, 1152);
+
+    offset = s->samples_offset[ch];
+    out = &s->sb_samples[ch][0][0][0];
+    for(j=0;j<36;j++) {
+        /* 32 samples at once */
+        for(i=0;i<32;i++) {
+            s->samples_buf[ch][offset + (31 - i)] = samples[0];
+            samples += incr;
+        }
+
+        /* filter */
+        p = s->samples_buf[ch] + offset;
+        q = filter_bank;
+        /* maxsum = 23169 */
+        for(i=0;i<64;i++) {
+            sum = p[0*64] * q[0*64];
+            sum += p[1*64] * q[1*64];
+            sum += p[2*64] * q[2*64];
+            sum += p[3*64] * q[3*64];
+            sum += p[4*64] * q[4*64];
+            sum += p[5*64] * q[5*64];
+            sum += p[6*64] * q[6*64];
+            sum += p[7*64] * q[7*64];
+            tmp[i] = sum >> 14;
+            p++;
+            q++;
+        }
+        tmp1[0] = tmp[16];
+        for( i=1; i<=16; i++ ) tmp1[i] = tmp[i+16]+tmp[16-i];
+        for( i=17; i<=31; i++ ) tmp1[i] = tmp[i+16]-tmp[80-i];
+
+        /* integer IDCT 32 with normalization. XXX: There may be some
+           overflow left */
+        norm = 0;
+        for(i=0;i<32;i++) {
+            norm |= abs(tmp1[i]);
+        }
+        n = log2(norm) - 12;
+        if (n > 0) {
+            for(i=0;i<32;i++) 
+                tmp1[i] >>= n;
+        } else {
+            n = 0;
+        }
+
+        idct32(out, tmp1, s->sblimit, n);
+
+        /* advance of 32 samples */
+        offset -= 32;
+        out += 32;
+        /* handle the wrap around */
+        if (offset < 0) {
+            memmove(s->samples_buf[ch] + SAMPLES_BUF_SIZE - (512 - 32), 
+                    s->samples_buf[ch], (512 - 32) * 2);
+            offset = SAMPLES_BUF_SIZE - 512;
+        }
+    }
+    s->samples_offset[ch] = offset;
+
+    //    print_pow(s->sb_samples, 1152);
+}
+
+static void compute_scale_factors(unsigned char scale_code[SBLIMIT],
+                                  unsigned char scale_factors[SBLIMIT][3], 
+                                  int sb_samples[3][12][SBLIMIT],
+                                  int sblimit)
+{
+    int *p, vmax, v, n, i, j, k, code;
+    int index, d1, d2;
+    unsigned char *sf = &scale_factors[0][0];
+    
+    for(j=0;j<sblimit;j++) {
+        for(i=0;i<3;i++) {
+            /* find the max absolute value */
+            p = &sb_samples[i][0][j];
+            vmax = abs(*p);
+            for(k=1;k<12;k++) {
+                p += SBLIMIT;
+                v = abs(*p);
+                if (v > vmax)
+                    vmax = v;
+            }
+            /* compute the scale factor index using log 2 computations */
+            if (vmax > 0) {
+                n = log2(vmax);
+                /* n is the position of the MSB of vmax. now 
+                   use at most 2 compares to find the index */
+                index = (21 - n) * 3 - 3;
+                if (index >= 0) {
+                    while (vmax <= scale_factor_table[index+1])
+                        index++;
+                } else {
+                    index = 0; /* very unlikely case of overflow */
+                }
+            } else {
+                index = 63;
+            }
+            
+#if 0
+            printf("%2d:%d in=%x %x %d\n", 
+                   j, i, vmax, scale_factor_table[index], index);
+#endif
+            /* store the scale factor */
+            assert(index >=0 && index <= 63);
+            sf[i] = index;
+        }
+
+        /* compute the transmission factor : look if the scale factors
+           are close enough to each other */
+        d1 = scale_diff_table[sf[0] - sf[1] + 64];
+        d2 = scale_diff_table[sf[1] - sf[2] + 64];
+        
+        /* handle the 25 cases */
+        switch(d1 * 5 + d2) {
+        case 0*5+0:
+        case 0*5+4:
+        case 3*5+4:
+        case 4*5+0:
+        case 4*5+4:
+            code = 0;
+            break;
+        case 0*5+1:
+        case 0*5+2:
+        case 4*5+1:
+        case 4*5+2:
+            code = 3;
+            sf[2] = sf[1];
+            break;
+        case 0*5+3:
+        case 4*5+3:
+            code = 3;
+            sf[1] = sf[2];
+            break;
+        case 1*5+0:
+        case 1*5+4:
+        case 2*5+4:
+            code = 1;
+            sf[1] = sf[0];
+            break;
+        case 1*5+1:
+        case 1*5+2:
+        case 2*5+0:
+        case 2*5+1:
+        case 2*5+2:
+            code = 2;
+            sf[1] = sf[2] = sf[0];
+            break;
+        case 2*5+3:
+        case 3*5+3:
+            code = 2;
+            sf[0] = sf[1] = sf[2];
+            break;
+        case 3*5+0:
+        case 3*5+1:
+        case 3*5+2:
+            code = 2;
+            sf[0] = sf[2] = sf[1];
+            break;
+        case 1*5+3:
+            code = 2;
+            if (sf[0] > sf[2])
+              sf[0] = sf[2];
+            sf[1] = sf[2] = sf[0];
+            break;
+        default:
+            abort();
+        }
+        
+#if 0
+        printf("%d: %2d %2d %2d %d %d -> %d\n", j, 
+               sf[0], sf[1], sf[2], d1, d2, code);
+#endif
+        scale_code[j] = code;
+        sf += 3;
+    }
+}
+
+/* The most important function : psycho acoustic module. In this
+   encoder there is basically none, so this is the worst you can do,
+   but also this is the simpler. */
+static void psycho_acoustic_model(MpegAudioContext *s, short smr[SBLIMIT])
+{
+    int i;
+
+    for(i=0;i<s->sblimit;i++) {
+        smr[i] = (int)(fixed_smr[i] * 10);
+    }
+}
+
+
+#define SB_NOTALLOCATED  0
+#define SB_ALLOCATED     1
+#define SB_NOMORE        2
+
+/* Try to maximize the smr while using a number of bits inferior to
+   the frame size. I tried to make the code simpler, faster and
+   smaller than other encoders :-) */
+static void compute_bit_allocation(MpegAudioContext *s, 
+                                   short smr1[MPA_MAX_CHANNELS][SBLIMIT],
+                                   unsigned char bit_alloc[MPA_MAX_CHANNELS][SBLIMIT],
+                                   int *padding)
+{
+    int i, ch, b, max_smr, max_ch, max_sb, current_frame_size, max_frame_size;
+    int incr;
+    short smr[MPA_MAX_CHANNELS][SBLIMIT];
+    unsigned char subband_status[MPA_MAX_CHANNELS][SBLIMIT];
+    const unsigned char *alloc;
+
+    memcpy(smr, smr1, s->nb_channels * sizeof(short) * SBLIMIT);
+    memset(subband_status, SB_NOTALLOCATED, s->nb_channels * SBLIMIT);
+    memset(bit_alloc, 0, s->nb_channels * SBLIMIT);
+    
+    /* compute frame size and padding */
+    max_frame_size = s->frame_size;
+    s->frame_frac += s->frame_frac_incr;
+    if (s->frame_frac >= 65536) {
+        s->frame_frac -= 65536;
+        s->do_padding = 1;
+        max_frame_size += 8;
+    } else {
+        s->do_padding = 0;
+    }
+
+    /* compute the header + bit alloc size */
+    current_frame_size = 32;
+    alloc = s->alloc_table;
+    for(i=0;i<s->sblimit;i++) {
+        incr = alloc[0];
+        current_frame_size += incr * s->nb_channels;
+        alloc += 1 << incr;
+    }
+    for(;;) {
+        /* look for the subband with the largest signal to mask ratio */
+        max_sb = -1;
+        max_ch = -1;
+        max_smr = 0x80000000;
+        for(ch=0;ch<s->nb_channels;ch++) {
+            for(i=0;i<s->sblimit;i++) {
+                if (smr[ch][i] > max_smr && subband_status[ch][i] != SB_NOMORE) {
+                    max_smr = smr[ch][i];
+                    max_sb = i;
+                    max_ch = ch;
+                }
+            }
+        }
+#if 0
+        printf("current=%d max=%d max_sb=%d alloc=%d\n", 
+               current_frame_size, max_frame_size, max_sb,
+               bit_alloc[max_sb]);
+#endif        
+        if (max_sb < 0)
+            break;
+        
+        /* find alloc table entry (XXX: not optimal, should use
+           pointer table) */
+        alloc = s->alloc_table;
+        for(i=0;i<max_sb;i++) {
+            alloc += 1 << alloc[0];
+        }
+
+        if (subband_status[max_ch][max_sb] == SB_NOTALLOCATED) {
+            /* nothing was coded for this band: add the necessary bits */
+            incr = 2 + nb_scale_factors[s->scale_code[max_ch][max_sb]] * 6;
+            incr += total_quant_bits[alloc[1]];
+        } else {
+            /* increments bit allocation */
+            b = bit_alloc[max_ch][max_sb];
+            incr = total_quant_bits[alloc[b + 1]] - 
+                total_quant_bits[alloc[b]];
+        }
+
+        if (current_frame_size + incr <= max_frame_size) {
+            /* can increase size */
+            b = ++bit_alloc[max_ch][max_sb];
+            current_frame_size += incr;
+            /* decrease smr by the resolution we added */
+            smr[max_ch][max_sb] = smr1[max_ch][max_sb] - quant_snr[alloc[b]];
+            /* max allocation size reached ? */
+            if (b == ((1 << alloc[0]) - 1))
+                subband_status[max_ch][max_sb] = SB_NOMORE;
+            else
+                subband_status[max_ch][max_sb] = SB_ALLOCATED;
+        } else {
+            /* cannot increase the size of this subband */
+            subband_status[max_ch][max_sb] = SB_NOMORE;
+        }
+    }
+    *padding = max_frame_size - current_frame_size;
+    assert(*padding >= 0);
+
+#if 0
+    for(i=0;i<s->sblimit;i++) {
+        printf("%d ", bit_alloc[i]);
+    }
+    printf("\n");
+#endif
+}
+
+/*
+ * Output the mpeg audio layer 2 frame. Note how the code is small
+ * compared to other encoders :-)
+ */
+static void encode_frame(MpegAudioContext *s,
+                         unsigned char bit_alloc[MPA_MAX_CHANNELS][SBLIMIT],
+                         int padding)
+{
+    int i, j, k, l, bit_alloc_bits, b, ch;
+    unsigned char *sf;
+    int q[3];
+    PutBitContext *p = &s->pb;
+
+    /* header */
+
+    put_bits(p, 12, 0xfff);
+    put_bits(p, 1, 1 - s->lsf); /* 1 = mpeg1 ID, 0 = mpeg2 lsf ID */
+    put_bits(p, 2, 4-2);  /* layer 2 */
+    put_bits(p, 1, 1); /* no error protection */
+    put_bits(p, 4, s->bitrate_index);
+    put_bits(p, 2, s->freq_index);
+    put_bits(p, 1, s->do_padding); /* use padding */
+    put_bits(p, 1, 0);             /* private_bit */
+    put_bits(p, 2, s->nb_channels == 2 ? MPA_STEREO : MPA_MONO);
+    put_bits(p, 2, 0); /* mode_ext */
+    put_bits(p, 1, 0); /* no copyright */
+    put_bits(p, 1, 1); /* original */
+    put_bits(p, 2, 0); /* no emphasis */
+
+    /* bit allocation */
+    j = 0;
+    for(i=0;i<s->sblimit;i++) {
+        bit_alloc_bits = s->alloc_table[j];
+        for(ch=0;ch<s->nb_channels;ch++) {
+            put_bits(p, bit_alloc_bits, bit_alloc[ch][i]);
+        }
+        j += 1 << bit_alloc_bits;
+    }
+    
+    /* scale codes */
+    for(i=0;i<s->sblimit;i++) {
+        for(ch=0;ch<s->nb_channels;ch++) {
+            if (bit_alloc[ch][i]) 
+                put_bits(p, 2, s->scale_code[ch][i]);
+        }
+    }
+
+    /* scale factors */
+    for(i=0;i<s->sblimit;i++) {
+        for(ch=0;ch<s->nb_channels;ch++) {
+            if (bit_alloc[ch][i]) {
+                sf = &s->scale_factors[ch][i][0];
+                switch(s->scale_code[ch][i]) {
+                case 0:
+                    put_bits(p, 6, sf[0]);
+                    put_bits(p, 6, sf[1]);
+                    put_bits(p, 6, sf[2]);
+                    break;
+                case 3:
+                case 1:
+                    put_bits(p, 6, sf[0]);
+                    put_bits(p, 6, sf[2]);
+                    break;
+                case 2:
+                    put_bits(p, 6, sf[0]);
+                    break;
+                }
+            }
+        }
+    }
+    
+    /* quantization & write sub band samples */
+
+    for(k=0;k<3;k++) {
+        for(l=0;l<12;l+=3) {
+            j = 0;
+            for(i=0;i<s->sblimit;i++) {
+                bit_alloc_bits = s->alloc_table[j];
+                for(ch=0;ch<s->nb_channels;ch++) {
+                    b = bit_alloc[ch][i];
+                    if (b) {
+                        int qindex, steps, m, sample, bits;
+                        /* we encode 3 sub band samples of the same sub band at a time */
+                        qindex = s->alloc_table[j+b];
+                        steps = quant_steps[qindex];
+                        for(m=0;m<3;m++) {
+                            sample = s->sb_samples[ch][k][l + m][i];
+                            /* divide by scale factor */
+#ifdef USE_FLOATS
+                            {
+                                float a;
+                                a = (float)sample * scale_factor_inv_table[s->scale_factors[ch][i][k]];
+                                q[m] = (int)((a + 1.0) * steps * 0.5);
+                            }
+#else
+                            {
+                                int q1, e, shift, mult;
+                                e = s->scale_factors[ch][i][k];
+                                shift = scale_factor_shift[e];
+                                mult = scale_factor_mult[e];
+                                
+                                /* normalize to P bits */
+                                if (shift < 0)
+                                    q1 = sample << (-shift);
+                                else
+                                    q1 = sample >> shift;
+                                q1 = (q1 * mult) >> P;
+                                q[m] = ((q1 + (1 << P)) * steps) >> (P + 1);
+                            }
+#endif
+                            if (q[m] >= steps)
+                                q[m] = steps - 1;
+                            assert(q[m] >= 0 && q[m] < steps);
+                        }
+                        bits = quant_bits[qindex];
+                        if (bits < 0) {
+                            /* group the 3 values to save bits */
+                            put_bits(p, -bits, 
+                                     q[0] + steps * (q[1] + steps * q[2]));
+#if 0
+                            printf("%d: gr1 %d\n", 
+                                   i, q[0] + steps * (q[1] + steps * q[2]));
+#endif
+                        } else {
+#if 0
+                            printf("%d: gr3 %d %d %d\n", 
+                                   i, q[0], q[1], q[2]);
+#endif                               
+                            put_bits(p, bits, q[0]);
+                            put_bits(p, bits, q[1]);
+                            put_bits(p, bits, q[2]);
+                        }
+                    }
+                }
+                /* next subband in alloc table */
+                j += 1 << bit_alloc_bits; 
+            }
+        }
+    }
+
+    /* padding */
+    for(i=0;i<padding;i++)
+        put_bits(p, 1, 0);
+
+    /* flush */
+    flush_put_bits(p);
+}
+
+int MPA_encode_frame(AVCodecContext *avctx,
+                     unsigned char *frame, int buf_size, void *data)
+{
+    MpegAudioContext *s = avctx->priv_data;
+    short *samples = data;
+    short smr[MPA_MAX_CHANNELS][SBLIMIT];
+    unsigned char bit_alloc[MPA_MAX_CHANNELS][SBLIMIT];
+    int padding, i;
+
+    for(i=0;i<s->nb_channels;i++) {
+        filter(s, i, samples + i, s->nb_channels);
+    }
+
+    for(i=0;i<s->nb_channels;i++) {
+        compute_scale_factors(s->scale_code[i], s->scale_factors[i], 
+                              s->sb_samples[i], s->sblimit);
+    }
+    for(i=0;i<s->nb_channels;i++) {
+        psycho_acoustic_model(s, smr[i]);
+    }
+    compute_bit_allocation(s, smr, bit_alloc, &padding);
+
+    init_put_bits(&s->pb, frame, MPA_MAX_CODED_FRAME_SIZE, NULL, NULL);
+
+    encode_frame(s, bit_alloc, padding);
+    
+    s->nb_samples += MPA_FRAME_SIZE;
+    return s->pb.buf_ptr - s->pb.buf;
+}
+
+
+AVCodec mp2_encoder = {
+    "mp2",
+    CODEC_TYPE_AUDIO,
+    CODEC_ID_MP2,
+    sizeof(MpegAudioContext),
+    MPA_encode_init,
+    MPA_encode_frame,
+    NULL,
+};
diff --git a/libavcodec/mpegaudio.h b/libavcodec/mpegaudio.h
new file mode 100644
index 0000000000..006650a99c
--- /dev/null
+++ b/libavcodec/mpegaudio.h
@@ -0,0 +1,34 @@
+
+/* max compressed frame size */
+#define MPA_MAX_CODED_FRAME_SIZE 1200
+
+#define MPA_FRAME_SIZE 1152 
+#define MPA_MAX_CHANNELS 2
+
+#define SAMPLES_BUF_SIZE 4096
+#define SBLIMIT 32 /* number of subbands */
+#define DCT_BITS 14 /* number of bits for the DCT */
+#define MUL(a,b) (((a) * (b)) >> DCT_BITS)
+#define FIX(a)   ((int)((a) * (1 << DCT_BITS)))
+
+typedef struct MpegAudioContext {
+    PutBitContext pb;
+    int nb_channels;
+    int freq, bit_rate;
+    int lsf;           /* 1 if mpeg2 low bitrate selected */
+    int bitrate_index; /* bit rate */
+    int freq_index;
+    int frame_size; /* frame size, in bits, without padding */
+    long long nb_samples; /* total number of samples encoded */
+    /* padding computation */
+    int frame_frac, frame_frac_incr, do_padding;
+    short samples_buf[MPA_MAX_CHANNELS][SAMPLES_BUF_SIZE]; /* buffer for filter */
+    int samples_offset[MPA_MAX_CHANNELS];       /* offset in samples_buf */
+    int sb_samples[MPA_MAX_CHANNELS][3][12][SBLIMIT];
+    unsigned char scale_factors[MPA_MAX_CHANNELS][SBLIMIT][3]; /* scale factors */
+    /* code to group 3 scale factors */
+    unsigned char scale_code[MPA_MAX_CHANNELS][SBLIMIT];       
+    int sblimit; /* number of used subbands */
+    const unsigned char *alloc_table;
+} MpegAudioContext;
+
diff --git a/libavcodec/mpegaudiodec.c b/libavcodec/mpegaudiodec.c
new file mode 100644
index 0000000000..f3fa90af55
--- /dev/null
+++ b/libavcodec/mpegaudiodec.c
@@ -0,0 +1,293 @@
+/*
+ * MPEG Audio decoder
+ * Copyright (c) 2001 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include "avcodec.h"
+#include "mpglib/mpg123.h"
+
+/*
+ * TODO: 
+ *  - add free format
+ *  - do not rely anymore on mpglib (first step: implement dct64 and decoding filter)
+ */
+
+#define HEADER_SIZE 4
+#define BACKSTEP_SIZE 512
+
+typedef struct MPADecodeContext {
+    struct mpstr mpstr;
+    UINT8 inbuf1[2][MAXFRAMESIZE + BACKSTEP_SIZE];	/* input buffer */
+    int inbuf_index;
+    UINT8 *inbuf_ptr, *inbuf;
+    int frame_size;
+    int error_protection;
+    int layer;
+    int sample_rate;
+    int bit_rate;
+    int old_frame_size;
+    GetBitContext gb;
+} MPADecodeContext;
+
+/* XXX: suppress that mess */
+struct mpstr *gmp;
+GetBitContext *gmp_gb;
+static MPADecodeContext *gmp_s;
+
+/* XXX: merge constants with encoder */
+static const unsigned short mp_bitrate_tab[2][3][15] = {
+    { {0, 32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448 },
+      {0, 32, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 384 },
+      {0, 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320 } },
+    { {0, 32, 48, 56, 64, 80, 96, 112, 128, 144, 160, 176, 192, 224, 256},
+      {0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160},
+      {0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160}
+    }
+};
+
+static unsigned short mp_freq_tab[3] = { 44100, 48000, 32000 };
+
+static int decode_init(AVCodecContext * avctx)
+{
+    MPADecodeContext *s = avctx->priv_data;
+    struct mpstr *mp = &s->mpstr;
+    static int init;
+
+    mp->fr.single = -1;
+    mp->synth_bo = 1;
+
+    if(!init) {
+        init = 1;
+        make_decode_tables(32767);
+        init_layer2();
+        init_layer3(SBLIMIT);
+    }
+
+    s->inbuf_index = 0;
+    s->inbuf = &s->inbuf1[s->inbuf_index][BACKSTEP_SIZE];
+    s->inbuf_ptr = s->inbuf;
+    
+    return 0;
+}
+
+/* fast header check for resync */
+static int check_header(UINT32 header)
+{
+    /* header */
+    if ((header & 0xffe00000) != 0xffe00000)
+	return -1;
+    /* layer check */
+    if (((header >> 17) & 3) == 0)
+	return -1;
+    /* bit rate : currently no free format supported */
+    if (((header >> 12) & 0xf) == 0xf ||
+        ((header >> 12) & 0xf) == 0x0)
+	return -1;
+    /* frequency */
+    if (((header >> 10) & 3) == 3)
+	return -1;
+    return 0;
+}
+
+/* header decoding. MUST check the header before because no
+   consistency check is done there */
+static void decode_header(MPADecodeContext *s, UINT32 header)
+{
+    struct frame *fr = &s->mpstr.fr;
+    int sample_rate, frame_size;
+
+    if (header & (1<<20)) {
+        fr->lsf = (header & (1<<19)) ? 0 : 1;
+        fr->mpeg25 = 0;
+    } else {
+        fr->lsf = 1;
+        fr->mpeg25 = 1;
+    }
+    
+    s->layer = 4 - ((header >> 17) & 3);
+    /* extract frequency */
+    fr->sampling_frequency = ((header >> 10) & 3);
+    sample_rate = mp_freq_tab[fr->sampling_frequency] >> (fr->lsf + fr->mpeg25);
+    fr->sampling_frequency += 3 * (fr->lsf + fr->mpeg25);
+
+    s->error_protection = ((header>>16) & 1) ^ 1;
+
+    fr->bitrate_index = ((header>>12)&0xf);
+    fr->padding   = ((header>>9)&0x1);
+    fr->extension = ((header>>8)&0x1);
+    fr->mode      = ((header>>6)&0x3);
+    fr->mode_ext  = ((header>>4)&0x3);
+    fr->copyright = ((header>>3)&0x1);
+    fr->original  = ((header>>2)&0x1);
+    fr->emphasis  = header & 0x3;
+
+    fr->stereo    = (fr->mode == MPG_MD_MONO) ? 1 : 2;
+
+    
+    frame_size = mp_bitrate_tab[fr->lsf][s->layer - 1][fr->bitrate_index];
+    s->bit_rate = frame_size * 1000;
+    switch(s->layer) {
+    case 1:
+        frame_size = (frame_size * 12000) / sample_rate;
+        frame_size = ((frame_size + fr->padding) << 2);
+        break;
+    case 2:
+        frame_size = (frame_size * 144000) / sample_rate;
+        frame_size += fr->padding;
+        break;
+    case 3:
+        frame_size = (frame_size * 144000) / (sample_rate << fr->lsf);
+        frame_size += fr->padding;
+        break;
+    }
+    s->frame_size = frame_size;
+    s->sample_rate = sample_rate;
+
+#if 0
+    printf("layer%d, %d Hz, %d kbits/s, %s\n",
+           s->layer, s->sample_rate, s->bit_rate, fr->stereo ? "stereo" : "mono");
+#endif
+}
+
+static int mp_decode_frame(MPADecodeContext *s, 
+                           short *samples)
+{
+    int nb_bytes;
+    
+    init_get_bits(&s->gb, s->inbuf + HEADER_SIZE, s->inbuf_ptr - s->inbuf - HEADER_SIZE);
+    
+    /* skip error protection field */
+    if (s->error_protection)
+        get_bits(&s->gb, 16);
+
+    /* XXX: horrible: global! */
+    gmp = &s->mpstr;
+    gmp_s = s;
+    gmp_gb = &s->gb;
+
+    nb_bytes = 0;
+    switch(s->layer) {
+    case 1:
+        do_layer1(&s->mpstr.fr,(unsigned char *)samples, &nb_bytes);
+        break;
+    case 2:
+        do_layer2(&s->mpstr.fr,(unsigned char *)samples, &nb_bytes);
+        break;
+    case 3:
+        do_layer3(&s->mpstr.fr,(unsigned char *)samples, &nb_bytes);
+        s->inbuf_index ^= 1;
+        s->inbuf = &s->inbuf1[s->inbuf_index][BACKSTEP_SIZE];
+        s->old_frame_size = s->frame_size;
+        break;
+    default:
+        break;
+    }
+    return nb_bytes;
+}
+
+/*
+ * seek back in the stream for backstep bytes (at most 511 bytes, and
+ * at most in last frame). Note that this is slightly incorrect (data
+ * can span more than one block!)  
+ */
+int set_pointer(long backstep)
+{
+    UINT8 *ptr;
+
+    /* compute current position in stream */
+    ptr = gmp_gb->buf_ptr - (gmp_gb->bit_cnt >> 3);
+    /* copy old data before current one */
+    ptr -= backstep;
+    memcpy(ptr, gmp_s->inbuf1[gmp_s->inbuf_index ^ 1] + 
+           BACKSTEP_SIZE + gmp_s->old_frame_size - backstep, backstep);
+    /* init get bits again */
+    init_get_bits(gmp_gb, ptr, gmp_s->frame_size + backstep);
+
+    return 0;
+}
+
+static int decode_frame(AVCodecContext * avctx,
+			void *data, int *data_size,
+			UINT8 * buf, int buf_size)
+{
+    MPADecodeContext *s = avctx->priv_data;
+    UINT32 header;
+    UINT8 *buf_ptr;
+    int len, out_size;
+    short *out_samples = data;
+
+    *data_size = 0;
+    buf_ptr = buf;
+    while (buf_size > 0) {
+	len = s->inbuf_ptr - s->inbuf;
+	if (s->frame_size == 0) {
+	    /* no header seen : find one. We need at least 7 bytes to parse it */
+	    len = HEADER_SIZE - len;
+	    if (len > buf_size)
+		len = buf_size;
+	    memcpy(s->inbuf_ptr, buf_ptr, len);
+	    buf_ptr += len;
+	    s->inbuf_ptr += len;
+	    buf_size -= len;
+	    if ((s->inbuf_ptr - s->inbuf) == HEADER_SIZE) {
+		header = (s->inbuf[0] << 24) | (s->inbuf[1] << 16) |
+		    (s->inbuf[2] << 8) | s->inbuf[3];
+		if (check_header(header) < 0) {
+		    /* no sync found : move by one byte (inefficient, but simple!) */
+		    memcpy(s->inbuf, s->inbuf + 1, HEADER_SIZE - 1);
+		    s->inbuf_ptr--;
+		} else {
+		    decode_header(s, header);
+		    /* update codec info */
+		    avctx->sample_rate = s->sample_rate;
+                    avctx->channels = s->mpstr.fr.stereo ? 2 : 1;
+		    avctx->bit_rate = s->bit_rate;
+		}
+	    }
+	} else if (len < s->frame_size) {
+	    len = s->frame_size - len;
+	    if (len > buf_size)
+		len = buf_size;
+
+	    memcpy(s->inbuf_ptr, buf_ptr, len);
+	    buf_ptr += len;
+	    s->inbuf_ptr += len;
+	    buf_size -= len;
+	} else {
+            out_size = mp_decode_frame(s, out_samples);
+	    s->inbuf_ptr = s->inbuf;
+	    s->frame_size = 0;
+	    *data_size = out_size;
+	    break;
+	}
+    }
+    return buf_ptr - buf;
+}
+
+AVCodec mp3_decoder =
+{
+    "mpegaudio",
+    CODEC_TYPE_AUDIO,
+    CODEC_ID_MP2,
+    sizeof(MPADecodeContext),
+    decode_init,
+    NULL,
+    NULL,
+    decode_frame,
+};
diff --git a/libavcodec/mpegaudiotab.h b/libavcodec/mpegaudiotab.h
new file mode 100644
index 0000000000..05bdb9eea1
--- /dev/null
+++ b/libavcodec/mpegaudiotab.h
@@ -0,0 +1,310 @@
+/*
+ * mpeg audio layer 2 tables. Most of them come from the mpeg audio
+ * specification.
+ * 
+ * Copyright (c) 2000 Gerard Lantau.
+ *
+ * The licence of this code is contained in file LICENCE found in the
+ * same archive 
+ */
+
+static const unsigned short bitrate_tab[2][15] = {
+    {0,8,16,24,32,40,48,56,64,80,96,112,128,144,160}, /* mpeg2 lsf */
+    {0,32,48,56,64,80,96,112,128,160,192,224,256,320,384}, /* mpeg1 */
+};
+
+static const unsigned short freq_tab[3] = { 44100, 48000, 32000 };
+
+#define SQRT2 1.41421356237309514547
+
+static const int costab32[30] = {
+    FIX(0.54119610014619701222),
+    FIX(1.3065629648763763537),
+    
+    FIX(0.50979557910415917998),
+    FIX(2.5629154477415054814),
+    FIX(0.89997622313641556513),
+    FIX(0.60134488693504528634),
+    
+    FIX(0.5024192861881556782),
+    FIX(5.1011486186891552563),
+    FIX(0.78815462345125020249),
+    FIX(0.64682178335999007679),
+    FIX(0.56694403481635768927),
+    FIX(1.0606776859903470633),
+    FIX(1.7224470982383341955),
+    FIX(0.52249861493968885462),
+    
+    FIX(10.19000812354803287),
+    FIX(0.674808341455005678),
+    FIX(1.1694399334328846596),
+    FIX(0.53104259108978413284),
+    FIX(2.0577810099534108446),
+    FIX(0.58293496820613388554),
+    FIX(0.83934964541552681272),
+    FIX(0.50547095989754364798),
+    FIX(3.4076084184687189804),
+    FIX(0.62250412303566482475),
+    FIX(0.97256823786196078263),
+    FIX(0.51544730992262455249),
+    FIX(1.4841646163141661852),
+    FIX(0.5531038960344445421),
+    FIX(0.74453627100229857749),
+    FIX(0.5006029982351962726),
+};
+
+static const int bitinv32[32] = {
+    0,  16,  8, 24,  4,  20,  12,  28,
+    2,  18, 10, 26,  6,  22,  14,  30,
+    1,  17,  9, 25,  5,  21,  13,  29,
+    3,  19, 11, 27,  7,  23,  15,  31
+};
+
+
+static short filter_bank[512];
+
+static const double enwindow[512] = {0.000000000,
+                        -0.000000477, -0.000000477, -0.000000477, -0.000000477, -0.000000477, -0.000000477, -0.000000954, -0.000000954,
+                        -0.000000954, -0.000000954, -0.000001431, -0.000001431, -0.000001907, -0.000001907, -0.000002384, -0.000002384,
+                        -0.000002861, -0.000003338, -0.000003338, -0.000003815, -0.000004292, -0.000004768, -0.000005245, -0.000006199,
+                        -0.000006676, -0.000007629, -0.000008106, -0.000009060, -0.000010014, -0.000011444, -0.000012398, -0.000013828,
+                        -0.000014782, -0.000016689, -0.000018120, -0.000019550, -0.000021458, -0.000023365, -0.000025272, -0.000027657,
+                        -0.000030041, -0.000032425, -0.000034809, -0.000037670, -0.000040531, -0.000043392, -0.000046253, -0.000049591,
+                        -0.000052929, -0.000055790, -0.000059605, -0.000062943, -0.000066280, -0.000070095, -0.000073433, -0.000076771,
+                        -0.000080585, -0.000083923, -0.000087261, -0.000090599, -0.000093460, -0.000096321, -0.000099182, 0.000101566,
+                        0.000103951, 0.000105858, 0.000107288, 0.000108242, 0.000108719, 0.000108719, 0.000108242, 0.000106812,
+                        0.000105381, 0.000102520, 0.000099182, 0.000095367, 0.000090122, 0.000084400, 0.000077724, 0.000069618,
+                        0.000060558, 0.000050545, 0.000039577, 0.000027180, 0.000013828, -0.000000954, -0.000017166, -0.000034332,
+                        -0.000052929, -0.000072956, -0.000093937, -0.000116348, -0.000140190, -0.000165462, -0.000191212, -0.000218868,
+                        -0.000247478, -0.000277042, -0.000307560, -0.000339031, -0.000371456, -0.000404358, -0.000438213, -0.000472546,
+                        -0.000507355, -0.000542164, -0.000576973, -0.000611782, -0.000646591, -0.000680923, -0.000714302, -0.000747204,
+                        -0.000779152, -0.000809669, -0.000838757, -0.000866413, -0.000891685, -0.000915051, -0.000935555, -0.000954151,
+                        -0.000968933, -0.000980854, -0.000989437, -0.000994205, -0.000995159, -0.000991821, -0.000983715, 0.000971317,
+                        0.000953674, 0.000930786, 0.000902653, 0.000868797, 0.000829220, 0.000783920, 0.000731945, 0.000674248,
+                        0.000610352, 0.000539303, 0.000462532, 0.000378609, 0.000288486, 0.000191689, 0.000088215, -0.000021458,
+                        -0.000137329, -0.000259876, -0.000388145, -0.000522137, -0.000661850, -0.000806808, -0.000956535, -0.001111031,
+                        -0.001269817, -0.001432419, -0.001597881, -0.001766682, -0.001937389, -0.002110004, -0.002283096, -0.002457142,
+                        -0.002630711, -0.002803326, -0.002974033, -0.003141880, -0.003306866, -0.003467083, -0.003622532, -0.003771782,
+                        -0.003914356, -0.004048824, -0.004174709, -0.004290581, -0.004395962, -0.004489899, -0.004570484, -0.004638195,
+                        -0.004691124, -0.004728317, -0.004748821, -0.004752159, -0.004737377, -0.004703045, -0.004649162, -0.004573822,
+                        -0.004477024, -0.004357815, -0.004215240, -0.004049301, -0.003858566, -0.003643036, -0.003401756, 0.003134727,
+                        0.002841473, 0.002521515, 0.002174854, 0.001800537, 0.001399517, 0.000971317, 0.000515938, 0.000033379,
+                        -0.000475883, -0.001011848, -0.001573563, -0.002161503, -0.002774239, -0.003411293, -0.004072189, -0.004756451,
+                        -0.005462170, -0.006189346, -0.006937027, -0.007703304, -0.008487225, -0.009287834, -0.010103703, -0.010933399,
+                        -0.011775017, -0.012627602, -0.013489246, -0.014358521, -0.015233517, -0.016112804, -0.016994476, -0.017876148,
+                        -0.018756866, -0.019634247, -0.020506859, -0.021372318, -0.022228718, -0.023074150, -0.023907185, -0.024725437,
+                        -0.025527000, -0.026310921, -0.027073860, -0.027815342, -0.028532982, -0.029224873, -0.029890060, -0.030526638,
+                        -0.031132698, -0.031706810, -0.032248020, -0.032754898, -0.033225536, -0.033659935, -0.034055710, -0.034412861,
+                        -0.034730434, -0.035007000, -0.035242081, -0.035435200, -0.035586357, -0.035694122, -0.035758972, 0.035780907,
+                        0.035758972, 0.035694122, 0.035586357, 0.035435200, 0.035242081, 0.035007000, 0.034730434, 0.034412861,
+                        0.034055710, 0.033659935, 0.033225536, 0.032754898, 0.032248020, 0.031706810, 0.031132698, 0.030526638,
+                        0.029890060, 0.029224873, 0.028532982, 0.027815342, 0.027073860, 0.026310921, 0.025527000, 0.024725437,
+                        0.023907185, 0.023074150, 0.022228718, 0.021372318, 0.020506859, 0.019634247, 0.018756866, 0.017876148,
+                        0.016994476, 0.016112804, 0.015233517, 0.014358521, 0.013489246, 0.012627602, 0.011775017, 0.010933399,
+                        0.010103703, 0.009287834, 0.008487225, 0.007703304, 0.006937027, 0.006189346, 0.005462170, 0.004756451,
+                        0.004072189, 0.003411293, 0.002774239, 0.002161503, 0.001573563, 0.001011848, 0.000475883, -0.000033379,
+                        -0.000515938, -0.000971317, -0.001399517, -0.001800537, -0.002174854, -0.002521515, -0.002841473, 0.003134727,
+                        0.003401756, 0.003643036, 0.003858566, 0.004049301, 0.004215240, 0.004357815, 0.004477024, 0.004573822,
+                        0.004649162, 0.004703045, 0.004737377, 0.004752159, 0.004748821, 0.004728317, 0.004691124, 0.004638195,
+                        0.004570484, 0.004489899, 0.004395962, 0.004290581, 0.004174709, 0.004048824, 0.003914356, 0.003771782,
+                        0.003622532, 0.003467083, 0.003306866, 0.003141880, 0.002974033, 0.002803326, 0.002630711, 0.002457142,
+                        0.002283096, 0.002110004, 0.001937389, 0.001766682, 0.001597881, 0.001432419, 0.001269817, 0.001111031,
+                        0.000956535, 0.000806808, 0.000661850, 0.000522137, 0.000388145, 0.000259876, 0.000137329, 0.000021458,
+                        -0.000088215, -0.000191689, -0.000288486, -0.000378609, -0.000462532, -0.000539303, -0.000610352, -0.000674248,
+                        -0.000731945, -0.000783920, -0.000829220, -0.000868797, -0.000902653, -0.000930786, -0.000953674, 0.000971317,
+                        0.000983715, 0.000991821, 0.000995159, 0.000994205, 0.000989437, 0.000980854, 0.000968933, 0.000954151,
+                        0.000935555, 0.000915051, 0.000891685, 0.000866413, 0.000838757, 0.000809669, 0.000779152, 0.000747204,
+                        0.000714302, 0.000680923, 0.000646591, 0.000611782, 0.000576973, 0.000542164, 0.000507355, 0.000472546,
+                        0.000438213, 0.000404358, 0.000371456, 0.000339031, 0.000307560, 0.000277042, 0.000247478, 0.000218868,
+                        0.000191212, 0.000165462, 0.000140190, 0.000116348, 0.000093937, 0.000072956, 0.000052929, 0.000034332,
+                        0.000017166, 0.000000954, -0.000013828, -0.000027180, -0.000039577, -0.000050545, -0.000060558, -0.000069618,
+                        -0.000077724, -0.000084400, -0.000090122, -0.000095367, -0.000099182, -0.000102520, -0.000105381, -0.000106812,
+                        -0.000108242, -0.000108719, -0.000108719, -0.000108242, -0.000107288, -0.000105858, -0.000103951, 0.000101566,
+                        0.000099182, 0.000096321, 0.000093460, 0.000090599, 0.000087261, 0.000083923, 0.000080585, 0.000076771,
+                        0.000073433, 0.000070095, 0.000066280, 0.000062943, 0.000059605, 0.000055790, 0.000052929, 0.000049591,
+                        0.000046253, 0.000043392, 0.000040531, 0.000037670, 0.000034809, 0.000032425, 0.000030041, 0.000027657,
+                        0.000025272, 0.000023365, 0.000021458, 0.000019550, 0.000018120, 0.000016689, 0.000014782, 0.000013828,
+                        0.000012398, 0.000011444, 0.000010014, 0.000009060, 0.000008106, 0.000007629, 0.000006676, 0.000006199,
+                        0.000005245, 0.000004768, 0.000004292, 0.000003815, 0.000003338, 0.000003338, 0.000002861, 0.000002384,
+                        0.000002384, 0.000001907, 0.000001907, 0.000001431, 0.000001431, 0.000000954, 0.000000954, 0.000000954,
+                        0.000000954, 0.000000477, 0.000000477, 0.000000477, 0.000000477, 0.000000477, 0.000000477
+                       };
+
+static int scale_factor_table[64];
+#ifdef USE_FLOATS
+static float scale_factor_inv_table[64];
+#else
+static INT8 scale_factor_shift[64];
+static unsigned short scale_factor_mult[64];
+#endif
+static unsigned char scale_diff_table[128];
+
+static const int sblimit_table[5] = { 27 , 30 , 8, 12 , 30 };
+
+static const int quant_steps[17] = {
+    3,     5,    7,    9,    15,
+    31,    63,  127,  255,   511,
+    1023,  2047, 4095, 8191, 16383,
+    32767, 65535
+};
+
+/* we use a negative value if grouped */
+static const int quant_bits[17] = {
+    -5,  -7,  3, -10, 4, 
+     5,  6,  7,  8,  9,
+    10, 11, 12, 13, 14,
+    15, 16 
+};
+
+/* signal to noise ratio of each quantification step (could be
+   computed from quant_steps[]). The values are dB multiplied by 10 
+*/
+static unsigned short quant_snr[17] = { 
+     70, 110, 160, 208,
+    253, 316, 378, 439,
+    499, 559, 620, 680, 
+    740, 800, 861, 920, 
+    980
+};
+
+
+/* total number of bits per allocation group */
+static unsigned short total_quant_bits[17];
+
+/* encoding tables which give the quantization index. Note how it is
+   possible to store them efficiently ! */
+static const unsigned char alloc_table_0[] = {
+ 4,  0,  2,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 
+ 4,  0,  2,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 
+ 4,  0,  2,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 16, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 16, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 16, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 16, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 16, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 16, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 16, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 2,  0,  1, 16, 
+ 2,  0,  1, 16, 
+ 2,  0,  1, 16, 
+ 2,  0,  1, 16, 
+};
+
+static const unsigned char alloc_table_1[] = {
+ 4,  0,  2,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 
+ 4,  0,  2,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 
+ 4,  0,  2,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 16, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 16, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 16, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 16, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 16, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 16, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 16, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 3,  0,  1,  2,  3,  4,  5, 16, 
+ 2,  0,  1, 16, 
+ 2,  0,  1, 16, 
+ 2,  0,  1, 16, 
+ 2,  0,  1, 16, 
+ 2,  0,  1, 16, 
+ 2,  0,  1, 16, 
+ 2,  0,  1, 16, 
+};
+
+static const unsigned char alloc_table_2[] = {
+ 4,  0,  1,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 
+ 4,  0,  1,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+};
+
+static const unsigned char alloc_table_3[] = {
+ 4,  0,  1,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 
+ 4,  0,  1,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+};
+
+static const unsigned char alloc_table_4[] = {
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14,
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 
+ 4,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 3,  0,  1,  3,  4,  5,  6,  7, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+ 2,  0,  1,  3, 
+};
+
+const unsigned char *alloc_tables[5] = 
+{ alloc_table_0, alloc_table_1, alloc_table_2, alloc_table_3, alloc_table_4, };
+
+/* fixed psycho acoustic model. Values of SNR taken from the 'toolame'
+   project */
+const float fixed_smr[SBLIMIT] =  {
+    30, 17, 16, 10, 3, 12, 8, 2.5,
+    5, 5, 6, 6, 5, 6, 10, 6,
+    -4, -10, -21, -30, -42, -55, -68, -75,
+    -75, -75, -75, -75, -91, -107, -110, -108
+};
+
+const unsigned char nb_scale_factors[4] = { 3, 2, 1, 2 };
diff --git a/libavcodec/mpegvideo.c b/libavcodec/mpegvideo.c
new file mode 100644
index 0000000000..467062ddf6
--- /dev/null
+++ b/libavcodec/mpegvideo.c
@@ -0,0 +1,1281 @@
+/*
+ * The simplest mpeg encoder (well, it was the simplest!)
+ * Copyright (c) 2000,2001 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include <math.h>
+#include <string.h>
+#include "avcodec.h"
+#include "dsputil.h"
+#include "mpegvideo.h"
+
+#define EDGE_WIDTH 16
+
+/* enable all paranoid tests for rounding, overflows, etc... */
+//#define PARANOID
+
+//#define DEBUG
+
+/* for jpeg fast DCT */
+#define CONST_BITS 14
+
+static const unsigned short aanscales[64] = {
+    /* precomputed values scaled up by 14 bits */
+    16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+    22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
+    21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
+    19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
+    16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+    12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
+    8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
+    4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
+};
+
+static UINT8 h263_chroma_roundtab[16] = {
+    0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2,
+};
+
+static void encode_picture(MpegEncContext *s, int picture_number);
+static void rate_control_init(MpegEncContext *s);
+static int rate_estimate_qscale(MpegEncContext *s);
+
+/* default motion estimation */
+int motion_estimation_method = ME_LOG;
+
+/* XXX: should use variable shift ? */
+#define QMAT_SHIFT_MMX 19
+#define QMAT_SHIFT 25
+
+static void convert_matrix(int *qmat, const UINT16 *quant_matrix, int qscale)
+{
+    int i;
+
+    if (av_fdct == jpeg_fdct_ifast) {
+        for(i=0;i<64;i++) {
+            /* 16 <= qscale * quant_matrix[i] <= 7905 */
+            /* 19952 <= aanscales[i] * qscale * quant_matrix[i] <= 249205026 */
+            
+            qmat[i] = (int)((1ULL << (QMAT_SHIFT + 11)) / (aanscales[i] * qscale * quant_matrix[i]));
+        }
+    } else {
+        for(i=0;i<64;i++) {
+            /* We can safely suppose that 16 <= quant_matrix[i] <= 255
+               So 16 <= qscale * quant_matrix[i] <= 7905
+               so (1 << QMAT_SHIFT) / 16 >= qmat[i] >= (1 << QMAT_SHIFT) / 7905
+            */
+            qmat[i] = (1 << QMAT_SHIFT_MMX) / (qscale * quant_matrix[i]);
+        }
+    }
+}
+
+/* init common structure for both encoder and decoder */
+int MPV_common_init(MpegEncContext *s)
+{
+    int c_size, i;
+    UINT8 *pict;
+
+    s->mb_width = (s->width + 15) / 16;
+    s->mb_height = (s->height + 15) / 16;
+    s->linesize = s->mb_width * 16 + 2 * EDGE_WIDTH;
+
+    for(i=0;i<3;i++) {
+        int w, h, shift, pict_start;
+
+        w = s->linesize;
+        h = s->mb_height * 16 + 2 * EDGE_WIDTH;
+        shift = (i == 0) ? 0 : 1;
+        c_size = (w >> shift) * (h >> shift);
+        pict_start = (w >> shift) * (EDGE_WIDTH >> shift) + (EDGE_WIDTH >> shift);
+
+        pict = av_mallocz(c_size);
+        if (pict == NULL)
+            goto fail;
+        s->last_picture_base[i] = pict;
+        s->last_picture[i] = pict + pict_start;
+    
+        pict = av_mallocz(c_size);
+        if (pict == NULL)
+            goto fail;
+        s->next_picture_base[i] = pict;
+        s->next_picture[i] = pict + pict_start;
+
+        if (s->has_b_frames) {
+            pict = av_mallocz(c_size);
+            if (pict == NULL) 
+                goto fail;
+            s->aux_picture_base[i] = pict;
+            s->aux_picture[i] = pict + pict_start;
+        }
+    }
+
+    if (s->out_format == FMT_H263) {
+        int size;
+        /* MV prediction */
+        size = (2 * s->mb_width + 2) * (2 * s->mb_height + 2);
+        s->motion_val = malloc(size * 2 * sizeof(INT16));
+        if (s->motion_val == NULL)
+            goto fail;
+        memset(s->motion_val, 0, size * 2 * sizeof(INT16));
+    }
+
+    if (s->h263_pred) {
+        int y_size, c_size, i, size;
+        
+        /* dc values */
+
+        y_size = (2 * s->mb_width + 2) * (2 * s->mb_height + 2);
+        c_size = (s->mb_width + 2) * (s->mb_height + 2);
+        size = y_size + 2 * c_size;
+        s->dc_val[0] = malloc(size * sizeof(INT16));
+        if (s->dc_val[0] == NULL)
+            goto fail;
+        s->dc_val[1] = s->dc_val[0] + y_size;
+        s->dc_val[2] = s->dc_val[1] + c_size;
+        for(i=0;i<size;i++)
+            s->dc_val[0][i] = 1024;
+
+        /* ac values */
+        s->ac_val[0] = av_mallocz(size * sizeof(INT16) * 16);
+        if (s->ac_val[0] == NULL)
+            goto fail;
+        s->ac_val[1] = s->ac_val[0] + y_size;
+        s->ac_val[2] = s->ac_val[1] + c_size;
+        
+        /* cbp values */
+        s->coded_block = av_mallocz(y_size);
+        if (!s->coded_block)
+            goto fail;
+    }
+    /* default structure is frame */
+    s->picture_structure = PICT_FRAME;
+
+    /* init default q matrix (only for mpeg and mjpeg) */
+    for(i=0;i<64;i++) {
+        s->intra_matrix[i] = default_intra_matrix[i];
+        s->chroma_intra_matrix[i] = default_intra_matrix[i];
+        s->non_intra_matrix[i] = default_non_intra_matrix[i];
+        s->chroma_non_intra_matrix[i] = default_non_intra_matrix[i];
+    }
+    s->context_initialized = 1;
+    return 0;
+ fail:
+    if (s->motion_val)
+        free(s->motion_val);
+    if (s->dc_val[0])
+        free(s->dc_val[0]);
+    if (s->ac_val[0])
+        free(s->ac_val[0]);
+    if (s->coded_block)
+        free(s->coded_block);
+    for(i=0;i<3;i++) {
+        if (s->last_picture_base[i])
+            free(s->last_picture_base[i]);
+        if (s->next_picture_base[i])
+            free(s->next_picture_base[i]);
+        if (s->aux_picture_base[i])
+            free(s->aux_picture_base[i]);
+    }
+    return -1;
+}
+
+/* init common structure for both encoder and decoder */
+void MPV_common_end(MpegEncContext *s)
+{
+    int i;
+
+    if (s->motion_val)
+        free(s->motion_val);
+    if (s->h263_pred) {
+        free(s->dc_val[0]);
+        free(s->ac_val[0]);
+        free(s->coded_block);
+    }
+    for(i=0;i<3;i++) {
+        free(s->last_picture_base[i]);
+        free(s->next_picture_base[i]);
+        if (s->has_b_frames)
+            free(s->aux_picture_base[i]);
+    }
+    s->context_initialized = 0;
+}
+
+/* init video encoder */
+int MPV_encode_init(AVCodecContext *avctx)
+{
+    MpegEncContext *s = avctx->priv_data;
+
+    s->bit_rate = avctx->bit_rate;
+    s->frame_rate = avctx->frame_rate;
+    s->width = avctx->width;
+    s->height = avctx->height;
+    s->gop_size = avctx->gop_size;
+    if (s->gop_size <= 1) {
+        s->intra_only = 1;
+        s->gop_size = 12;
+    } else {
+        s->intra_only = 0;
+    }
+    s->full_search = motion_estimation_method;
+
+    s->fixed_qscale = (avctx->flags & CODEC_FLAG_QSCALE);
+
+    switch(avctx->codec->id) {
+    case CODEC_ID_MPEG1VIDEO:
+        s->out_format = FMT_MPEG1;
+        break;
+    case CODEC_ID_MJPEG:
+        s->out_format = FMT_MJPEG;
+        s->intra_only = 1; /* force intra only for jpeg */
+        if (mjpeg_init(s) < 0)
+            return -1;
+        break;
+    case CODEC_ID_H263:
+        if (h263_get_picture_format(s->width, s->height) == 7)
+            return -1;
+        s->out_format = FMT_H263;
+        break;
+    case CODEC_ID_H263P:
+        s->out_format = FMT_H263;
+        s->h263_plus = 1;
+        /* XXX: not unrectricted mv yet */
+        break;
+    case CODEC_ID_RV10:
+        s->out_format = FMT_H263;
+        s->h263_rv10 = 1;
+        break;
+    case CODEC_ID_OPENDIVX:
+        s->out_format = FMT_H263;
+        s->h263_pred = 1;
+        s->unrestricted_mv = 1;
+        break;
+    case CODEC_ID_MSMPEG4:
+        s->out_format = FMT_H263;
+        s->h263_msmpeg4 = 1;
+        s->h263_pred = 1;
+        s->unrestricted_mv = 1;
+        break;
+    default:
+        return -1;
+    }
+
+    if (s->out_format == FMT_H263)
+        h263_encode_init_vlc(s);
+
+    /* init */
+    if (MPV_common_init(s) < 0)
+        return -1;
+    
+    /* rate control init */
+    rate_control_init(s);
+
+    s->picture_number = 0;
+    s->fake_picture_number = 0;
+    /* motion detector init */
+    s->f_code = 1;
+
+    return 0;
+}
+
+int MPV_encode_end(AVCodecContext *avctx)
+{
+    MpegEncContext *s = avctx->priv_data;
+
+#ifdef STATS
+    print_stats();
+#endif
+    MPV_common_end(s);
+    if (s->out_format == FMT_MJPEG)
+        mjpeg_close(s);
+    return 0;
+}
+
+/* draw the edges of width 'w' of an image of size width, height */
+static void draw_edges(UINT8 *buf, int wrap, int width, int height, int w)
+{
+    UINT8 *ptr, *last_line;
+    int i;
+
+    last_line = buf + (height - 1) * wrap;
+    for(i=0;i<w;i++) {
+        /* top and bottom */
+        memcpy(buf - (i + 1) * wrap, buf, width);
+        memcpy(last_line + (i + 1) * wrap, last_line, width);
+    }
+    /* left and right */
+    ptr = buf;
+    for(i=0;i<height;i++) {
+        memset(ptr - w, ptr[0], w);
+        memset(ptr + width, ptr[width-1], w);
+        ptr += wrap;
+    }
+    /* corners */
+    for(i=0;i<w;i++) {
+        memset(buf - (i + 1) * wrap - w, buf[0], w); /* top left */
+        memset(buf - (i + 1) * wrap + width, buf[width-1], w); /* top right */
+        memset(last_line + (i + 1) * wrap - w, last_line[0], w); /* top left */
+        memset(last_line + (i + 1) * wrap + width, last_line[width-1], w); /* top right */
+    }
+}
+
+
+/* generic function for encode/decode called before a frame is coded/decoded */
+void MPV_frame_start(MpegEncContext *s)
+{
+    int i;
+    UINT8 *tmp;
+
+    if (s->pict_type == B_TYPE) {
+        for(i=0;i<3;i++) {
+            s->current_picture[i] = s->aux_picture[i];
+        }
+    } else {
+        for(i=0;i<3;i++) {
+            /* swap next and last */
+            tmp = s->last_picture[i];
+            s->last_picture[i] = s->next_picture[i];
+            s->next_picture[i] = tmp;
+            s->current_picture[i] = tmp;
+        }
+    }
+}
+
+/* generic function for encode/decode called after a frame has been coded/decoded */
+void MPV_frame_end(MpegEncContext *s)
+{
+    /* draw edge for correct motion prediction if outside */
+    if (s->pict_type != B_TYPE) {
+        draw_edges(s->current_picture[0], s->linesize, s->width, s->height, EDGE_WIDTH);
+        draw_edges(s->current_picture[1], s->linesize/2, s->width/2, s->height/2, EDGE_WIDTH/2);
+        draw_edges(s->current_picture[2], s->linesize/2, s->width/2, s->height/2, EDGE_WIDTH/2);
+    }
+}
+
+int MPV_encode_picture(AVCodecContext *avctx,
+                       unsigned char *buf, int buf_size, void *data)
+{
+    MpegEncContext *s = avctx->priv_data;
+    AVPicture *pict = data;
+    int i, j;
+
+    if (s->fixed_qscale) 
+        s->qscale = avctx->quality;
+
+    init_put_bits(&s->pb, buf, buf_size, NULL, NULL);
+
+    if (!s->intra_only) {
+        /* first picture of GOP is intra */
+        if ((s->picture_number % s->gop_size) == 0)
+            s->pict_type = I_TYPE;
+        else
+            s->pict_type = P_TYPE;
+    } else {
+        s->pict_type = I_TYPE;
+    }
+    avctx->key_frame = (s->pict_type == I_TYPE);
+    
+    MPV_frame_start(s);
+
+    for(i=0;i<3;i++) {
+        UINT8 *src = pict->data[i];
+        UINT8 *dest = s->current_picture[i];
+        int src_wrap = pict->linesize[i];
+        int dest_wrap = s->linesize;
+        int w = s->width;
+        int h = s->height;
+
+        if (i >= 1) {
+            dest_wrap >>= 1;
+            w >>= 1;
+            h >>= 1;
+        }
+
+        for(j=0;j<h;j++) {
+            memcpy(dest, src, w);
+            dest += dest_wrap;
+            src += src_wrap;
+        }
+        s->new_picture[i] = s->current_picture[i];
+    }
+
+    encode_picture(s, s->picture_number);
+    
+    MPV_frame_end(s);
+    s->picture_number++;
+
+    if (s->out_format == FMT_MJPEG)
+        mjpeg_picture_trailer(s);
+
+    flush_put_bits(&s->pb);
+    s->total_bits += (s->pb.buf_ptr - s->pb.buf) * 8;
+    avctx->quality = s->qscale;
+    return s->pb.buf_ptr - s->pb.buf;
+}
+
+static inline int clip(int a, int amin, int amax)
+{
+    if (a < amin)
+        return amin;
+    else if (a > amax)
+        return amax;
+    else
+        return a;
+}
+
+static int dct_quantize(MpegEncContext *s, DCTELEM *block, int n, int qscale);
+static int dct_quantize_mmx(MpegEncContext *s, 
+                            DCTELEM *block, int n,
+                            int qscale);
+static void dct_unquantize(MpegEncContext *s, DCTELEM *block, int n, int qscale);
+
+/* apply one mpeg motion vector to the three components */
+static inline void mpeg_motion(MpegEncContext *s,
+                               UINT8 *dest_y, UINT8 *dest_cb, UINT8 *dest_cr,
+                               int dest_offset,
+                               UINT8 **ref_picture, int src_offset,
+                               int field_based, op_pixels_func *pix_op,
+                               int motion_x, int motion_y, int h)
+{
+    UINT8 *ptr;
+    int dxy, offset, mx, my, src_x, src_y, height, linesize;
+    
+    dxy = ((motion_y & 1) << 1) | (motion_x & 1);
+    src_x = s->mb_x * 16 + (motion_x >> 1);
+    src_y = s->mb_y * (16 >> field_based) + (motion_y >> 1);
+                
+    /* WARNING: do no forget half pels */
+    height = s->height >> field_based;
+    src_x = clip(src_x, -16, s->width);
+    if (src_x == s->width)
+        dxy &= ~1;
+    src_y = clip(src_y, -16, height);
+    if (src_y == height)
+        dxy &= ~2;
+    linesize = s->linesize << field_based;
+    ptr = ref_picture[0] + (src_y * linesize) + (src_x) + src_offset;
+    dest_y += dest_offset;
+    pix_op[dxy](dest_y, ptr, linesize, h);
+    pix_op[dxy](dest_y + 8, ptr + 8, linesize, h);
+
+    if (s->out_format == FMT_H263) {
+        dxy = 0;
+        if ((motion_x & 3) != 0)
+            dxy |= 1;
+        if ((motion_y & 3) != 0)
+            dxy |= 2;
+        mx = motion_x >> 2;
+        my = motion_y >> 2;
+    } else {
+        mx = motion_x / 2;
+        my = motion_y / 2;
+        dxy = ((my & 1) << 1) | (mx & 1);
+        mx >>= 1;
+        my >>= 1;
+    }
+    
+    src_x = s->mb_x * 8 + mx;
+    src_y = s->mb_y * (8 >> field_based) + my;
+    src_x = clip(src_x, -8, s->width >> 1);
+    if (src_x == (s->width >> 1))
+        dxy &= ~1;
+    src_y = clip(src_y, -8, height >> 1);
+    if (src_y == (height >> 1))
+        dxy &= ~2;
+
+    offset = (src_y * (linesize >> 1)) + src_x + (src_offset >> 1);
+    ptr = ref_picture[1] + offset;
+    pix_op[dxy](dest_cb + (dest_offset >> 1), ptr, linesize >> 1, h >> 1);
+    ptr = ref_picture[2] + offset;
+    pix_op[dxy](dest_cr + (dest_offset >> 1), ptr, linesize >> 1, h >> 1);
+}
+
+static inline void MPV_motion(MpegEncContext *s, 
+                              UINT8 *dest_y, UINT8 *dest_cb, UINT8 *dest_cr,
+                              int dir, UINT8 **ref_picture, 
+                              op_pixels_func *pix_op)
+{
+    int dxy, offset, mx, my, src_x, src_y, motion_x, motion_y;
+    int mb_x, mb_y, i;
+    UINT8 *ptr, *dest;
+
+    mb_x = s->mb_x;
+    mb_y = s->mb_y;
+
+    switch(s->mv_type) {
+    case MV_TYPE_16X16:
+        mpeg_motion(s, dest_y, dest_cb, dest_cr, 0,
+                    ref_picture, 0,
+                    0, pix_op,
+                    s->mv[dir][0][0], s->mv[dir][0][1], 16);
+        break;
+    case MV_TYPE_8X8:
+        for(i=0;i<4;i++) {
+            motion_x = s->mv[dir][i][0];
+            motion_y = s->mv[dir][i][1];
+
+            dxy = ((motion_y & 1) << 1) | (motion_x & 1);
+            src_x = mb_x * 16 + (motion_x >> 1) + (i & 1) * 8;
+            src_y = mb_y * 16 + (motion_y >> 1) + ((i >> 1) & 1) * 8;
+                    
+            /* WARNING: do no forget half pels */
+            src_x = clip(src_x, -16, s->width);
+            if (src_x == s->width)
+                dxy &= ~1;
+            src_y = clip(src_y, -16, s->height);
+            if (src_y == s->height)
+                dxy &= ~2;
+                    
+            ptr = ref_picture[0] + (src_y * s->linesize) + (src_x);
+            dest = dest_y + ((i & 1) * 8) + (i >> 1) * 8 * s->linesize;
+            pix_op[dxy](dest, ptr, s->linesize, 8);
+        }
+        /* In case of 8X8, we construct a single chroma motion vector
+           with a special rounding */
+        mx = 0;
+        my = 0;
+        for(i=0;i<4;i++) {
+            mx += s->mv[dir][i][0];
+            my += s->mv[dir][i][1];
+        }
+        if (mx >= 0)
+            mx = (h263_chroma_roundtab[mx & 0xf] + ((mx >> 3) & ~1));
+        else {
+            mx = -mx;
+            mx = -(h263_chroma_roundtab[mx & 0xf] + ((mx >> 3) & ~1));
+        }
+        if (my >= 0)
+            my = (h263_chroma_roundtab[my & 0xf] + ((my >> 3) & ~1));
+        else {
+            my = -my;
+            my = -(h263_chroma_roundtab[my & 0xf] + ((my >> 3) & ~1));
+        }
+        dxy = ((my & 1) << 1) | (mx & 1);
+        mx >>= 1;
+        my >>= 1;
+
+        src_x = mb_x * 8 + mx;
+        src_y = mb_y * 8 + my;
+        src_x = clip(src_x, -8, s->width/2);
+        if (src_x == s->width/2)
+            dxy &= ~1;
+        src_y = clip(src_y, -8, s->height/2);
+        if (src_y == s->height/2)
+            dxy &= ~2;
+        
+        offset = (src_y * (s->linesize >> 1)) + src_x;
+        ptr = ref_picture[1] + offset;
+        pix_op[dxy](dest_cb, ptr, s->linesize >> 1, 8);
+        ptr = ref_picture[2] + offset;
+        pix_op[dxy](dest_cr, ptr, s->linesize >> 1, 8);
+        break;
+    case MV_TYPE_FIELD:
+        if (s->picture_structure == PICT_FRAME) {
+            /* top field */
+            mpeg_motion(s, dest_y, dest_cb, dest_cr, 0,
+                        ref_picture, s->field_select[dir][0] ? s->linesize : 0,
+                        1, pix_op,
+                        s->mv[dir][0][0], s->mv[dir][0][1], 8);
+            /* bottom field */
+            mpeg_motion(s, dest_y, dest_cb, dest_cr, s->linesize,
+                        ref_picture, s->field_select[dir][1] ? s->linesize : 0,
+                        1, pix_op,
+                        s->mv[dir][1][0], s->mv[dir][1][1], 8);
+        } else {
+            
+
+        }
+        break;
+    }
+}
+
+
+/* put block[] to dest[] */
+static inline void put_dct(MpegEncContext *s, 
+                           DCTELEM *block, int i, UINT8 *dest, int line_size)
+{
+    if (!s->mpeg2)
+        dct_unquantize(s, block, i, s->qscale);
+    j_rev_dct (block);
+    put_pixels_clamped(block, dest, line_size);
+}
+
+/* add block[] to dest[] */
+static inline void add_dct(MpegEncContext *s, 
+                           DCTELEM *block, int i, UINT8 *dest, int line_size)
+{
+    if (s->block_last_index[i] >= 0) {
+        if (!s->mpeg2)
+            dct_unquantize(s, block, i, s->qscale);
+        j_rev_dct (block);
+        add_pixels_clamped(block, dest, line_size);
+    }
+}
+
+/* generic function called after a macroblock has been parsed by the
+   decoder or after it has been encoded by the encoder.
+
+   Important variables used:
+   s->mb_intra : true if intra macroblock
+   s->mv_dir   : motion vector direction
+   s->mv_type  : motion vector type
+   s->mv       : motion vector
+   s->interlaced_dct : true if interlaced dct used (mpeg2)
+ */
+void MPV_decode_mb(MpegEncContext *s, DCTELEM block[6][64])
+{
+    int mb_x, mb_y, motion_x, motion_y;
+    int dct_linesize, dct_offset;
+    op_pixels_func *op_pix;
+
+    mb_x = s->mb_x;
+    mb_y = s->mb_y;
+
+    /* update DC predictors for P macroblocks */
+    if (!s->mb_intra) {
+        if (s->h263_pred) {
+            int wrap, x, y, v;
+            wrap = 2 * s->mb_width + 2;
+            v = 1024;
+            x = 2 * mb_x + 1;
+            y = 2 * mb_y + 1;
+            s->dc_val[0][(x) + (y) * wrap] = v;
+            s->dc_val[0][(x + 1) + (y) * wrap] = v;
+            s->dc_val[0][(x) + (y + 1) * wrap] = v;
+            s->dc_val[0][(x + 1) + (y + 1) * wrap] = v;
+            /* ac pred */
+            memset(s->ac_val[0][(x) + (y) * wrap], 0, 16 * sizeof(INT16));
+            memset(s->ac_val[0][(x + 1) + (y) * wrap], 0, 16 * sizeof(INT16));
+            memset(s->ac_val[0][(x) + (y + 1) * wrap], 0, 16 * sizeof(INT16));
+            memset(s->ac_val[0][(x + 1) + (y + 1) * wrap], 0, 16 * sizeof(INT16));
+            if (s->h263_msmpeg4) {
+                s->coded_block[(x) + (y) * wrap] = 0;
+                s->coded_block[(x + 1) + (y) * wrap] = 0;
+                s->coded_block[(x) + (y + 1) * wrap] = 0;
+                s->coded_block[(x + 1) + (y + 1) * wrap] = 0;
+            }
+            /* chroma */
+            wrap = s->mb_width + 2;
+            x = mb_x + 1;
+            y = mb_y + 1;
+            s->dc_val[1][(x) + (y) * wrap] = v;
+            s->dc_val[2][(x) + (y) * wrap] = v;
+            /* ac pred */
+            memset(s->ac_val[1][(x) + (y) * wrap], 0, 16 * sizeof(INT16));
+            memset(s->ac_val[2][(x) + (y) * wrap], 0, 16 * sizeof(INT16));
+        } else {
+            s->last_dc[0] = 128 << s->intra_dc_precision;
+            s->last_dc[1] = 128 << s->intra_dc_precision;
+            s->last_dc[2] = 128 << s->intra_dc_precision;
+        }
+    }
+    
+    /* update motion predictor */
+    if (s->out_format == FMT_H263) {
+        int x, y, wrap;
+        
+        x = 2 * mb_x + 1;
+        y = 2 * mb_y + 1;
+        wrap = 2 * s->mb_width + 2;
+        if (s->mb_intra) {
+            motion_x = 0;
+            motion_y = 0;
+            goto motion_init;
+        } else if (s->mv_type == MV_TYPE_16X16) {
+            motion_x = s->mv[0][0][0];
+            motion_y = s->mv[0][0][1];
+        motion_init:
+            /* no update if 8X8 because it has been done during parsing */
+            s->motion_val[(x) + (y) * wrap][0] = motion_x;
+            s->motion_val[(x) + (y) * wrap][1] = motion_y;
+            s->motion_val[(x + 1) + (y) * wrap][0] = motion_x;
+            s->motion_val[(x + 1) + (y) * wrap][1] = motion_y;
+            s->motion_val[(x) + (y + 1) * wrap][0] = motion_x;
+            s->motion_val[(x) + (y + 1) * wrap][1] = motion_y;
+            s->motion_val[(x + 1) + (y + 1) * wrap][0] = motion_x;
+            s->motion_val[(x + 1) + (y + 1) * wrap][1] = motion_y;
+        }
+    }
+    
+    if (!s->intra_only) {
+        UINT8 *dest_y, *dest_cb, *dest_cr;
+
+        dest_y = s->current_picture[0] + (mb_y * 16 * s->linesize) + mb_x * 16;
+        dest_cb = s->current_picture[1] + (mb_y * 8 * (s->linesize >> 1)) + mb_x * 8;
+        dest_cr = s->current_picture[2] + (mb_y * 8 * (s->linesize >> 1)) + mb_x * 8;
+
+        if (s->interlaced_dct) {
+            dct_linesize = s->linesize * 2;
+            dct_offset = s->linesize;
+        } else {
+            dct_linesize = s->linesize;
+            dct_offset = s->linesize * 8;
+        }
+
+        if (!s->mb_intra) {
+            /* motion handling */
+            if (!s->no_rounding) 
+                op_pix = put_pixels_tab;
+            else
+                op_pix = put_no_rnd_pixels_tab;
+
+            if (s->mv_dir & MV_DIR_FORWARD) {
+                MPV_motion(s, dest_y, dest_cb, dest_cr, 0, s->last_picture, op_pix);
+                if (!s->no_rounding) 
+                    op_pix = avg_pixels_tab;
+                else
+                    op_pix = avg_no_rnd_pixels_tab;
+            }
+            if (s->mv_dir & MV_DIR_BACKWARD) {
+                MPV_motion(s, dest_y, dest_cb, dest_cr, 1, s->next_picture, op_pix);
+            }
+
+            /* add dct residue */
+            add_dct(s, block[0], 0, dest_y, dct_linesize);
+            add_dct(s, block[1], 1, dest_y + 8, dct_linesize);
+            add_dct(s, block[2], 2, dest_y + dct_offset, dct_linesize);
+            add_dct(s, block[3], 3, dest_y + dct_offset + 8, dct_linesize);
+
+            add_dct(s, block[4], 4, dest_cb, dct_linesize >> 1);
+            add_dct(s, block[5], 5, dest_cr, dct_linesize >> 1);
+        } else {
+            /* dct only in intra block */
+            put_dct(s, block[0], 0, dest_y, dct_linesize);
+            put_dct(s, block[1], 1, dest_y + 8, dct_linesize);
+            put_dct(s, block[2], 2, dest_y + dct_offset, dct_linesize);
+            put_dct(s, block[3], 3, dest_y + dct_offset + 8, dct_linesize);
+
+            put_dct(s, block[4], 4, dest_cb, dct_linesize >> 1);
+            put_dct(s, block[5], 5, dest_cr, dct_linesize >> 1);
+        }
+    }
+}
+
+static void encode_picture(MpegEncContext *s, int picture_number)
+{
+    int mb_x, mb_y, wrap;
+    UINT8 *ptr;
+    DCTELEM block[6][64];
+    int i, motion_x, motion_y;
+
+    s->picture_number = picture_number;
+    if (!s->fixed_qscale) 
+        s->qscale = rate_estimate_qscale(s);
+
+    /* precompute matrix */
+    if (s->out_format == FMT_MJPEG) {
+        /* for mjpeg, we do include qscale in the matrix */
+        s->intra_matrix[0] = default_intra_matrix[0];
+        for(i=1;i<64;i++)
+            s->intra_matrix[i] = (default_intra_matrix[i] * s->qscale) >> 3;
+        convert_matrix(s->q_intra_matrix, s->intra_matrix, 8);
+    } else {
+        convert_matrix(s->q_intra_matrix, s->intra_matrix, s->qscale);
+        convert_matrix(s->q_non_intra_matrix, s->non_intra_matrix, s->qscale);
+    }
+
+    switch(s->out_format) {
+    case FMT_MJPEG:
+        mjpeg_picture_header(s);
+        break;
+    case FMT_H263:
+        if (s->h263_msmpeg4) 
+            msmpeg4_encode_picture_header(s, picture_number);
+        else if (s->h263_pred)
+            mpeg4_encode_picture_header(s, picture_number);
+        else if (s->h263_rv10) 
+            rv10_encode_picture_header(s, picture_number);
+        else
+            h263_encode_picture_header(s, picture_number);
+        break;
+    case FMT_MPEG1:
+        mpeg1_encode_picture_header(s, picture_number);
+        break;
+    }
+        
+    /* init last dc values */
+    /* note: quant matrix value (8) is implied here */
+    s->last_dc[0] = 128;
+    s->last_dc[1] = 128;
+    s->last_dc[2] = 128;
+    s->mb_incr = 1;
+    s->last_mv[0][0][0] = 0;
+    s->last_mv[0][0][1] = 0;
+    s->mv_type = MV_TYPE_16X16;
+    s->mv_dir = MV_DIR_FORWARD;
+
+    for(mb_y=0; mb_y < s->mb_height; mb_y++) {
+        for(mb_x=0; mb_x < s->mb_width; mb_x++) {
+
+            s->mb_x = mb_x;
+            s->mb_y = mb_y;
+
+            /* compute motion vector and macro block type (intra or non intra) */
+            motion_x = 0;
+            motion_y = 0;
+            if (s->pict_type == P_TYPE) {
+                s->mb_intra = estimate_motion(s, mb_x, mb_y,
+                                              &motion_x,
+                                              &motion_y);
+            } else {
+                s->mb_intra = 1;
+            }
+
+            /* get the pixels */
+            wrap = s->linesize;
+            ptr = s->new_picture[0] + (mb_y * 16 * wrap) + mb_x * 16;
+            get_pixels(block[0], ptr, wrap);
+            get_pixels(block[1], ptr + 8, wrap);
+            get_pixels(block[2], ptr + 8 * wrap, wrap);
+            get_pixels(block[3], ptr + 8 * wrap + 8, wrap);
+            wrap = s->linesize >> 1;
+            ptr = s->new_picture[1] + (mb_y * 8 * wrap) + mb_x * 8;
+            get_pixels(block[4], ptr, wrap);
+
+            wrap = s->linesize >> 1;
+            ptr = s->new_picture[2] + (mb_y * 8 * wrap) + mb_x * 8;
+            get_pixels(block[5], ptr, wrap);
+
+            /* subtract previous frame if non intra */
+            if (!s->mb_intra) {
+                int dxy, offset, mx, my;
+
+                dxy = ((motion_y & 1) << 1) | (motion_x & 1);
+                ptr = s->last_picture[0] + 
+                    ((mb_y * 16 + (motion_y >> 1)) * s->linesize) + 
+                    (mb_x * 16 + (motion_x >> 1));
+
+                sub_pixels_2(block[0], ptr, s->linesize, dxy);
+                sub_pixels_2(block[1], ptr + 8, s->linesize, dxy);
+                sub_pixels_2(block[2], ptr + s->linesize * 8, s->linesize, dxy);
+                sub_pixels_2(block[3], ptr + 8 + s->linesize * 8, s->linesize ,dxy);
+
+                if (s->out_format == FMT_H263) {
+                    /* special rounding for h263 */
+                    dxy = 0;
+                    if ((motion_x & 3) != 0)
+                        dxy |= 1;
+                    if ((motion_y & 3) != 0)
+                        dxy |= 2;
+                    mx = motion_x >> 2;
+                    my = motion_y >> 2;
+                } else {
+                    mx = motion_x / 2;
+                    my = motion_y / 2;
+                    dxy = ((my & 1) << 1) | (mx & 1);
+                    mx >>= 1;
+                    my >>= 1;
+                }
+                offset = ((mb_y * 8 + my) * (s->linesize >> 1)) + (mb_x * 8 + mx);
+                ptr = s->last_picture[1] + offset;
+                sub_pixels_2(block[4], ptr, s->linesize >> 1, dxy);
+                ptr = s->last_picture[2] + offset;
+                sub_pixels_2(block[5], ptr, s->linesize >> 1, dxy);
+            }
+
+            /* DCT & quantize */
+            if (s->h263_msmpeg4) {
+                msmpeg4_dc_scale(s);
+            } else if (s->h263_pred) {
+                h263_dc_scale(s);
+            } else {
+                /* default quantization values */
+                s->y_dc_scale = 8;
+                s->c_dc_scale = 8;
+            }
+
+            for(i=0;i<6;i++) {
+                int last_index;
+                if (av_fdct == jpeg_fdct_ifast)
+                    last_index = dct_quantize(s, block[i], i, s->qscale);
+                else
+                    last_index = dct_quantize_mmx(s, block[i], i, s->qscale);
+                s->block_last_index[i] = last_index;
+            }
+
+            /* huffman encode */
+            switch(s->out_format) {
+            case FMT_MPEG1:
+                mpeg1_encode_mb(s, block, motion_x, motion_y);
+                break;
+            case FMT_H263:
+                if (s->h263_msmpeg4)
+                    msmpeg4_encode_mb(s, block, motion_x, motion_y);
+                else
+                    h263_encode_mb(s, block, motion_x, motion_y);
+                break;
+            case FMT_MJPEG:
+                mjpeg_encode_mb(s, block);
+                break;
+            }
+
+            /* decompress blocks so that we keep the state of the decoder */
+            s->mv[0][0][0] = motion_x;
+            s->mv[0][0][1] = motion_y;
+
+            MPV_decode_mb(s, block);
+        }
+    }
+}
+
+static int dct_quantize(MpegEncContext *s, 
+                        DCTELEM *block, int n,
+                        int qscale)
+{
+    int i, j, level, last_non_zero, q;
+    const int *qmat;
+
+    av_fdct (block);
+
+    if (s->mb_intra) {
+        if (n < 4)
+            q = s->y_dc_scale;
+        else
+            q = s->c_dc_scale;
+        q = q << 3;
+        
+        /* note: block[0] is assumed to be positive */
+        block[0] = (block[0] + (q >> 1)) / q;
+        i = 1;
+        last_non_zero = 0;
+        if (s->out_format == FMT_H263) {
+            qmat = s->q_non_intra_matrix;
+        } else {
+            qmat = s->q_intra_matrix;
+        }
+    } else {
+        i = 0;
+        last_non_zero = -1;
+        qmat = s->q_non_intra_matrix;
+    }
+
+    for(;i<64;i++) {
+        j = zigzag_direct[i];
+        level = block[j];
+        level = level * qmat[j];
+#ifdef PARANOID
+        {
+            static int count = 0;
+            int level1, level2, qmat1;
+            double val;
+            if (qmat == s->q_non_intra_matrix) {
+                qmat1 = default_non_intra_matrix[j] * s->qscale;
+            } else {
+                qmat1 = default_intra_matrix[j] * s->qscale;
+            }
+            if (av_fdct != jpeg_fdct_ifast)
+                val = ((double)block[j] * 8.0) / (double)qmat1;
+            else
+                val = ((double)block[j] * 8.0 * 2048.0) / 
+                    ((double)qmat1 * aanscales[j]);
+            level1 = (int)val;
+            level2 = level / (1 << (QMAT_SHIFT - 3));
+            if (level1 != level2) {
+                fprintf(stderr, "%d: quant error qlevel=%d wanted=%d level=%d qmat1=%d qmat=%d wantedf=%0.6f\n", 
+                        count, level2, level1, block[j], qmat1, qmat[j],
+                        val);
+                count++;
+            }
+
+        }
+#endif
+        /* XXX: slight error for the low range. Test should be equivalent to
+           (level <= -(1 << (QMAT_SHIFT - 3)) || level >= (1 <<
+           (QMAT_SHIFT - 3)))
+        */
+        if (((level << (31 - (QMAT_SHIFT - 3))) >> (31 - (QMAT_SHIFT - 3))) != 
+            level) {
+            level = level / (1 << (QMAT_SHIFT - 3));
+            /* XXX: currently, this code is not optimal. the range should be:
+               mpeg1: -255..255
+               mpeg2: -2048..2047
+               h263:  -128..127
+               mpeg4: -2048..2047
+            */
+            if (level > 127)
+                level = 127;
+            else if (level < -128)
+                level = -128;
+            block[j] = level;
+            last_non_zero = i;
+        } else {
+            block[j] = 0;
+        }
+    }
+    return last_non_zero;
+}
+
+static int dct_quantize_mmx(MpegEncContext *s, 
+                            DCTELEM *block, int n,
+                            int qscale)
+{
+    int i, j, level, last_non_zero, q;
+    const int *qmat;
+
+    av_fdct (block);
+
+    if (s->mb_intra) {
+        if (n < 4)
+            q = s->y_dc_scale;
+        else
+            q = s->c_dc_scale;
+        
+        /* note: block[0] is assumed to be positive */
+        block[0] = (block[0] + (q >> 1)) / q;
+        i = 1;
+        last_non_zero = 0;
+        if (s->out_format == FMT_H263) {
+            qmat = s->q_non_intra_matrix;
+        } else {
+            qmat = s->q_intra_matrix;
+        }
+    } else {
+        i = 0;
+        last_non_zero = -1;
+        qmat = s->q_non_intra_matrix;
+    }
+
+    for(;i<64;i++) {
+        j = zigzag_direct[i];
+        level = block[j];
+        level = level * qmat[j];
+        /* XXX: slight error for the low range. Test should be equivalent to
+           (level <= -(1 << (QMAT_SHIFT_MMX - 3)) || level >= (1 <<
+           (QMAT_SHIFT_MMX - 3)))
+        */
+        if (((level << (31 - (QMAT_SHIFT_MMX - 3))) >> (31 - (QMAT_SHIFT_MMX - 3))) != 
+            level) {
+            level = level / (1 << (QMAT_SHIFT_MMX - 3));
+            /* XXX: currently, this code is not optimal. the range should be:
+               mpeg1: -255..255
+               mpeg2: -2048..2047
+               h263:  -128..127
+               mpeg4: -2048..2047
+            */
+            if (level > 127)
+                level = 127;
+            else if (level < -128)
+                level = -128;
+            block[j] = level;
+            last_non_zero = i;
+        } else {
+            block[j] = 0;
+        }
+    }
+    return last_non_zero;
+}
+
+static void dct_unquantize(MpegEncContext *s, 
+                           DCTELEM *block, int n, int qscale)
+{
+    int i, level;
+    const UINT16 *quant_matrix;
+
+    if (s->mb_intra) {
+        if (n < 4) 
+            block[0] = block[0] * s->y_dc_scale;
+        else
+            block[0] = block[0] * s->c_dc_scale;
+        if (s->out_format == FMT_H263) {
+            i = 1;
+            goto unquant_even;
+        }
+        /* XXX: only mpeg1 */
+        quant_matrix = s->intra_matrix;
+        for(i=1;i<64;i++) {
+            level = block[i];
+            if (level) {
+                if (level < 0) {
+                    level = -level;
+                    level = (int)(level * qscale * quant_matrix[i]) >> 3;
+                    level = (level - 1) | 1;
+                    level = -level;
+                } else {
+                    level = (int)(level * qscale * quant_matrix[i]) >> 3;
+                    level = (level - 1) | 1;
+                }
+#ifdef PARANOID
+                if (level < -2048 || level > 2047)
+                    fprintf(stderr, "unquant error %d %d\n", i, level);
+#endif
+                block[i] = level;
+            }
+        }
+    } else {
+        i = 0;
+    unquant_even:
+        quant_matrix = s->non_intra_matrix;
+        for(;i<64;i++) {
+            level = block[i];
+            if (level) {
+                if (level < 0) {
+                    level = -level;
+                    level = (((level << 1) + 1) * qscale *
+                             ((int) (quant_matrix[i]))) >> 4;
+                    level = (level - 1) | 1;
+                    level = -level;
+                } else {
+                    level = (((level << 1) + 1) * qscale *
+                             ((int) (quant_matrix[i]))) >> 4;
+                    level = (level - 1) | 1;
+                }
+#ifdef PARANOID
+                if (level < -2048 || level > 2047)
+                    fprintf(stderr, "unquant error %d %d\n", i, level);
+#endif
+                block[i] = level;
+            }
+        }
+    }
+}
+                         
+
+/* rate control */
+
+/* an I frame is I_FRAME_SIZE_RATIO bigger than a P frame */
+#define I_FRAME_SIZE_RATIO 3.0
+#define QSCALE_K           20
+
+static void rate_control_init(MpegEncContext *s)
+{
+    s->wanted_bits = 0;
+
+    if (s->intra_only) {
+        s->I_frame_bits = ((INT64)s->bit_rate * FRAME_RATE_BASE) / s->frame_rate;
+        s->P_frame_bits = s->I_frame_bits;
+    } else {
+        s->P_frame_bits = (int) ((float)(s->gop_size * s->bit_rate) / 
+                                 (float)((float)s->frame_rate / FRAME_RATE_BASE * (I_FRAME_SIZE_RATIO + s->gop_size - 1)));
+        s->I_frame_bits = (int)(s->P_frame_bits * I_FRAME_SIZE_RATIO);
+    }
+    
+#if defined(DEBUG)
+    printf("I_frame_size=%d P_frame_size=%d\n",
+           s->I_frame_bits, s->P_frame_bits);
+#endif
+}
+
+
+/*
+ * This heuristic is rather poor, but at least we do not have to
+ * change the qscale at every macroblock.
+ */
+static int rate_estimate_qscale(MpegEncContext *s)
+{
+    long long total_bits = s->total_bits;
+    float q;
+    int qscale, diff, qmin;
+
+    if (s->pict_type == I_TYPE) {
+        s->wanted_bits += s->I_frame_bits;
+    } else {
+        s->wanted_bits += s->P_frame_bits;
+    }
+    diff = s->wanted_bits - total_bits;
+    q = 31.0 - (float)diff / (QSCALE_K * s->mb_height * s->mb_width);
+    /* adjust for I frame */
+    if (s->pict_type == I_TYPE && !s->intra_only) {
+        q /= I_FRAME_SIZE_RATIO;
+    }
+
+    /* using a too small Q scale leeds to problems in mpeg1 and h263
+       because AC coefficients are clamped to 255 or 127 */
+    qmin = 3;
+    if (q < qmin)
+        q = qmin;
+    else if (q > 31)
+        q = 31;
+    qscale = (int)(q + 0.5);
+#if defined(DEBUG)
+    printf("%d: total=%Ld br=%0.1f diff=%d qest=%0.1f\n", 
+           s->picture_number, 
+           total_bits, 
+           (float)s->frame_rate / FRAME_RATE_BASE * 
+           total_bits / s->picture_number, 
+           diff, q);
+#endif
+    return qscale;
+}
+
+AVCodec mpeg1video_encoder = {
+    "mpeg1video",
+    CODEC_TYPE_VIDEO,
+    CODEC_ID_MPEG1VIDEO,
+    sizeof(MpegEncContext),
+    MPV_encode_init,
+    MPV_encode_picture,
+    MPV_encode_end,
+};
+
+AVCodec h263_encoder = {
+    "h263",
+    CODEC_TYPE_VIDEO,
+    CODEC_ID_H263,
+    sizeof(MpegEncContext),
+    MPV_encode_init,
+    MPV_encode_picture,
+    MPV_encode_end,
+};
+
+AVCodec h263p_encoder = {
+    "h263p",
+    CODEC_TYPE_VIDEO,
+    CODEC_ID_H263P,
+    sizeof(MpegEncContext),
+    MPV_encode_init,
+    MPV_encode_picture,
+    MPV_encode_end,
+};
+
+AVCodec rv10_encoder = {
+    "rv10",
+    CODEC_TYPE_VIDEO,
+    CODEC_ID_RV10,
+    sizeof(MpegEncContext),
+    MPV_encode_init,
+    MPV_encode_picture,
+    MPV_encode_end,
+};
+
+AVCodec mjpeg_encoder = {
+    "mjpeg",
+    CODEC_TYPE_VIDEO,
+    CODEC_ID_MJPEG,
+    sizeof(MpegEncContext),
+    MPV_encode_init,
+    MPV_encode_picture,
+    MPV_encode_end,
+};
+
+AVCodec opendivx_encoder = {
+    "opendivx",
+    CODEC_TYPE_VIDEO,
+    CODEC_ID_OPENDIVX,
+    sizeof(MpegEncContext),
+    MPV_encode_init,
+    MPV_encode_picture,
+    MPV_encode_end,
+};
+
+AVCodec msmpeg4_encoder = {
+    "msmpeg4",
+    CODEC_TYPE_VIDEO,
+    CODEC_ID_MSMPEG4,
+    sizeof(MpegEncContext),
+    MPV_encode_init,
+    MPV_encode_picture,
+    MPV_encode_end,
+};
diff --git a/libavcodec/mpegvideo.h b/libavcodec/mpegvideo.h
new file mode 100644
index 0000000000..b04d7b7693
--- /dev/null
+++ b/libavcodec/mpegvideo.h
@@ -0,0 +1,273 @@
+/*
+ * Generic DCT based hybrid video encoder
+ * Copyright (c) 2000,2001 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+/* Macros for picture code type. */
+#define I_TYPE 1
+#define P_TYPE 2
+#define B_TYPE 3
+
+enum OutputFormat {
+    FMT_MPEG1,
+    FMT_H263,
+    FMT_MJPEG,
+};
+
+#define MPEG_BUF_SIZE (16 * 1024)
+
+typedef struct MpegEncContext {
+    /* the following parameters must be initialized before encoding */
+    int width, height; /* picture size. must be a multiple of 16 */
+    int gop_size;
+    int frame_rate; /* number of frames per second */
+    int intra_only; /* if true, only intra pictures are generated */
+    int bit_rate;        /* wanted bit rate */
+    enum OutputFormat out_format; /* output format */
+    int h263_plus; /* h263 plus headers */
+    int h263_rv10; /* use RV10 variation for H263 */
+    int h263_pred; /* use OpenDIVX (aka mpeg4) ac/dc predictions */
+    int h263_msmpeg4; /* generate MSMPEG4 compatible stream */
+    int h263_intel; /* use I263 intel h263 header */
+    int fixed_qscale; /* fixed qscale if non zero */
+    /* the following fields are managed internally by the encoder */
+
+    /* bit output */
+    PutBitContext pb;
+
+    /* sequence parameters */
+    int context_initialized;
+    int picture_number;
+    int fake_picture_number; /* picture number at the bitstream frame rate */
+    int gop_picture_number; /* index of the first picture of a GOP */
+    int mb_width, mb_height;
+    int linesize;              /* line size, in bytes, may be different from width */
+    UINT8 *new_picture[3];     /* picture to be compressed */
+    UINT8 *last_picture[3];    /* previous picture */
+    UINT8 *last_picture_base[3]; /* real start of the picture */
+    UINT8 *next_picture[3];    /* previous picture (for bidir pred) */
+    UINT8 *next_picture_base[3]; /* real start of the picture */
+    UINT8 *aux_picture[3];    /* aux picture (for B frames only) */
+    UINT8 *aux_picture_base[3]; /* real start of the picture */
+    UINT8 *current_picture[3]; /* buffer to store the decompressed current picture */
+    int last_dc[3]; /* last DC values for MPEG1 */
+    INT16 *dc_val[3]; /* used for mpeg4 DC prediction */
+    int y_dc_scale, c_dc_scale;
+    UINT8 *coded_block; /* used for coded block pattern prediction */
+    INT16 (*ac_val[3])[16]; /* used for for mpeg4 AC prediction */
+    int ac_pred;
+
+    int qscale;
+    int pict_type;
+    int frame_rate_index;
+    /* motion compensation */
+    int unrestricted_mv;
+    int h263_long_vectors; /* use horrible h263v1 long vector mode */
+
+    int f_code; /* resolution */
+    INT16 (*motion_val)[2]; /* used for MV prediction */
+    int full_search;
+    int mv_dir;
+#define MV_DIR_BACKWARD  1
+#define MV_DIR_FORWARD   2
+    int mv_type;
+#define MV_TYPE_16X16       0   /* 1 vector for the whole mb */
+#define MV_TYPE_8X8         1   /* 4 vectors (h263) */
+#define MV_TYPE_16X8        2   /* 2 vectors, one per 16x8 block */ 
+#define MV_TYPE_FIELD       3   /* 2 vectors, one per field */ 
+#define MV_TYPE_DMV         4   /* 2 vectors, special mpeg2 Dual Prime Vectors */
+    /* motion vectors for a macroblock 
+       first coordinate : 0 = forward 1 = backward
+       second "         : depend on type
+       third  "         : 0 = x, 1 = y
+    */
+    int mv[2][4][2];
+    int field_select[2][2];
+    int last_mv[2][2][2];
+
+    int has_b_frames;
+    int no_rounding; /* apply no rounding to motion estimation (MPEG4) */
+
+    /* macroblock layer */
+    int mb_x, mb_y;
+    int mb_incr;
+    int mb_intra;
+    /* matrix transmitted in the bitstream */
+    UINT16 intra_matrix[64];
+    UINT16 chroma_intra_matrix[64];
+    UINT16 non_intra_matrix[64];
+    UINT16 chroma_non_intra_matrix[64];
+    /* precomputed matrix (combine qscale and DCT renorm) */
+    int q_intra_matrix[64];
+    int q_non_intra_matrix[64];
+    int block_last_index[6];  /* last non zero coefficient in block */
+
+    void *opaque; /* private data for the user */
+
+    /* bit rate control */
+    int I_frame_bits;    /* wanted number of bits per I frame */
+    int P_frame_bits;    /* same for P frame */
+    long long wanted_bits;
+    long long total_bits;
+
+    /* mpeg4 specific */
+    int time_increment_bits;
+
+    /* RV10 specific */
+    int rv10_version; /* RV10 version: 0 or 3 */
+    int rv10_first_dc_coded[3];
+    
+    /* MJPEG specific */
+    struct MJpegContext *mjpeg_ctx;
+
+    /* MSMPEG4 specific */
+    int mv_table_index;
+    int rl_table_index;
+    int rl_chroma_table_index;
+    int dc_table_index;
+    int use_skip_mb_code;
+    int slice_height;      /* in macroblocks */
+    int first_slice_line;  
+    /* decompression specific */
+    GetBitContext gb;
+
+    /* MPEG2 specific - I wish I had not to support this mess. */
+    int progressive_sequence;
+    int mpeg_f_code[2][2];
+    int picture_structure;
+/* picture type */
+#define PICT_TOP_FIELD     1
+#define PICT_BOTTOM_FIELD  2
+#define PICT_FRAME         3
+
+    int intra_dc_precision;
+    int frame_pred_frame_dct;
+    int top_field_first;
+    int concealment_motion_vectors;
+    int q_scale_type;
+    int intra_vlc_format;
+    int alternate_scan;
+    int repeat_first_field;
+    int chroma_420_type;
+    int progressive_frame;
+    int mpeg2;
+    int full_pel[2];
+    int interlaced_dct;
+    int last_qscale;
+    int first_slice;
+} MpegEncContext;
+
+extern const UINT8 zigzag_direct[64];
+
+int MPV_common_init(MpegEncContext *s);
+void MPV_common_end(MpegEncContext *s);
+void MPV_decode_mb(MpegEncContext *s, DCTELEM block[6][64]);
+void MPV_frame_start(MpegEncContext *s);
+void MPV_frame_end(MpegEncContext *s);
+
+/* motion_est.c */
+
+int estimate_motion(MpegEncContext *s, 
+                    int mb_x, int mb_y,
+                    int *mx_ptr, int *my_ptr);
+
+/* mpeg12.c */
+extern const UINT8 default_intra_matrix[64];
+extern const UINT8 default_non_intra_matrix[64];
+
+void mpeg1_encode_picture_header(MpegEncContext *s, int picture_number);
+void mpeg1_encode_mb(MpegEncContext *s,
+                     DCTELEM block[6][64],
+                     int motion_x, int motion_y);
+
+/* h263enc.c */
+
+/* run length table */
+#define MAX_RUN    64
+#define MAX_LEVEL  64
+
+typedef struct RLTable {
+    int n; /* number of entries of table_vlc minus 1 */
+    int last; /* number of values for last = 0 */
+    const UINT16 (*table_vlc)[2];
+    const INT8 *table_run;
+    const INT8 *table_level;
+    UINT8 *index_run[2]; /* encoding only */
+    INT8 *max_level[2]; /* encoding & decoding */
+    INT8 *max_run[2];   /* encoding & decoding */
+    VLC vlc;            /* decoding only */
+} RLTable;
+
+void init_rl(RLTable *rl);
+void init_vlc_rl(RLTable *rl);
+
+extern inline int get_rl_index(const RLTable *rl, int last, int run, int level)
+{
+    int index;
+    index = rl->index_run[last][run];
+    if (index >= rl->n)
+        return rl->n;
+    if (level > rl->max_level[last][run])
+        return rl->n;
+    return index + level - 1;
+}
+
+void h263_encode_mb(MpegEncContext *s, 
+                    DCTELEM block[6][64],
+                    int motion_x, int motion_y);
+void h263_encode_picture_header(MpegEncContext *s, int picture_number);
+void h263_dc_scale(MpegEncContext *s);
+INT16 *h263_pred_motion(MpegEncContext * s, int block, 
+                        int *px, int *py);
+void mpeg4_pred_ac(MpegEncContext * s, INT16 *block, int n, 
+                   int dir);
+void mpeg4_encode_picture_header(MpegEncContext *s, int picture_number);
+void h263_encode_init_vlc(MpegEncContext *s);
+
+void h263_decode_init_vlc(MpegEncContext *s);
+int h263_decode_picture_header(MpegEncContext *s);
+int mpeg4_decode_picture_header(MpegEncContext * s);
+int intel_h263_decode_picture_header(MpegEncContext *s);
+int h263_decode_mb(MpegEncContext *s,
+                   DCTELEM block[6][64]);
+int h263_get_picture_format(int width, int height);
+extern UINT8 ff_alternate_horizontal_scan[64];
+extern UINT8 ff_alternate_vertical_scan[64];
+
+/* rv10.c */
+void rv10_encode_picture_header(MpegEncContext *s, int picture_number);
+int rv_decode_dc(MpegEncContext *s, int n);
+
+/* msmpeg4.c */
+void msmpeg4_encode_picture_header(MpegEncContext * s, int picture_number);
+void msmpeg4_encode_mb(MpegEncContext * s, 
+                       DCTELEM block[6][64],
+                       int motion_x, int motion_y);
+void msmpeg4_dc_scale(MpegEncContext * s);
+int msmpeg4_decode_picture_header(MpegEncContext * s);
+int msmpeg4_decode_mb(MpegEncContext *s, 
+                      DCTELEM block[6][64]);
+int msmpeg4_decode_init_vlc(MpegEncContext *s);
+
+/* mjpegenc.c */
+
+int mjpeg_init(MpegEncContext *s);
+void mjpeg_close(MpegEncContext *s);
+void mjpeg_encode_mb(MpegEncContext *s, 
+                     DCTELEM block[6][64]);
+void mjpeg_picture_header(MpegEncContext *s);
+void mjpeg_picture_trailer(MpegEncContext *s);
diff --git a/libavcodec/msmpeg4.c b/libavcodec/msmpeg4.c
new file mode 100644
index 0000000000..044d964ee5
--- /dev/null
+++ b/libavcodec/msmpeg4.c
@@ -0,0 +1,931 @@
+/*
+ * MSMPEG4 backend for ffmpeg encoder and decoder
+ * Copyright (c) 2001 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include "common.h"
+#include "dsputil.h"
+#include "mpegvideo.h"
+
+/*
+ * You can also call this codec : MPEG4 with a twist ! 
+ *
+ * TODO: 
+ *        - (encoding) select best mv table (two choices)
+ *        - (encoding) select best vlc/dc table 
+ *        - (decoding) handle slice indication
+ */
+//#define DEBUG
+
+/* motion vector table */
+typedef struct MVTable {
+    int n;
+    const UINT16 *table_mv_code;
+    const UINT8 *table_mv_bits;
+    const UINT8 *table_mvx;
+    const UINT8 *table_mvy;
+    UINT16 *table_mv_index; /* encoding: convert mv to index in table_mv */
+    VLC vlc;                /* decoding: vlc */
+} MVTable;
+
+static void msmpeg4_encode_block(MpegEncContext * s, DCTELEM * block, int n);
+static int msmpeg4_decode_block(MpegEncContext * s, DCTELEM * block,
+                                int n, int coded);
+static int msmpeg4_decode_dc(MpegEncContext * s, int n, int *dir_ptr);
+static int msmpeg4_decode_motion(MpegEncContext * s, 
+                                 int *mx_ptr, int *my_ptr);
+
+#ifdef DEBUG
+int intra_count = 0;
+int frame_count = 0;
+#endif
+/* XXX: move it to mpegvideo.h */
+
+static int init_done = 0;
+
+#include "msmpeg4data.h"
+
+#ifdef STATS
+
+const char *st_names[ST_NB] = {
+    "unknown",
+    "dc",
+    "intra_ac",
+    "inter_ac",
+    "intra_mb",
+    "inter_mb",
+    "mv",
+};
+
+int st_current_index = 0;
+unsigned int st_bit_counts[ST_NB];
+unsigned int st_out_bit_counts[ST_NB];
+
+#define set_stat(var) st_current_index = var;
+
+void print_stats(void)
+{
+    unsigned int total;
+    int i;
+
+    printf("Input:\n");
+    total = 0;
+    for(i=0;i<ST_NB;i++)
+        total += st_bit_counts[i];
+    if (total == 0)
+        total = 1;
+    for(i=0;i<ST_NB;i++) {
+        printf("%-10s : %10.1f %5.1f%%\n", 
+               st_names[i], 
+               (double)st_bit_counts[i] / 8.0, 
+               (double)st_bit_counts[i] * 100.0 / total);
+    }
+    printf("%-10s : %10.1f %5.1f%%\n",
+           "total", 
+           (double)total / 8.0, 
+           100.0);
+
+    printf("Output:\n");
+    total = 0;
+    for(i=0;i<ST_NB;i++)
+        total += st_out_bit_counts[i];
+    if (total == 0)
+        total = 1;
+    for(i=0;i<ST_NB;i++) {
+        printf("%-10s : %10.1f %5.1f%%\n", 
+               st_names[i], 
+               (double)st_out_bit_counts[i] / 8.0, 
+               (double)st_out_bit_counts[i] * 100.0 / total);
+    }
+    printf("%-10s : %10.1f %5.1f%%\n",
+           "total", 
+           (double)total / 8.0, 
+           100.0);
+}
+
+#else
+
+#define set_stat(var)
+
+#endif
+
+/* build the table which associate a (x,y) motion vector to a vlc */
+static void init_mv_table(MVTable *tab)
+{
+    int i, x, y;
+
+    tab->table_mv_index = malloc(sizeof(UINT16) * 4096);
+    /* mark all entries as not used */
+    for(i=0;i<4096;i++)
+        tab->table_mv_index[i] = tab->n;
+    
+    for(i=0;i<tab->n;i++) {
+        x = tab->table_mvx[i];
+        y = tab->table_mvy[i];
+        tab->table_mv_index[(x << 6) | y] = i;
+    }
+}
+
+static void code012(PutBitContext *pb, int n)
+{
+    if (n == 0) {
+        put_bits(pb, 1, 0);
+    } else {
+        put_bits(pb, 1, 1);
+        put_bits(pb, 1, (n >= 2));
+    }
+}
+
+/* write MSMPEG4 V3 compatible frame header */
+void msmpeg4_encode_picture_header(MpegEncContext * s, int picture_number)
+{
+    int i;
+
+    align_put_bits(&s->pb);
+
+    put_bits(&s->pb, 2, s->pict_type - 1);
+
+    put_bits(&s->pb, 5, s->qscale);
+
+    s->rl_table_index = 2;
+    s->rl_chroma_table_index = 1; /* only for I frame */
+    s->dc_table_index = 1;
+    s->mv_table_index = 1; /* only if P frame */
+    s->use_skip_mb_code = 1; /* only if P frame */
+    
+    if (s->pict_type == I_TYPE) {
+        put_bits(&s->pb, 5, 0x17); /* indicate only one "slice" */
+
+        code012(&s->pb, s->rl_chroma_table_index);
+        code012(&s->pb, s->rl_table_index);
+
+        put_bits(&s->pb, 1, s->dc_table_index);
+        s->no_rounding = 1;
+    } else {
+        put_bits(&s->pb, 1, s->use_skip_mb_code);
+        
+        s->rl_chroma_table_index = s->rl_table_index;
+        code012(&s->pb, s->rl_table_index);
+
+        put_bits(&s->pb, 1, s->dc_table_index);
+
+        put_bits(&s->pb, 1, s->mv_table_index);
+        s->no_rounding ^= 1;
+    }
+
+    if (!init_done) {
+        /* init various encoding tables */
+        init_done = 1;
+        init_mv_table(&mv_tables[0]);
+        init_mv_table(&mv_tables[1]);
+        for(i=0;i<NB_RL_TABLES;i++)
+            init_rl(&rl_table[i]);
+    }
+
+#ifdef DEBUG
+    intra_count = 0;
+    printf("*****frame %d:\n", frame_count++);
+#endif
+}
+
+/* predict coded block */
+static inline int coded_block_pred(MpegEncContext * s, int n, UINT8 **coded_block_ptr)
+{
+    int x, y, wrap, pred, a, b, c;
+
+    x = 2 * s->mb_x + 1 + (n & 1);
+    y = 2 * s->mb_y + 1 + ((n & 2) >> 1);
+    wrap = s->mb_width * 2 + 2;
+
+    /* B C
+     * A X 
+     */
+    a = s->coded_block[(x - 1) + (y) * wrap];
+    b = s->coded_block[(x - 1) + (y - 1) * wrap];
+    c = s->coded_block[(x) + (y - 1) * wrap];
+    
+    if (b == c) {
+        pred = a;
+    } else {
+        pred = c;
+    }
+    
+    /* store value */
+    *coded_block_ptr = &s->coded_block[(x) + (y) * wrap];
+
+    return pred;
+}
+
+static void msmpeg4_encode_motion(MpegEncContext * s, 
+                                  int mx, int my)
+{
+    int code;
+    MVTable *mv;
+
+    /* modulo encoding */
+    /* WARNING : you cannot reach all the MVs even with the modulo
+       encoding. This is a somewhat strange compromise they took !!!  */
+    if (mx <= -64)
+        mx += 64;
+    else if (mx >= 64)
+        mx -= 64;
+    if (my <= -64)
+        my += 64;
+    else if (my >= 64)
+        my -= 64;
+    
+    mx += 32;
+    my += 32;
+#if 0
+    if ((unsigned)mx >= 64 ||
+        (unsigned)my >= 64) 
+        fprintf(stderr, "error mx=%d my=%d\n", mx, my);
+#endif
+    mv = &mv_tables[s->mv_table_index];
+
+    code = mv->table_mv_index[(mx << 6) | my];
+    set_stat(ST_MV);
+    put_bits(&s->pb, 
+             mv->table_mv_bits[code], 
+             mv->table_mv_code[code]);
+    if (code == mv->n) {
+        /* escape : code litterally */
+        put_bits(&s->pb, 6, mx);
+        put_bits(&s->pb, 6, my);
+    }
+}
+
+void msmpeg4_encode_mb(MpegEncContext * s, 
+                       DCTELEM block[6][64],
+                       int motion_x, int motion_y)
+{
+    int cbp, coded_cbp, i;
+    int pred_x, pred_y;
+    UINT8 *coded_block;
+
+    if (!s->mb_intra) {
+	/* compute cbp */
+        set_stat(ST_INTER_MB);
+	cbp = 0;
+	for (i = 0; i < 6; i++) {
+	    if (s->block_last_index[i] >= 0)
+		cbp |= 1 << (5 - i);
+	}
+	if (s->use_skip_mb_code && (cbp | motion_x | motion_y) == 0) {
+	    /* skip macroblock */
+	    put_bits(&s->pb, 1, 1);
+	    return;
+	}
+        if (s->use_skip_mb_code)
+            put_bits(&s->pb, 1, 0);	/* mb coded */
+        
+        put_bits(&s->pb, 
+                 table_mb_non_intra[cbp + 64][1], 
+                 table_mb_non_intra[cbp + 64][0]);
+
+        /* motion vector */
+        h263_pred_motion(s, 0, &pred_x, &pred_y);
+        msmpeg4_encode_motion(s, motion_x - pred_x, 
+                              motion_y - pred_y);
+    } else {
+	/* compute cbp */
+	cbp = 0;
+        coded_cbp = 0;
+	for (i = 0; i < 6; i++) {
+            int val, pred;
+            val = (s->block_last_index[i] >= 1);
+            cbp |= val << (5 - i);
+            if (i < 4) {
+                /* predict value for close blocks only for luma */
+                pred = coded_block_pred(s, i, &coded_block);
+                *coded_block = val;
+                val = val ^ pred;
+            }
+            coded_cbp |= val << (5 - i);
+	}
+#if 0
+        if (coded_cbp)
+            printf("cbp=%x %x\n", cbp, coded_cbp);
+#endif
+
+	if (s->pict_type == I_TYPE) {
+            set_stat(ST_INTRA_MB);
+            put_bits(&s->pb, 
+                     table_mb_intra[coded_cbp][1], table_mb_intra[coded_cbp][0]);
+        } else {
+            if (s->use_skip_mb_code)
+                put_bits(&s->pb, 1, 0);	/* mb coded */
+            put_bits(&s->pb, 
+                     table_mb_non_intra[cbp][1], 
+                     table_mb_non_intra[cbp][0]);
+        }
+        set_stat(ST_INTRA_MB);
+        put_bits(&s->pb, 1, 0);	/* no AC prediction yet */
+    }
+
+    for (i = 0; i < 6; i++) {
+        msmpeg4_encode_block(s, block[i], i);
+    }
+}
+
+
+/* strongly inspirated from MPEG4, but not exactly the same ! */
+void msmpeg4_dc_scale(MpegEncContext * s)
+{
+    int scale;
+
+    if (s->qscale < 5)
+        scale = 8;
+    else if (s->qscale < 9)
+        scale = 2 * s->qscale;
+    else 
+        scale = s->qscale + 8;
+    s->y_dc_scale = scale;
+    s->c_dc_scale = (s->qscale + 13) / 2;
+}
+
+/* dir = 0: left, dir = 1: top prediction */
+static int msmpeg4_pred_dc(MpegEncContext * s, int n, 
+                           UINT16 **dc_val_ptr, int *dir_ptr)
+{
+    int a, b, c, x, y, wrap, pred, scale;
+    UINT16 *dc_val;
+
+    /* find prediction */
+    if (n < 4) {
+	x = 2 * s->mb_x + 1 + (n & 1);
+	y = 2 * s->mb_y + 1 + ((n & 2) >> 1);
+	wrap = s->mb_width * 2 + 2;
+	dc_val = s->dc_val[0];
+	scale = s->y_dc_scale;
+    } else {
+	x = s->mb_x + 1;
+	y = s->mb_y + 1;
+	wrap = s->mb_width + 2;
+	dc_val = s->dc_val[n - 4 + 1];
+	scale = s->c_dc_scale;
+    }
+
+    /* B C
+     * A X 
+     */
+    a = dc_val[(x - 1) + (y) * wrap];
+    b = dc_val[(x - 1) + (y - 1) * wrap];
+    c = dc_val[(x) + (y - 1) * wrap];
+
+    /* XXX: the following solution consumes divisions, but it does not
+       necessitate to modify mpegvideo.c. The problem comes from the
+       fact they decided to store the quantized DC (which would lead
+       to problems if Q could vary !) */
+    a = (a + (scale >> 1)) / scale;
+    b = (b + (scale >> 1)) / scale;
+    c = (c + (scale >> 1)) / scale;
+
+    /* XXX: WARNING: they did not choose the same test as MPEG4. This
+       is very important ! */
+    if (abs(a - b) <= abs(b - c)) {
+	pred = c;
+        *dir_ptr = 1;
+    } else {
+	pred = a;
+        *dir_ptr = 0;
+    }
+
+    /* update predictor */
+    *dc_val_ptr = &dc_val[(x) + (y) * wrap];
+    return pred;
+}
+
+#define DC_MAX 119
+
+static void msmpeg4_encode_dc(MpegEncContext * s, int level, int n, int *dir_ptr)
+{
+    int sign, code;
+    int pred;
+    UINT16 *dc_val;
+
+    pred = msmpeg4_pred_dc(s, n, &dc_val, dir_ptr);
+
+    /* update predictor */
+    if (n < 4) {
+        *dc_val = level * s->y_dc_scale;
+    } else {
+        *dc_val = level * s->c_dc_scale;
+    }
+
+    /* do the prediction */
+    level -= pred;
+
+    sign = 0;
+    if (level < 0) {
+        level = -level;
+        sign = 1;
+    }
+    
+    code = level;
+    if (code > DC_MAX) 
+        code = DC_MAX;
+
+    if (s->dc_table_index == 0) {
+        if (n < 4) {
+            put_bits(&s->pb, table0_dc_lum[code][1], table0_dc_lum[code][0]);
+        } else {
+            put_bits(&s->pb, table0_dc_chroma[code][1], table0_dc_chroma[code][0]);
+        }
+    } else {
+        if (n < 4) {
+            put_bits(&s->pb, table1_dc_lum[code][1], table1_dc_lum[code][0]);
+        } else {
+            put_bits(&s->pb, table1_dc_chroma[code][1], table1_dc_chroma[code][0]);
+        }
+    }
+        
+    if (code == DC_MAX)
+        put_bits(&s->pb, 8, level);
+        
+    if (level != 0) {
+        put_bits(&s->pb, 1, sign);
+    }
+}
+
+/* Encoding of a block. Very similar to MPEG4 except for a different
+   escape coding (same as H263) and more vlc tables.
+ */
+static void msmpeg4_encode_block(MpegEncContext * s, DCTELEM * block, int n)
+{
+    int level, run, last, i, j, last_index;
+    int last_non_zero, sign, slevel;
+    int code, run_diff, dc_pred_dir;
+    const RLTable *rl;
+
+    if (s->mb_intra) {
+        set_stat(ST_DC);
+        msmpeg4_encode_dc(s, block[0], n, &dc_pred_dir);
+        i = 1;
+        if (n < 4) {
+            rl = &rl_table[s->rl_table_index];
+        } else {
+            rl = &rl_table[3 + s->rl_chroma_table_index];
+        }
+        run_diff = 0;
+        set_stat(ST_INTRA_AC);
+    } else {
+        i = 0;
+        rl = &rl_table[3 + s->rl_table_index];
+        run_diff = 1;
+        set_stat(ST_INTER_AC);
+    }
+
+    /* AC coefs */
+    last_index = s->block_last_index[n];
+    last_non_zero = i - 1;
+    for (; i <= last_index; i++) {
+	j = zigzag_direct[i];
+	level = block[j];
+	if (level) {
+	    run = i - last_non_zero - 1;
+	    last = (i == last_index);
+	    sign = 0;
+	    slevel = level;
+	    if (level < 0) {
+		sign = 1;
+		level = -level;
+	    }
+            code = get_rl_index(rl, last, run, level);
+            put_bits(&s->pb, rl->table_vlc[code][1], rl->table_vlc[code][0]);
+            if (code == rl->n) {
+                int level1, run1;
+
+                level1 = level - rl->max_level[last][run];
+                if (level1 < 1) 
+                    goto esc2;
+                code = get_rl_index(rl, last, run, level1);
+                if (code == rl->n) {
+                esc2:
+                    put_bits(&s->pb, 1, 0);
+                    if (level > MAX_LEVEL)
+                        goto esc3;
+                    run1 = run - rl->max_run[last][level] - run_diff;
+                    if (run1 < 0)
+                        goto esc3;
+                    code = get_rl_index(rl, last, run1, level);
+                    if (code == rl->n) {
+                    esc3:
+                        /* third escape */
+                        put_bits(&s->pb, 1, 0);
+                        put_bits(&s->pb, 1, last);
+                        put_bits(&s->pb, 6, run);
+                        put_bits(&s->pb, 8, slevel & 0xff);
+                    } else {
+                        /* second escape */
+                        put_bits(&s->pb, 1, 1);
+                        put_bits(&s->pb, rl->table_vlc[code][1], rl->table_vlc[code][0]);
+                        put_bits(&s->pb, 1, sign);
+                    }
+                } else {
+                    /* first escape */
+                    put_bits(&s->pb, 1, 1);
+                    put_bits(&s->pb, rl->table_vlc[code][1], rl->table_vlc[code][0]);
+                    put_bits(&s->pb, 1, sign);
+                }
+            } else {
+                put_bits(&s->pb, 1, sign);
+            }
+	    last_non_zero = i;
+	}
+    }
+}
+
+/****************************************/
+/* decoding stuff */
+
+static VLC mb_non_intra_vlc;
+static VLC mb_intra_vlc;
+static VLC dc_lum_vlc[2];
+static VLC dc_chroma_vlc[2];
+
+/* init all vlc decoding tables */
+int msmpeg4_decode_init_vlc(MpegEncContext *s)
+{
+    int i;
+    MVTable *mv;
+
+    for(i=0;i<NB_RL_TABLES;i++) {
+        init_rl(&rl_table[i]);
+        init_vlc_rl(&rl_table[i]);
+    }
+    for(i=0;i<2;i++) {
+        mv = &mv_tables[i];
+        init_vlc(&mv->vlc, 9, mv->n + 1, 
+                 mv->table_mv_bits, 1, 1,
+                 mv->table_mv_code, 2, 2);
+    }
+
+    init_vlc(&dc_lum_vlc[0], 9, 120, 
+             &table0_dc_lum[0][1], 8, 4,
+             &table0_dc_lum[0][0], 8, 4);
+    init_vlc(&dc_chroma_vlc[0], 9, 120, 
+             &table0_dc_chroma[0][1], 8, 4,
+             &table0_dc_chroma[0][0], 8, 4);
+    init_vlc(&dc_lum_vlc[1], 9, 120, 
+             &table1_dc_lum[0][1], 8, 4,
+             &table1_dc_lum[0][0], 8, 4);
+    init_vlc(&dc_chroma_vlc[1], 9, 120, 
+             &table1_dc_chroma[0][1], 8, 4,
+             &table1_dc_chroma[0][0], 8, 4);
+
+    init_vlc(&mb_non_intra_vlc, 9, 128, 
+             &table_mb_non_intra[0][1], 8, 4,
+             &table_mb_non_intra[0][0], 8, 4);
+    init_vlc(&mb_intra_vlc, 9, 128, 
+             &table_mb_intra[0][1], 4, 2,
+             &table_mb_intra[0][0], 4, 2);
+    return 0;
+}
+
+static int decode012(GetBitContext *gb)
+{
+    int n;
+    n = get_bits(gb, 1);
+    if (n == 0)
+        return 0;
+    else
+        return get_bits(gb, 1) + 1;
+}
+
+int msmpeg4_decode_picture_header(MpegEncContext * s)
+{
+    int code;
+
+    s->pict_type = get_bits(&s->gb, 2) + 1;
+    if (s->pict_type != I_TYPE &&
+        s->pict_type != P_TYPE)
+        return -1;
+
+    s->qscale = get_bits(&s->gb, 5);
+
+    if (s->pict_type == I_TYPE) {
+        code = get_bits(&s->gb, 5); 
+        /* 0x17: one slice, 0x18: three slices */
+        /* XXX: implement it */
+        if (code < 0x17)
+            return -1;
+        s->slice_height = s->mb_height / (code - 0x16);
+        s->rl_chroma_table_index = decode012(&s->gb);
+        s->rl_table_index = decode012(&s->gb);
+
+        s->dc_table_index = get_bits(&s->gb, 1);
+        s->no_rounding = 1;
+    } else {
+        s->use_skip_mb_code = get_bits(&s->gb, 1);
+        
+        s->rl_table_index = decode012(&s->gb);
+        s->rl_chroma_table_index = s->rl_table_index;
+
+        s->dc_table_index = get_bits(&s->gb, 1);
+
+        s->mv_table_index = get_bits(&s->gb, 1);
+        s->no_rounding ^= 1;
+    }
+#ifdef DEBUG
+    printf("*****frame %d:\n", frame_count++);
+#endif
+    return 0;
+}
+
+void memsetw(short *tab, int val, int n)
+{
+    int i;
+    for(i=0;i<n;i++)
+        tab[i] = val;
+}
+
+int msmpeg4_decode_mb(MpegEncContext *s, 
+                      DCTELEM block[6][64])
+{
+    int cbp, code, i;
+    int pred, val;
+    UINT8 *coded_val;
+
+    /* special slice handling */
+    if (s->mb_x == 0) {
+        if ((s->mb_y % s->slice_height) == 0) {
+            int wrap;
+            /* reset DC pred (set previous line to 1024) */
+            wrap = 2 * s->mb_width + 2;
+            memsetw(&s->dc_val[0][(1) + (2 * s->mb_y) * wrap], 
+                    1024, 2 * s->mb_width);
+            wrap = s->mb_width + 2;
+            memsetw(&s->dc_val[1][(1) + (s->mb_y) * wrap], 
+                    1024, s->mb_width);
+            memsetw(&s->dc_val[2][(1) + (s->mb_y) * wrap], 
+                    1024, s->mb_width);
+
+            s->first_slice_line = 1;
+        } else {
+            s->first_slice_line = 0; 
+        }
+    }
+
+    if (s->pict_type == P_TYPE) {
+        set_stat(ST_INTER_MB);
+        if (s->use_skip_mb_code) {
+            if (get_bits(&s->gb, 1)) {
+                /* skip mb */
+                s->mb_intra = 0;
+                for(i=0;i<6;i++)
+                    s->block_last_index[i] = -1;
+                s->mv_dir = MV_DIR_FORWARD;
+                s->mv_type = MV_TYPE_16X16;
+                s->mv[0][0][0] = 0;
+                s->mv[0][0][1] = 0;
+                return 0;
+            }
+        }
+        
+        code = get_vlc(&s->gb, &mb_non_intra_vlc);
+        if (code < 0)
+            return -1;
+        if (code & 0x40)
+            s->mb_intra = 0;
+        else
+            s->mb_intra = 1;
+            
+        cbp = code & 0x3f;
+    } else {
+        set_stat(ST_INTRA_MB);
+        s->mb_intra = 1;
+        code = get_vlc(&s->gb, &mb_intra_vlc);
+        if (code < 0)
+            return -1;
+        /* predict coded block pattern */
+        cbp = 0;
+        for(i=0;i<6;i++) {
+            val = ((code >> (5 - i)) & 1);
+            if (i < 4) {
+                pred = coded_block_pred(s, i, &coded_val);
+                val = val ^ pred;
+                *coded_val = val;
+            }
+            cbp |= val << (5 - i);
+        }
+    }
+
+    if (!s->mb_intra) {
+        int mx, my;
+        set_stat(ST_MV);
+        h263_pred_motion(s, 0, &mx, &my);
+        if (msmpeg4_decode_motion(s, &mx, &my) < 0)
+            return -1;
+        s->mv_dir = MV_DIR_FORWARD;
+        s->mv_type = MV_TYPE_16X16;
+        s->mv[0][0][0] = mx;
+        s->mv[0][0][1] = my;
+    } else {
+        set_stat(ST_INTRA_MB);
+        s->ac_pred = get_bits(&s->gb, 1);
+    }
+
+    for (i = 0; i < 6; i++) {
+        if (msmpeg4_decode_block(s, block[i], i, (cbp >> (5 - i)) & 1) < 0)
+            return -1;
+    }
+    return 0;
+}
+
+static int msmpeg4_decode_block(MpegEncContext * s, DCTELEM * block,
+                              int n, int coded)
+{
+    int code, level, i, j, last, run, run_diff;
+    int dc_pred_dir;
+    RLTable *rl;
+    const UINT8 *scan_table;
+
+    if (s->mb_intra) {
+	/* DC coef */
+        set_stat(ST_DC);
+        level = msmpeg4_decode_dc(s, n, &dc_pred_dir);
+        if (level < 0)
+            return -1;
+        block[0] = level;
+        if (n < 4) {
+            rl = &rl_table[s->rl_table_index];
+        } else {
+            rl = &rl_table[3 + s->rl_chroma_table_index];
+        }
+        run_diff = 0;
+	i = 1;
+        if (!coded) {
+            goto not_coded;
+        }
+        if (s->ac_pred) {
+            if (dc_pred_dir == 0) 
+                scan_table = ff_alternate_vertical_scan; /* left */
+            else
+                scan_table = ff_alternate_horizontal_scan; /* top */
+        } else {
+            scan_table = zigzag_direct;
+        }
+        set_stat(ST_INTRA_AC);
+    } else {
+	i = 0;
+        rl = &rl_table[3 + s->rl_table_index];
+        run_diff = 1;
+        if (!coded) {
+            s->block_last_index[n] = i - 1;
+            return 0;
+        }
+        scan_table = zigzag_direct;
+        set_stat(ST_INTER_AC);
+    }
+
+    for(;;) {
+        code = get_vlc(&s->gb, &rl->vlc);
+        if (code < 0)
+            return -1;
+        if (code == rl->n) {
+            /* escape */
+            if (get_bits(&s->gb, 1) == 0) {
+                if (get_bits(&s->gb, 1) == 0) {
+                    /* third escape */
+                    last = get_bits(&s->gb, 1);
+                    run = get_bits(&s->gb, 6);
+                    level = get_bits(&s->gb, 8);
+                    level = (level << 24) >> 24; /* sign extend */
+                } else {
+                    /* second escape */
+                    code = get_vlc(&s->gb, &rl->vlc);
+                    if (code < 0 || code >= rl->n)
+                        return -1;
+                    run = rl->table_run[code];
+                    level = rl->table_level[code];
+                    last = code >= rl->last;
+                    run += rl->max_run[last][level] + run_diff;
+                    if (get_bits(&s->gb, 1))
+                        level = -level;
+                }
+            } else {
+                /* first escape */
+                code = get_vlc(&s->gb, &rl->vlc);
+                if (code < 0 || code >= rl->n)
+                    return -1;
+                run = rl->table_run[code];
+                level = rl->table_level[code];
+                last = code >= rl->last;
+                level += rl->max_level[last][run];
+                if (get_bits(&s->gb, 1))
+                    level = -level;
+            }
+        } else {
+            run = rl->table_run[code];
+            level = rl->table_level[code];
+            last = code >= rl->last;
+            if (get_bits(&s->gb, 1))
+                level = -level;
+        }
+        i += run;
+        if (i >= 64)
+            return -1;
+	j = scan_table[i];
+        block[j] = level;
+        i++;
+        if (last)
+            break;
+    }
+ not_coded:
+    if (s->mb_intra) {
+        mpeg4_pred_ac(s, block, n, dc_pred_dir);
+        if (s->ac_pred) {
+            i = 64; /* XXX: not optimal */
+        }
+    }
+    s->block_last_index[n] = i - 1;
+
+    return 0;
+}
+
+static int msmpeg4_decode_dc(MpegEncContext * s, int n, int *dir_ptr)
+{
+    int level, pred;
+    UINT16 *dc_val;
+
+    if (n < 4) {
+        level = get_vlc(&s->gb, &dc_lum_vlc[s->dc_table_index]);
+    } else {
+        level = get_vlc(&s->gb, &dc_chroma_vlc[s->dc_table_index]);
+    }
+    if (level < 0)
+        return -1;
+
+    if (level == DC_MAX) {
+        level = get_bits(&s->gb, 8);
+        if (get_bits(&s->gb, 1))
+            level = -level;
+    } else if (level != 0) {
+        if (get_bits(&s->gb, 1))
+            level = -level;
+    }
+
+    pred = msmpeg4_pred_dc(s, n, &dc_val, dir_ptr);
+    level += pred;
+
+    /* update predictor */
+    if (n < 4) {
+        *dc_val = level * s->y_dc_scale;
+    } else {
+        *dc_val = level * s->c_dc_scale;
+    }
+
+    return level;
+}
+
+static int msmpeg4_decode_motion(MpegEncContext * s, 
+                                 int *mx_ptr, int *my_ptr)
+{
+    MVTable *mv;
+    int code, mx, my;
+
+    mv = &mv_tables[s->mv_table_index];
+
+    code = get_vlc(&s->gb, &mv->vlc);
+    if (code < 0)
+        return -1;
+    if (code == mv->n) {
+        mx = get_bits(&s->gb, 6);
+        my = get_bits(&s->gb, 6);
+    } else {
+        mx = mv->table_mvx[code];
+        my = mv->table_mvy[code];
+    }
+
+    mx += *mx_ptr - 32;
+    my += *my_ptr - 32;
+    /* WARNING : they do not do exactly modulo encoding */
+    if (mx <= -64)
+        mx += 64;
+    else if (mx >= 64)
+        mx -= 64;
+
+    if (my <= -64)
+        my += 64;
+    else if (my >= 64)
+        my -= 64;
+    *mx_ptr = mx;
+    *my_ptr = my;
+    return 0;
+}
diff --git a/libavcodec/msmpeg4data.h b/libavcodec/msmpeg4data.h
new file mode 100644
index 0000000000..03a2612114
--- /dev/null
+++ b/libavcodec/msmpeg4data.h
@@ -0,0 +1,1767 @@
+/*
+ * MSMPEG4 data tables. 
+ */
+
+/* intra picture macro block coded block pattern */
+const UINT16 table_mb_intra[64][2] = {
+{ 0x1, 1 },{ 0x17, 6 },{ 0x9, 5 },{ 0x5, 5 },
+{ 0x6, 5 },{ 0x47, 9 },{ 0x20, 7 },{ 0x10, 7 },
+{ 0x2, 5 },{ 0x7c, 9 },{ 0x3a, 7 },{ 0x1d, 7 },
+{ 0x2, 6 },{ 0xec, 9 },{ 0x77, 8 },{ 0x0, 8 },
+{ 0x3, 5 },{ 0xb7, 9 },{ 0x2c, 7 },{ 0x13, 7 },
+{ 0x1, 6 },{ 0x168, 10 },{ 0x46, 8 },{ 0x3f, 8 },
+{ 0x1e, 6 },{ 0x712, 13 },{ 0xb5, 9 },{ 0x42, 8 },
+{ 0x22, 7 },{ 0x1c5, 11 },{ 0x11e, 10 },{ 0x87, 9 },
+{ 0x6, 4 },{ 0x3, 9 },{ 0x1e, 7 },{ 0x1c, 6 },
+{ 0x12, 7 },{ 0x388, 12 },{ 0x44, 9 },{ 0x70, 9 },
+{ 0x1f, 6 },{ 0x23e, 11 },{ 0x39, 8 },{ 0x8e, 9 },
+{ 0x1, 7 },{ 0x1c6, 11 },{ 0xb6, 9 },{ 0x45, 9 },
+{ 0x14, 6 },{ 0x23f, 11 },{ 0x7d, 9 },{ 0x18, 9 },
+{ 0x7, 7 },{ 0x1c7, 11 },{ 0x86, 9 },{ 0x19, 9 },
+{ 0x15, 6 },{ 0x1db, 10 },{ 0x2, 9 },{ 0x46, 9 },
+{ 0xd, 8 },{ 0x713, 13 },{ 0x1da, 10 },{ 0x169, 10 },
+};
+
+/* non intra picture macro block coded block pattern + mb type */
+const UINT32 table_mb_non_intra[128][2] = {
+{ 0x40, 7 },{ 0x13c9, 13 },{ 0x9fd, 12 },{ 0x1fc, 15 },
+{ 0x9fc, 12 },{ 0xa83, 18 },{ 0x12d34, 17 },{ 0x83bc, 16 },
+{ 0x83a, 12 },{ 0x7f8, 17 },{ 0x3fd, 16 },{ 0x3ff, 16 },
+{ 0x79, 13 },{ 0xa82, 18 },{ 0x969d, 16 },{ 0x2a4, 16 },
+{ 0x978, 12 },{ 0x543, 17 },{ 0x41df, 15 },{ 0x7f9, 17 },
+{ 0x12f3, 13 },{ 0x25a6b, 18 },{ 0x25ef9, 18 },{ 0x3fa, 16 },
+{ 0x20ee, 14 },{ 0x969ab, 20 },{ 0x969c, 16 },{ 0x25ef8, 18 },
+{ 0x12d2, 13 },{ 0xa85, 18 },{ 0x969e, 16 },{ 0x4bc8, 15 },
+{ 0x3d, 12 },{ 0x12f7f, 17 },{ 0x2a2, 16 },{ 0x969f, 16 },
+{ 0x25ee, 14 },{ 0x12d355, 21 },{ 0x12f7d, 17 },{ 0x12f7e, 17 },
+{ 0x9e5, 12 },{ 0xa81, 18 },{ 0x4b4d4, 19 },{ 0x83bd, 16 },
+{ 0x78, 13 },{ 0x969b, 16 },{ 0x3fe, 16 },{ 0x2a5, 16 },
+{ 0x7e, 13 },{ 0xa80, 18 },{ 0x2a3, 16 },{ 0x3fb, 16 },
+{ 0x1076, 13 },{ 0xa84, 18 },{ 0x153, 15 },{ 0x4bc9, 15 },
+{ 0x55, 13 },{ 0x12d354, 21 },{ 0x4bde, 15 },{ 0x25e5, 14 },
+{ 0x25b, 10 },{ 0x4b4c, 15 },{ 0x96b, 12 },{ 0x96a, 12 },
+{ 0x1, 2 },{ 0x0, 7 },{ 0x26, 6 },{ 0x12b, 9 },
+{ 0x7, 3 },{ 0x20f, 10 },{ 0x4, 9 },{ 0x28, 12 },
+{ 0x6, 3 },{ 0x20a, 10 },{ 0x128, 9 },{ 0x2b, 12 },
+{ 0x11, 5 },{ 0x1b, 11 },{ 0x13a, 9 },{ 0x4ff, 11 },
+{ 0x3, 4 },{ 0x277, 10 },{ 0x106, 9 },{ 0x839, 12 },
+{ 0xb, 4 },{ 0x27b, 10 },{ 0x12c, 9 },{ 0x4bf, 11 },
+{ 0x9, 6 },{ 0x35, 12 },{ 0x27e, 10 },{ 0x13c8, 13 },
+{ 0x1, 6 },{ 0x4aa, 11 },{ 0x208, 10 },{ 0x29, 12 },
+{ 0x1, 4 },{ 0x254, 10 },{ 0x12e, 9 },{ 0x838, 12 },
+{ 0x24, 6 },{ 0x4f3, 11 },{ 0x276, 10 },{ 0x12f6, 13 },
+{ 0x1, 5 },{ 0x27a, 10 },{ 0x13e, 9 },{ 0x3e, 12 },
+{ 0x8, 6 },{ 0x413, 11 },{ 0xc, 10 },{ 0x4be, 11 },
+{ 0x14, 5 },{ 0x412, 11 },{ 0x253, 10 },{ 0x97a, 12 },
+{ 0x21, 6 },{ 0x4ab, 11 },{ 0x20b, 10 },{ 0x34, 12 },
+{ 0x15, 5 },{ 0x278, 10 },{ 0x252, 10 },{ 0x968, 12 },
+{ 0x5, 5 },{ 0xb, 10 },{ 0x9c, 8 },{ 0xe, 10 },
+};
+
+/* dc table 0 */
+
+static const UINT32 table0_dc_lum[120][2] = {
+{ 0x1, 1 },{ 0x1, 2 },{ 0x1, 4 },{ 0x1, 5 },
+{ 0x5, 5 },{ 0x7, 5 },{ 0x8, 6 },{ 0xc, 6 },
+{ 0x0, 7 },{ 0x2, 7 },{ 0x12, 7 },{ 0x1a, 7 },
+{ 0x3, 8 },{ 0x7, 8 },{ 0x27, 8 },{ 0x37, 8 },
+{ 0x5, 9 },{ 0x4c, 9 },{ 0x6c, 9 },{ 0x6d, 9 },
+{ 0x8, 10 },{ 0x19, 10 },{ 0x9b, 10 },{ 0x1b, 10 },
+{ 0x9a, 10 },{ 0x13, 11 },{ 0x34, 11 },{ 0x35, 11 },
+{ 0x61, 12 },{ 0x48, 13 },{ 0xc4, 13 },{ 0x4a, 13 },
+{ 0xc6, 13 },{ 0xc7, 13 },{ 0x92, 14 },{ 0x18b, 14 },
+{ 0x93, 14 },{ 0x183, 14 },{ 0x182, 14 },{ 0x96, 14 },
+{ 0x97, 14 },{ 0x180, 14 },{ 0x314, 15 },{ 0x315, 15 },
+{ 0x605, 16 },{ 0x604, 16 },{ 0x606, 16 },{ 0xc0e, 17 },
+{ 0x303cd, 23 },{ 0x303c9, 23 },{ 0x303c8, 23 },{ 0x303ca, 23 },
+{ 0x303cb, 23 },{ 0x303cc, 23 },{ 0x303ce, 23 },{ 0x303cf, 23 },
+{ 0x303d0, 23 },{ 0x303d1, 23 },{ 0x303d2, 23 },{ 0x303d3, 23 },
+{ 0x303d4, 23 },{ 0x303d5, 23 },{ 0x303d6, 23 },{ 0x303d7, 23 },
+{ 0x303d8, 23 },{ 0x303d9, 23 },{ 0x303da, 23 },{ 0x303db, 23 },
+{ 0x303dc, 23 },{ 0x303dd, 23 },{ 0x303de, 23 },{ 0x303df, 23 },
+{ 0x303e0, 23 },{ 0x303e1, 23 },{ 0x303e2, 23 },{ 0x303e3, 23 },
+{ 0x303e4, 23 },{ 0x303e5, 23 },{ 0x303e6, 23 },{ 0x303e7, 23 },
+{ 0x303e8, 23 },{ 0x303e9, 23 },{ 0x303ea, 23 },{ 0x303eb, 23 },
+{ 0x303ec, 23 },{ 0x303ed, 23 },{ 0x303ee, 23 },{ 0x303ef, 23 },
+{ 0x303f0, 23 },{ 0x303f1, 23 },{ 0x303f2, 23 },{ 0x303f3, 23 },
+{ 0x303f4, 23 },{ 0x303f5, 23 },{ 0x303f6, 23 },{ 0x303f7, 23 },
+{ 0x303f8, 23 },{ 0x303f9, 23 },{ 0x303fa, 23 },{ 0x303fb, 23 },
+{ 0x303fc, 23 },{ 0x303fd, 23 },{ 0x303fe, 23 },{ 0x303ff, 23 },
+{ 0x60780, 24 },{ 0x60781, 24 },{ 0x60782, 24 },{ 0x60783, 24 },
+{ 0x60784, 24 },{ 0x60785, 24 },{ 0x60786, 24 },{ 0x60787, 24 },
+{ 0x60788, 24 },{ 0x60789, 24 },{ 0x6078a, 24 },{ 0x6078b, 24 },
+{ 0x6078c, 24 },{ 0x6078d, 24 },{ 0x6078e, 24 },{ 0x6078f, 24 },
+};
+
+static const UINT32 table0_dc_chroma[120][2] = {
+{ 0x0, 2 },{ 0x1, 2 },{ 0x5, 3 },{ 0x9, 4 },
+{ 0xd, 4 },{ 0x11, 5 },{ 0x1d, 5 },{ 0x1f, 5 },
+{ 0x21, 6 },{ 0x31, 6 },{ 0x38, 6 },{ 0x33, 6 },
+{ 0x39, 6 },{ 0x3d, 6 },{ 0x61, 7 },{ 0x79, 7 },
+{ 0x80, 8 },{ 0xc8, 8 },{ 0xca, 8 },{ 0xf0, 8 },
+{ 0x81, 8 },{ 0xc0, 8 },{ 0xc9, 8 },{ 0x107, 9 },
+{ 0x106, 9 },{ 0x196, 9 },{ 0x183, 9 },{ 0x1e3, 9 },
+{ 0x1e2, 9 },{ 0x20a, 10 },{ 0x20b, 10 },{ 0x609, 11 },
+{ 0x412, 11 },{ 0x413, 11 },{ 0x60b, 11 },{ 0x411, 11 },
+{ 0x60a, 11 },{ 0x65f, 11 },{ 0x410, 11 },{ 0x65d, 11 },
+{ 0x65e, 11 },{ 0xcb8, 12 },{ 0xc10, 12 },{ 0xcb9, 12 },
+{ 0x1823, 13 },{ 0x3045, 14 },{ 0x6089, 15 },{ 0xc110, 16 },
+{ 0x304448, 22 },{ 0x304449, 22 },{ 0x30444a, 22 },{ 0x30444b, 22 },
+{ 0x30444c, 22 },{ 0x30444d, 22 },{ 0x30444e, 22 },{ 0x30444f, 22 },
+{ 0x304450, 22 },{ 0x304451, 22 },{ 0x304452, 22 },{ 0x304453, 22 },
+{ 0x304454, 22 },{ 0x304455, 22 },{ 0x304456, 22 },{ 0x304457, 22 },
+{ 0x304458, 22 },{ 0x304459, 22 },{ 0x30445a, 22 },{ 0x30445b, 22 },
+{ 0x30445c, 22 },{ 0x30445d, 22 },{ 0x30445e, 22 },{ 0x30445f, 22 },
+{ 0x304460, 22 },{ 0x304461, 22 },{ 0x304462, 22 },{ 0x304463, 22 },
+{ 0x304464, 22 },{ 0x304465, 22 },{ 0x304466, 22 },{ 0x304467, 22 },
+{ 0x304468, 22 },{ 0x304469, 22 },{ 0x30446a, 22 },{ 0x30446b, 22 },
+{ 0x30446c, 22 },{ 0x30446d, 22 },{ 0x30446e, 22 },{ 0x30446f, 22 },
+{ 0x304470, 22 },{ 0x304471, 22 },{ 0x304472, 22 },{ 0x304473, 22 },
+{ 0x304474, 22 },{ 0x304475, 22 },{ 0x304476, 22 },{ 0x304477, 22 },
+{ 0x304478, 22 },{ 0x304479, 22 },{ 0x30447a, 22 },{ 0x30447b, 22 },
+{ 0x30447c, 22 },{ 0x30447d, 22 },{ 0x30447e, 22 },{ 0x30447f, 22 },
+{ 0x608880, 23 },{ 0x608881, 23 },{ 0x608882, 23 },{ 0x608883, 23 },
+{ 0x608884, 23 },{ 0x608885, 23 },{ 0x608886, 23 },{ 0x608887, 23 },
+{ 0x608888, 23 },{ 0x608889, 23 },{ 0x60888a, 23 },{ 0x60888b, 23 },
+{ 0x60888c, 23 },{ 0x60888d, 23 },{ 0x60888e, 23 },{ 0x60888f, 23 },
+};
+
+/* dc table 1 */
+
+const UINT32 table1_dc_lum[120][2] = {
+{ 0x2, 2 },{ 0x3, 2 },{ 0x3, 3 },{ 0x2, 4 },
+{ 0x5, 4 },{ 0x1, 5 },{ 0x3, 5 },{ 0x8, 5 },
+{ 0x0, 6 },{ 0x5, 6 },{ 0xd, 6 },{ 0xf, 6 },
+{ 0x13, 6 },{ 0x8, 7 },{ 0x18, 7 },{ 0x1c, 7 },
+{ 0x24, 7 },{ 0x4, 8 },{ 0x6, 8 },{ 0x12, 8 },
+{ 0x32, 8 },{ 0x3b, 8 },{ 0x4a, 8 },{ 0x4b, 8 },
+{ 0xb, 9 },{ 0x26, 9 },{ 0x27, 9 },{ 0x66, 9 },
+{ 0x74, 9 },{ 0x75, 9 },{ 0x14, 10 },{ 0x1c, 10 },
+{ 0x1f, 10 },{ 0x1d, 10 },{ 0x2b, 11 },{ 0x3d, 11 },
+{ 0x19d, 11 },{ 0x19f, 11 },{ 0x54, 12 },{ 0x339, 12 },
+{ 0x338, 12 },{ 0x33d, 12 },{ 0xab, 13 },{ 0xf1, 13 },
+{ 0x678, 13 },{ 0xf2, 13 },{ 0x1e0, 14 },{ 0x1e1, 14 },
+{ 0x154, 14 },{ 0xcf2, 14 },{ 0x3cc, 15 },{ 0x2ab, 15 },
+{ 0x19e7, 15 },{ 0x3ce, 15 },{ 0x19e6, 15 },{ 0x554, 16 },
+{ 0x79f, 16 },{ 0x555, 16 },{ 0xf3d, 17 },{ 0xf37, 17 },
+{ 0xf3c, 17 },{ 0xf35, 17 },{ 0x1e6d, 18 },{ 0x1e68, 18 },
+{ 0x3cd8, 19 },{ 0x3cd3, 19 },{ 0x3cd9, 19 },{ 0x79a4, 20 },
+{ 0xf34ba, 25 },{ 0xf34b4, 25 },{ 0xf34b5, 25 },{ 0xf34b6, 25 },
+{ 0xf34b7, 25 },{ 0xf34b8, 25 },{ 0xf34b9, 25 },{ 0xf34bb, 25 },
+{ 0xf34bc, 25 },{ 0xf34bd, 25 },{ 0xf34be, 25 },{ 0xf34bf, 25 },
+{ 0x1e6940, 26 },{ 0x1e6941, 26 },{ 0x1e6942, 26 },{ 0x1e6943, 26 },
+{ 0x1e6944, 26 },{ 0x1e6945, 26 },{ 0x1e6946, 26 },{ 0x1e6947, 26 },
+{ 0x1e6948, 26 },{ 0x1e6949, 26 },{ 0x1e694a, 26 },{ 0x1e694b, 26 },
+{ 0x1e694c, 26 },{ 0x1e694d, 26 },{ 0x1e694e, 26 },{ 0x1e694f, 26 },
+{ 0x1e6950, 26 },{ 0x1e6951, 26 },{ 0x1e6952, 26 },{ 0x1e6953, 26 },
+{ 0x1e6954, 26 },{ 0x1e6955, 26 },{ 0x1e6956, 26 },{ 0x1e6957, 26 },
+{ 0x1e6958, 26 },{ 0x1e6959, 26 },{ 0x1e695a, 26 },{ 0x1e695b, 26 },
+{ 0x1e695c, 26 },{ 0x1e695d, 26 },{ 0x1e695e, 26 },{ 0x1e695f, 26 },
+{ 0x1e6960, 26 },{ 0x1e6961, 26 },{ 0x1e6962, 26 },{ 0x1e6963, 26 },
+{ 0x1e6964, 26 },{ 0x1e6965, 26 },{ 0x1e6966, 26 },{ 0x1e6967, 26 },
+};
+
+const UINT32 table1_dc_chroma[120][2] = {
+{ 0x0, 2 },{ 0x1, 2 },{ 0x4, 3 },{ 0x7, 3 },
+{ 0xb, 4 },{ 0xd, 4 },{ 0x15, 5 },{ 0x28, 6 },
+{ 0x30, 6 },{ 0x32, 6 },{ 0x52, 7 },{ 0x62, 7 },
+{ 0x66, 7 },{ 0xa6, 8 },{ 0xc6, 8 },{ 0xcf, 8 },
+{ 0x14f, 9 },{ 0x18e, 9 },{ 0x19c, 9 },{ 0x29d, 10 },
+{ 0x33a, 10 },{ 0x538, 11 },{ 0x63c, 11 },{ 0x63e, 11 },
+{ 0x63f, 11 },{ 0x676, 11 },{ 0xa73, 12 },{ 0xc7a, 12 },
+{ 0xcef, 12 },{ 0x14e5, 13 },{ 0x19dd, 13 },{ 0x29c8, 14 },
+{ 0x29c9, 14 },{ 0x63dd, 15 },{ 0x33b8, 14 },{ 0x33b9, 14 },
+{ 0xc7b6, 16 },{ 0x63d8, 15 },{ 0x63df, 15 },{ 0xc7b3, 16 },
+{ 0xc7b4, 16 },{ 0xc7b5, 16 },{ 0x63de, 15 },{ 0xc7b7, 16 },
+{ 0xc7b8, 16 },{ 0xc7b9, 16 },{ 0x18f65, 17 },{ 0x31ec8, 18 },
+{ 0xc7b248, 24 },{ 0xc7b249, 24 },{ 0xc7b24a, 24 },{ 0xc7b24b, 24 },
+{ 0xc7b24c, 24 },{ 0xc7b24d, 24 },{ 0xc7b24e, 24 },{ 0xc7b24f, 24 },
+{ 0xc7b250, 24 },{ 0xc7b251, 24 },{ 0xc7b252, 24 },{ 0xc7b253, 24 },
+{ 0xc7b254, 24 },{ 0xc7b255, 24 },{ 0xc7b256, 24 },{ 0xc7b257, 24 },
+{ 0xc7b258, 24 },{ 0xc7b259, 24 },{ 0xc7b25a, 24 },{ 0xc7b25b, 24 },
+{ 0xc7b25c, 24 },{ 0xc7b25d, 24 },{ 0xc7b25e, 24 },{ 0xc7b25f, 24 },
+{ 0xc7b260, 24 },{ 0xc7b261, 24 },{ 0xc7b262, 24 },{ 0xc7b263, 24 },
+{ 0xc7b264, 24 },{ 0xc7b265, 24 },{ 0xc7b266, 24 },{ 0xc7b267, 24 },
+{ 0xc7b268, 24 },{ 0xc7b269, 24 },{ 0xc7b26a, 24 },{ 0xc7b26b, 24 },
+{ 0xc7b26c, 24 },{ 0xc7b26d, 24 },{ 0xc7b26e, 24 },{ 0xc7b26f, 24 },
+{ 0xc7b270, 24 },{ 0xc7b271, 24 },{ 0xc7b272, 24 },{ 0xc7b273, 24 },
+{ 0xc7b274, 24 },{ 0xc7b275, 24 },{ 0xc7b276, 24 },{ 0xc7b277, 24 },
+{ 0xc7b278, 24 },{ 0xc7b279, 24 },{ 0xc7b27a, 24 },{ 0xc7b27b, 24 },
+{ 0xc7b27c, 24 },{ 0xc7b27d, 24 },{ 0xc7b27e, 24 },{ 0xc7b27f, 24 },
+{ 0x18f6480, 25 },{ 0x18f6481, 25 },{ 0x18f6482, 25 },{ 0x18f6483, 25 },
+{ 0x18f6484, 25 },{ 0x18f6485, 25 },{ 0x18f6486, 25 },{ 0x18f6487, 25 },
+{ 0x18f6488, 25 },{ 0x18f6489, 25 },{ 0x18f648a, 25 },{ 0x18f648b, 25 },
+{ 0x18f648c, 25 },{ 0x18f648d, 25 },{ 0x18f648e, 25 },{ 0x18f648f, 25 },
+};
+
+/* vlc table 0, for intra luma */
+
+static const UINT16 table0_vlc[133][2] = {
+{ 0x1, 2 },{ 0x6, 3 },{ 0xf, 4 },{ 0x16, 5 },
+{ 0x20, 6 },{ 0x18, 7 },{ 0x8, 8 },{ 0x9a, 8 },
+{ 0x56, 9 },{ 0x13e, 9 },{ 0xf0, 10 },{ 0x3a5, 10 },
+{ 0x77, 11 },{ 0x1ef, 11 },{ 0x9a, 12 },{ 0x5d, 13 },
+{ 0x1, 4 },{ 0x11, 5 },{ 0x2, 7 },{ 0xb, 8 },
+{ 0x12, 9 },{ 0x1d6, 9 },{ 0x27e, 10 },{ 0x191, 11 },
+{ 0xea, 12 },{ 0x3dc, 12 },{ 0x13b, 13 },{ 0x4, 5 },
+{ 0x14, 7 },{ 0x9e, 8 },{ 0x9, 10 },{ 0x1ac, 11 },
+{ 0x1e2, 11 },{ 0x3ca, 12 },{ 0x5f, 13 },{ 0x17, 5 },
+{ 0x4e, 7 },{ 0x5e, 9 },{ 0xf3, 10 },{ 0x1ad, 11 },
+{ 0xec, 12 },{ 0x5f0, 13 },{ 0xe, 6 },{ 0xe1, 8 },
+{ 0x3a4, 10 },{ 0x9c, 12 },{ 0x13d, 13 },{ 0x3b, 6 },
+{ 0x1c, 9 },{ 0x14, 11 },{ 0x9be, 12 },{ 0x6, 7 },
+{ 0x7a, 9 },{ 0x190, 11 },{ 0x137, 13 },{ 0x1b, 7 },
+{ 0x8, 10 },{ 0x75c, 11 },{ 0x71, 7 },{ 0xd7, 10 },
+{ 0x9bf, 12 },{ 0x7, 8 },{ 0xaf, 10 },{ 0x4cc, 11 },
+{ 0x34, 8 },{ 0x265, 10 },{ 0x9f, 12 },{ 0xe0, 8 },
+{ 0x16, 11 },{ 0x327, 12 },{ 0x15, 9 },{ 0x17d, 11 },
+{ 0xebb, 12 },{ 0x14, 9 },{ 0xf6, 10 },{ 0x1e4, 11 },
+{ 0xcb, 10 },{ 0x99d, 12 },{ 0xca, 10 },{ 0x2fc, 12 },
+{ 0x17f, 11 },{ 0x4cd, 11 },{ 0x2fd, 12 },{ 0x4fe, 11 },
+{ 0x13a, 13 },{ 0xa, 4 },{ 0x42, 7 },{ 0x1d3, 9 },
+{ 0x4dd, 11 },{ 0x12, 5 },{ 0xe8, 8 },{ 0x4c, 11 },
+{ 0x136, 13 },{ 0x39, 6 },{ 0x264, 10 },{ 0xeba, 12 },
+{ 0x0, 7 },{ 0xae, 10 },{ 0x99c, 12 },{ 0x1f, 7 },
+{ 0x4de, 11 },{ 0x43, 7 },{ 0x4dc, 11 },{ 0x3, 8 },
+{ 0x3cb, 12 },{ 0x6, 8 },{ 0x99e, 12 },{ 0x2a, 8 },
+{ 0x5f1, 13 },{ 0xf, 8 },{ 0x9fe, 12 },{ 0x33, 8 },
+{ 0x9ff, 12 },{ 0x98, 8 },{ 0x99f, 12 },{ 0xea, 8 },
+{ 0x13c, 13 },{ 0x2e, 8 },{ 0x192, 11 },{ 0x136, 9 },
+{ 0x6a, 9 },{ 0x15, 11 },{ 0x3af, 10 },{ 0x1e3, 11 },
+{ 0x74, 11 },{ 0xeb, 12 },{ 0x2f9, 12 },{ 0x5c, 13 },
+{ 0xed, 12 },{ 0x3dd, 12 },{ 0x326, 12 },{ 0x5e, 13 },
+{ 0x16, 7 },
+};
+
+const INT8 table0_level[132] = {
+  1,  2,  3,  4,  5,  6,  7,  8,
+  9, 10, 11, 12, 13, 14, 15, 16,
+  1,  2,  3,  4,  5,  6,  7,  8,
+  9, 10, 11,  1,  2,  3,  4,  5,
+  6,  7,  8,  1,  2,  3,  4,  5,
+  6,  7,  1,  2,  3,  4,  5,  1,
+  2,  3,  4,  1,  2,  3,  4,  1,
+  2,  3,  1,  2,  3,  1,  2,  3,
+  1,  2,  3,  1,  2,  3,  1,  2,
+  3,  1,  2,  3,  1,  2,  1,  2,
+  1,  1,  1,  1,  1,  1,  2,  3,
+  4,  1,  2,  3,  4,  1,  2,  3,
+  1,  2,  3,  1,  2,  1,  2,  1,
+  2,  1,  2,  1,  2,  1,  2,  1,
+  2,  1,  2,  1,  2,  1,  2,  1,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,
+};
+
+const INT8 table0_run[132] = {
+  0,  0,  0,  0,  0,  0,  0,  0,
+  0,  0,  0,  0,  0,  0,  0,  0,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  2,  2,  2,  2,  2,
+  2,  2,  2,  3,  3,  3,  3,  3,
+  3,  3,  4,  4,  4,  4,  4,  5,
+  5,  5,  5,  6,  6,  6,  6,  7,
+  7,  7,  8,  8,  8,  9,  9,  9,
+ 10, 10, 10, 11, 11, 11, 12, 12,
+ 12, 13, 13, 13, 14, 14, 15, 15,
+ 16, 17, 18, 19, 20,  0,  0,  0,
+  0,  1,  1,  1,  1,  2,  2,  2,
+  3,  3,  3,  4,  4,  5,  5,  6,
+  6,  7,  7,  8,  8,  9,  9, 10,
+ 10, 11, 11, 12, 12, 13, 13, 14,
+ 15, 16, 17, 18, 19, 20, 21, 22,
+ 23, 24, 25, 26,
+};
+
+/* vlc table 1, for intra chroma and P macro blocks */
+
+const UINT16 table1_vlc[149][2] = {
+{ 0x4, 3 },{ 0x14, 5 },{ 0x17, 7 },{ 0x7f, 8 },
+{ 0x154, 9 },{ 0x1f2, 10 },{ 0xbf, 11 },{ 0x65, 12 },
+{ 0xaaa, 12 },{ 0x630, 13 },{ 0x1597, 13 },{ 0x3b7, 14 },
+{ 0x2b22, 14 },{ 0xbe6, 15 },{ 0xb, 4 },{ 0x37, 7 },
+{ 0x62, 9 },{ 0x7, 11 },{ 0x166, 12 },{ 0xce, 13 },
+{ 0x1590, 13 },{ 0x5f6, 14 },{ 0xbe7, 15 },{ 0x7, 5 },
+{ 0x6d, 8 },{ 0x3, 11 },{ 0x31f, 12 },{ 0x5f2, 14 },
+{ 0x2, 6 },{ 0x61, 9 },{ 0x55, 12 },{ 0x1df, 14 },
+{ 0x1a, 6 },{ 0x1e, 10 },{ 0xac9, 12 },{ 0x2b23, 14 },
+{ 0x1e, 6 },{ 0x1f, 10 },{ 0xac3, 12 },{ 0x2b2b, 14 },
+{ 0x6, 7 },{ 0x4, 11 },{ 0x2f8, 13 },{ 0x19, 7 },
+{ 0x6, 11 },{ 0x63d, 13 },{ 0x57, 7 },{ 0x182, 11 },
+{ 0x2aa2, 14 },{ 0x4, 8 },{ 0x180, 11 },{ 0x59c, 14 },
+{ 0x7d, 8 },{ 0x164, 12 },{ 0x76d, 15 },{ 0x2, 9 },
+{ 0x18d, 11 },{ 0x1581, 13 },{ 0xad, 8 },{ 0x60, 12 },
+{ 0xc67, 14 },{ 0x1c, 9 },{ 0xee, 13 },{ 0x3, 9 },
+{ 0x2cf, 13 },{ 0xd9, 9 },{ 0x1580, 13 },{ 0x2, 11 },
+{ 0x183, 11 },{ 0x57, 12 },{ 0x61, 12 },{ 0x31, 11 },
+{ 0x66, 12 },{ 0x631, 13 },{ 0x632, 13 },{ 0xac, 13 },
+{ 0x31d, 12 },{ 0x76, 12 },{ 0x3a, 11 },{ 0x165, 12 },
+{ 0xc66, 14 },{ 0x3, 2 },{ 0x54, 7 },{ 0x2ab, 10 },
+{ 0x16, 13 },{ 0x5f7, 14 },{ 0x5, 4 },{ 0xf8, 9 },
+{ 0xaa9, 12 },{ 0x5f, 15 },{ 0x4, 4 },{ 0x1c, 10 },
+{ 0x1550, 13 },{ 0x4, 5 },{ 0x77, 11 },{ 0x76c, 15 },
+{ 0xe, 5 },{ 0xa, 12 },{ 0xc, 5 },{ 0x562, 11 },
+{ 0x4, 6 },{ 0x31c, 12 },{ 0x6, 6 },{ 0xc8, 13 },
+{ 0xd, 6 },{ 0x1da, 13 },{ 0x7, 6 },{ 0xc9, 13 },
+{ 0x1, 7 },{ 0x2e, 14 },{ 0x14, 7 },{ 0x1596, 13 },
+{ 0xa, 7 },{ 0xac2, 12 },{ 0x16, 7 },{ 0x15b, 14 },
+{ 0x15, 7 },{ 0x15a, 14 },{ 0xf, 8 },{ 0x5e, 15 },
+{ 0x7e, 8 },{ 0xab, 8 },{ 0x2d, 9 },{ 0xd8, 9 },
+{ 0xb, 9 },{ 0x14, 10 },{ 0x2b3, 10 },{ 0x1f3, 10 },
+{ 0x3a, 10 },{ 0x0, 10 },{ 0x58, 10 },{ 0x2e, 9 },
+{ 0x5e, 10 },{ 0x563, 11 },{ 0xec, 12 },{ 0x54, 12 },
+{ 0xac1, 12 },{ 0x1556, 13 },{ 0x2fa, 13 },{ 0x181, 11 },
+{ 0x1557, 13 },{ 0x59d, 14 },{ 0x2aa3, 14 },{ 0x2b2a, 14 },
+{ 0x1de, 14 },{ 0x63c, 13 },{ 0xcf, 13 },{ 0x1594, 13 },
+{ 0xd, 9 },
+};
+
+const INT8 table1_level[148] = {
+  1,  2,  3,  4,  5,  6,  7,  8,
+  9, 10, 11, 12, 13, 14,  1,  2,
+  3,  4,  5,  6,  7,  8,  9,  1,
+  2,  3,  4,  5,  1,  2,  3,  4,
+  1,  2,  3,  4,  1,  2,  3,  4,
+  1,  2,  3,  1,  2,  3,  1,  2,
+  3,  1,  2,  3,  1,  2,  3,  1,
+  2,  3,  1,  2,  3,  1,  2,  1,
+  2,  1,  2,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  2,  3,  4,  5,  1,  2,
+  3,  4,  1,  2,  3,  1,  2,  3,
+  1,  2,  1,  2,  1,  2,  1,  2,
+  1,  2,  1,  2,  1,  2,  1,  2,
+  1,  2,  1,  2,  1,  2,  1,  2,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,
+};
+
+const INT8 table1_run[148] = {
+  0,  0,  0,  0,  0,  0,  0,  0,
+  0,  0,  0,  0,  0,  0,  1,  1,
+  1,  1,  1,  1,  1,  1,  1,  2,
+  2,  2,  2,  2,  3,  3,  3,  3,
+  4,  4,  4,  4,  5,  5,  5,  5,
+  6,  6,  6,  7,  7,  7,  8,  8,
+  8,  9,  9,  9, 10, 10, 10, 11,
+ 11, 11, 12, 12, 12, 13, 13, 14,
+ 14, 15, 15, 16, 17, 18, 19, 20,
+ 21, 22, 23, 24, 25, 26, 27, 28,
+ 29,  0,  0,  0,  0,  0,  1,  1,
+  1,  1,  2,  2,  2,  3,  3,  3,
+  4,  4,  5,  5,  6,  6,  7,  7,
+  8,  8,  9,  9, 10, 10, 11, 11,
+ 12, 12, 13, 13, 14, 14, 15, 15,
+ 16, 17, 18, 19, 20, 21, 22, 23,
+ 24, 25, 26, 27, 28, 29, 30, 31,
+ 32, 33, 34, 35, 36, 37, 38, 39,
+ 40, 41, 42, 43,
+};
+
+/* third vlc table */
+
+const UINT16 table2_vlc[186][2] = {
+{ 0x1, 2 },{ 0x5, 3 },{ 0xd, 4 },{ 0x12, 5 },
+{ 0xe, 6 },{ 0x15, 7 },{ 0x13, 8 },{ 0x3f, 8 },
+{ 0x4b, 9 },{ 0x11f, 9 },{ 0xb8, 10 },{ 0x3e3, 10 },
+{ 0x172, 11 },{ 0x24d, 12 },{ 0x3da, 12 },{ 0x2dd, 13 },
+{ 0x1f55, 13 },{ 0x5b9, 14 },{ 0x3eae, 14 },{ 0x0, 4 },
+{ 0x10, 5 },{ 0x8, 7 },{ 0x20, 8 },{ 0x29, 9 },
+{ 0x1f4, 9 },{ 0x233, 10 },{ 0x1e0, 11 },{ 0x12a, 12 },
+{ 0x3dd, 12 },{ 0x50a, 13 },{ 0x1f29, 13 },{ 0xa42, 14 },
+{ 0x1272, 15 },{ 0x1737, 15 },{ 0x3, 5 },{ 0x11, 7 },
+{ 0xc4, 8 },{ 0x4b, 10 },{ 0xb4, 11 },{ 0x7d4, 11 },
+{ 0x345, 12 },{ 0x2d7, 13 },{ 0x7bf, 13 },{ 0x938, 14 },
+{ 0xbbb, 14 },{ 0x95e, 15 },{ 0x13, 5 },{ 0x78, 7 },
+{ 0x69, 9 },{ 0x232, 10 },{ 0x461, 11 },{ 0x3ec, 12 },
+{ 0x520, 13 },{ 0x1f2a, 13 },{ 0x3e50, 14 },{ 0x3e51, 14 },
+{ 0x1486, 15 },{ 0xc, 6 },{ 0x24, 9 },{ 0x94, 11 },
+{ 0x8c0, 12 },{ 0xf09, 14 },{ 0x1ef0, 15 },{ 0x3d, 6 },
+{ 0x53, 9 },{ 0x1a0, 11 },{ 0x2d6, 13 },{ 0xf08, 14 },
+{ 0x13, 7 },{ 0x7c, 9 },{ 0x7c1, 11 },{ 0x4ac, 14 },
+{ 0x1b, 7 },{ 0xa0, 10 },{ 0x344, 12 },{ 0xf79, 14 },
+{ 0x79, 7 },{ 0x3e1, 10 },{ 0x2d4, 13 },{ 0x2306, 14 },
+{ 0x21, 8 },{ 0x23c, 10 },{ 0xfae, 12 },{ 0x23de, 14 },
+{ 0x35, 8 },{ 0x175, 11 },{ 0x7b3, 13 },{ 0xc5, 8 },
+{ 0x174, 11 },{ 0x785, 13 },{ 0x48, 9 },{ 0x1a3, 11 },
+{ 0x49e, 13 },{ 0x2c, 9 },{ 0xfa, 10 },{ 0x7d6, 11 },
+{ 0x92, 10 },{ 0x5cc, 13 },{ 0x1ef1, 15 },{ 0xa3, 10 },
+{ 0x3ed, 12 },{ 0x93e, 14 },{ 0x1e2, 11 },{ 0x1273, 15 },
+{ 0x7c4, 11 },{ 0x1487, 15 },{ 0x291, 12 },{ 0x293, 12 },
+{ 0xf8a, 12 },{ 0x509, 13 },{ 0x508, 13 },{ 0x78d, 13 },
+{ 0x7be, 13 },{ 0x78c, 13 },{ 0x4ae, 14 },{ 0xbba, 14 },
+{ 0x2307, 14 },{ 0xb9a, 14 },{ 0x1736, 15 },{ 0xe, 4 },
+{ 0x45, 7 },{ 0x1f3, 9 },{ 0x47a, 11 },{ 0x5dc, 13 },
+{ 0x23df, 14 },{ 0x19, 5 },{ 0x28, 9 },{ 0x176, 11 },
+{ 0x49d, 13 },{ 0x23dd, 14 },{ 0x30, 6 },{ 0xa2, 10 },
+{ 0x2ef, 12 },{ 0x5b8, 14 },{ 0x3f, 6 },{ 0xa5, 10 },
+{ 0x3db, 12 },{ 0x93f, 14 },{ 0x44, 7 },{ 0x7cb, 11 },
+{ 0x95f, 15 },{ 0x63, 7 },{ 0x3c3, 12 },{ 0x15, 8 },
+{ 0x8f6, 12 },{ 0x17, 8 },{ 0x498, 13 },{ 0x2c, 8 },
+{ 0x7b2, 13 },{ 0x2f, 8 },{ 0x1f54, 13 },{ 0x8d, 8 },
+{ 0x7bd, 13 },{ 0x8e, 8 },{ 0x1182, 13 },{ 0xfb, 8 },
+{ 0x50b, 13 },{ 0x2d, 8 },{ 0x7c0, 11 },{ 0x79, 9 },
+{ 0x1f5f, 13 },{ 0x7a, 9 },{ 0x1f56, 13 },{ 0x231, 10 },
+{ 0x3e4, 10 },{ 0x1a1, 11 },{ 0x143, 11 },{ 0x1f7, 11 },
+{ 0x16f, 12 },{ 0x292, 12 },{ 0x2e7, 12 },{ 0x16c, 12 },
+{ 0x16d, 12 },{ 0x3dc, 12 },{ 0xf8b, 12 },{ 0x499, 13 },
+{ 0x3d8, 12 },{ 0x78e, 13 },{ 0x2d5, 13 },{ 0x1f5e, 13 },
+{ 0x1f2b, 13 },{ 0x78f, 13 },{ 0x4ad, 14 },{ 0x3eaf, 14 },
+{ 0x23dc, 14 },{ 0x4a, 9 },
+};
+
+const INT8 table2_level[185] = {
+  1,  2,  3,  4,  5,  6,  7,  8,
+  9, 10, 11, 12, 13, 14, 15, 16,
+ 17, 18, 19,  1,  2,  3,  4,  5,
+  6,  7,  8,  9, 10, 11, 12, 13,
+ 14, 15,  1,  2,  3,  4,  5,  6,
+  7,  8,  9, 10, 11, 12,  1,  2,
+  3,  4,  5,  6,  7,  8,  9, 10,
+ 11,  1,  2,  3,  4,  5,  6,  1,
+  2,  3,  4,  5,  1,  2,  3,  4,
+  1,  2,  3,  4,  1,  2,  3,  4,
+  1,  2,  3,  4,  1,  2,  3,  1,
+  2,  3,  1,  2,  3,  1,  2,  3,
+  1,  2,  3,  1,  2,  3,  1,  2,
+  1,  2,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  2,  3,  4,  5,  6,  1,  2,  3,
+  4,  5,  1,  2,  3,  4,  1,  2,
+  3,  4,  1,  2,  3,  1,  2,  1,
+  2,  1,  2,  1,  2,  1,  2,  1,
+  2,  1,  2,  1,  2,  1,  2,  1,
+  2,  1,  2,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,
+};
+
+const INT8 table2_run[185] = {
+  0,  0,  0,  0,  0,  0,  0,  0,
+  0,  0,  0,  0,  0,  0,  0,  0,
+  0,  0,  0,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  2,  2,  2,  2,  2,  2,
+  2,  2,  2,  2,  2,  2,  3,  3,
+  3,  3,  3,  3,  3,  3,  3,  3,
+  3,  4,  4,  4,  4,  4,  4,  5,
+  5,  5,  5,  5,  6,  6,  6,  6,
+  7,  7,  7,  7,  8,  8,  8,  8,
+  9,  9,  9,  9, 10, 10, 10, 11,
+ 11, 11, 12, 12, 12, 13, 13, 13,
+ 14, 14, 14, 15, 15, 15, 16, 16,
+ 17, 17, 18, 19, 20, 21, 22, 23,
+ 24, 25, 26, 27, 28, 29, 30,  0,
+  0,  0,  0,  0,  0,  1,  1,  1,
+  1,  1,  2,  2,  2,  2,  3,  3,
+  3,  3,  4,  4,  4,  5,  5,  6,
+  6,  7,  7,  8,  8,  9,  9, 10,
+ 10, 11, 11, 12, 12, 13, 13, 14,
+ 14, 15, 15, 16, 17, 18, 19, 20,
+ 21, 22, 23, 24, 25, 26, 27, 28,
+ 29, 30, 31, 32, 33, 34, 35, 36,
+ 37,
+};
+
+/* second non intra vlc table */
+const UINT16 table4_vlc[169][2] = {
+{ 0x0, 3 },{ 0x3, 4 },{ 0xb, 5 },{ 0x14, 6 },
+{ 0x3f, 6 },{ 0x5d, 7 },{ 0xa2, 8 },{ 0xac, 9 },
+{ 0x16e, 9 },{ 0x20a, 10 },{ 0x2e2, 10 },{ 0x432, 11 },
+{ 0x5c9, 11 },{ 0x827, 12 },{ 0xb54, 12 },{ 0x4e6, 13 },
+{ 0x105f, 13 },{ 0x172a, 13 },{ 0x20b2, 14 },{ 0x2d4e, 14 },
+{ 0x39f0, 14 },{ 0x4175, 15 },{ 0x5a9e, 15 },{ 0x4, 4 },
+{ 0x1e, 5 },{ 0x42, 7 },{ 0xb6, 8 },{ 0x173, 9 },
+{ 0x395, 10 },{ 0x72e, 11 },{ 0xb94, 12 },{ 0x16a4, 13 },
+{ 0x20b3, 14 },{ 0x2e45, 14 },{ 0x5, 5 },{ 0x40, 7 },
+{ 0x49, 9 },{ 0x28f, 10 },{ 0x5cb, 11 },{ 0x48a, 13 },
+{ 0x9dd, 14 },{ 0x73e2, 15 },{ 0x18, 5 },{ 0x25, 8 },
+{ 0x8a, 10 },{ 0x51b, 11 },{ 0xe5f, 12 },{ 0x9c9, 14 },
+{ 0x139c, 15 },{ 0x29, 6 },{ 0x4f, 9 },{ 0x412, 11 },
+{ 0x48d, 13 },{ 0x2e41, 14 },{ 0x38, 6 },{ 0x10e, 9 },
+{ 0x5a8, 11 },{ 0x105c, 13 },{ 0x39f2, 14 },{ 0x58, 7 },
+{ 0x21f, 10 },{ 0xe7e, 12 },{ 0x39ff, 14 },{ 0x23, 8 },
+{ 0x2e3, 10 },{ 0x4e5, 13 },{ 0x2e40, 14 },{ 0xa1, 8 },
+{ 0x5be, 11 },{ 0x9c8, 14 },{ 0x83, 8 },{ 0x13a, 11 },
+{ 0x1721, 13 },{ 0x44, 9 },{ 0x276, 12 },{ 0x39f6, 14 },
+{ 0x8b, 10 },{ 0x4ef, 13 },{ 0x5a9b, 15 },{ 0x208, 10 },
+{ 0x1cfe, 13 },{ 0x399, 10 },{ 0x1cb4, 13 },{ 0x39e, 10 },
+{ 0x39f3, 14 },{ 0x5ab, 11 },{ 0x73e3, 15 },{ 0x737, 11 },
+{ 0x5a9f, 15 },{ 0x82d, 12 },{ 0xe69, 12 },{ 0xe68, 12 },
+{ 0x433, 11 },{ 0xb7b, 12 },{ 0x2df8, 14 },{ 0x2e56, 14 },
+{ 0x2e57, 14 },{ 0x39f7, 14 },{ 0x51a5, 15 },{ 0x3, 3 },
+{ 0x2a, 6 },{ 0xe4, 8 },{ 0x28e, 10 },{ 0x735, 11 },
+{ 0x1058, 13 },{ 0x1cfa, 13 },{ 0x2df9, 14 },{ 0x4174, 15 },
+{ 0x9, 4 },{ 0x54, 8 },{ 0x398, 10 },{ 0x48b, 13 },
+{ 0x139d, 15 },{ 0xd, 4 },{ 0xad, 9 },{ 0x826, 12 },
+{ 0x2d4c, 14 },{ 0x11, 5 },{ 0x16b, 9 },{ 0xb7f, 12 },
+{ 0x51a4, 15 },{ 0x19, 5 },{ 0x21b, 10 },{ 0x16fd, 13 },
+{ 0x1d, 5 },{ 0x394, 10 },{ 0x28d3, 14 },{ 0x2b, 6 },
+{ 0x5bc, 11 },{ 0x5a9a, 15 },{ 0x2f, 6 },{ 0x247, 12 },
+{ 0x10, 7 },{ 0xa35, 12 },{ 0x3e, 6 },{ 0xb7a, 12 },
+{ 0x59, 7 },{ 0x105e, 13 },{ 0x26, 8 },{ 0x9cf, 14 },
+{ 0x55, 8 },{ 0x1cb5, 13 },{ 0x57, 8 },{ 0xe5b, 12 },
+{ 0xa0, 8 },{ 0x1468, 13 },{ 0x170, 9 },{ 0x90, 10 },
+{ 0x1ce, 9 },{ 0x21a, 10 },{ 0x218, 10 },{ 0x168, 9 },
+{ 0x21e, 10 },{ 0x244, 12 },{ 0x736, 11 },{ 0x138, 11 },
+{ 0x519, 11 },{ 0xe5e, 12 },{ 0x72c, 11 },{ 0xb55, 12 },
+{ 0x9dc, 14 },{ 0x20bb, 14 },{ 0x48c, 13 },{ 0x1723, 13 },
+{ 0x2e44, 14 },{ 0x16a5, 13 },{ 0x518, 11 },{ 0x39fe, 14 },
+{ 0x169, 9 },
+};
+
+const INT8 table4_level[168] = {
+  1,  2,  3,  4,  5,  6,  7,  8,
+  9, 10, 11, 12, 13, 14, 15, 16,
+ 17, 18, 19, 20, 21, 22, 23,  1,
+  2,  3,  4,  5,  6,  7,  8,  9,
+ 10, 11,  1,  2,  3,  4,  5,  6,
+  7,  8,  1,  2,  3,  4,  5,  6,
+  7,  1,  2,  3,  4,  5,  1,  2,
+  3,  4,  5,  1,  2,  3,  4,  1,
+  2,  3,  4,  1,  2,  3,  1,  2,
+  3,  1,  2,  3,  1,  2,  3,  1,
+  2,  1,  2,  1,  2,  1,  2,  1,
+  2,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  2,  3,  4,  5,
+  6,  7,  8,  9,  1,  2,  3,  4,
+  5,  1,  2,  3,  4,  1,  2,  3,
+  4,  1,  2,  3,  1,  2,  3,  1,
+  2,  3,  1,  2,  1,  2,  1,  2,
+  1,  2,  1,  2,  1,  2,  1,  2,
+  1,  2,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  1,  1,  1,  1,  1,  1,
+};
+
+const INT8 table4_run[168] = {
+  0,  0,  0,  0,  0,  0,  0,  0,
+  0,  0,  0,  0,  0,  0,  0,  0,
+  0,  0,  0,  0,  0,  0,  0,  1,
+  1,  1,  1,  1,  1,  1,  1,  1,
+  1,  1,  2,  2,  2,  2,  2,  2,
+  2,  2,  3,  3,  3,  3,  3,  3,
+  3,  4,  4,  4,  4,  4,  5,  5,
+  5,  5,  5,  6,  6,  6,  6,  7,
+  7,  7,  7,  8,  8,  8,  9,  9,
+  9, 10, 10, 10, 11, 11, 11, 12,
+ 12, 13, 13, 14, 14, 15, 15, 16,
+ 16, 17, 18, 19, 20, 21, 22, 23,
+ 24, 25, 26,  0,  0,  0,  0,  0,
+  0,  0,  0,  0,  1,  1,  1,  1,
+  1,  2,  2,  2,  2,  3,  3,  3,
+  3,  4,  4,  4,  5,  5,  5,  6,
+  6,  6,  7,  7,  8,  8,  9,  9,
+ 10, 10, 11, 11, 12, 12, 13, 13,
+ 14, 14, 15, 16, 17, 18, 19, 20,
+ 21, 22, 23, 24, 25, 26, 27, 28,
+ 29, 30, 31, 32, 33, 34, 35, 36,
+};
+
+extern const UINT16 inter_vlc[103][2];
+extern const INT8 inter_level[102];
+extern const INT8 inter_run[102];
+
+extern const UINT16 intra_vlc[103][2];
+extern const INT8 intra_level[102];
+extern const INT8 intra_run[102];
+
+#define NB_RL_TABLES  6
+
+static RLTable rl_table[NB_RL_TABLES] = {
+    /* intra luminance tables */
+    {
+        132,
+        85,
+        table0_vlc,
+        table0_run,
+        table0_level,
+    },
+    {
+        185,
+        119,
+        table2_vlc,
+        table2_run,
+        table2_level,
+    },
+    {
+        102,
+        67,
+        intra_vlc,
+        intra_run,
+        intra_level,
+    },
+    /* intra chrominance / non intra tables */
+    {
+        148,
+        81,
+        table1_vlc,
+        table1_run,
+        table1_level,
+    },
+    {
+        168,
+        99,
+        table4_vlc,
+        table4_run,
+        table4_level,
+    },
+    {
+        102,
+        58,
+        inter_vlc,
+        inter_run,
+        inter_level,
+    },
+};
+
+/* motion vector table 0 */
+
+const UINT16 table0_mv_code[1100] = {
+ 0x0001, 0x0003, 0x0005, 0x0007, 0x0003, 0x0008, 0x000c, 0x0001,
+ 0x0002, 0x001b, 0x0006, 0x000b, 0x0015, 0x0002, 0x000e, 0x000f,
+ 0x0014, 0x0020, 0x0022, 0x0025, 0x0027, 0x0029, 0x002d, 0x004b,
+ 0x004d, 0x0003, 0x0022, 0x0023, 0x0025, 0x0027, 0x0042, 0x0048,
+ 0x0049, 0x0050, 0x005c, 0x0091, 0x009f, 0x000e, 0x0043, 0x004c,
+ 0x0054, 0x0056, 0x008c, 0x0098, 0x009a, 0x009b, 0x00b1, 0x00b2,
+ 0x0120, 0x0121, 0x0126, 0x0133, 0x0139, 0x01a1, 0x01a4, 0x01a5,
+ 0x01a6, 0x01a7, 0x01ae, 0x01af, 0x000b, 0x0019, 0x0085, 0x0090,
+ 0x009b, 0x00aa, 0x00af, 0x010c, 0x010e, 0x011c, 0x011e, 0x0133,
+ 0x0144, 0x0160, 0x0174, 0x0175, 0x0177, 0x0178, 0x0249, 0x024b,
+ 0x0252, 0x0261, 0x0265, 0x0270, 0x0352, 0x0353, 0x0355, 0x0359,
+ 0x0010, 0x0011, 0x0013, 0x0034, 0x0035, 0x0036, 0x0037, 0x003d,
+ 0x003e, 0x0109, 0x0126, 0x0156, 0x021a, 0x021e, 0x023a, 0x023e,
+ 0x028e, 0x028f, 0x02cf, 0x0491, 0x0494, 0x049f, 0x04a0, 0x04a3,
+ 0x04a6, 0x04a7, 0x04ad, 0x04ae, 0x04c0, 0x04c4, 0x04c6, 0x04c8,
+ 0x04c9, 0x04f5, 0x04f6, 0x04f7, 0x0680, 0x0682, 0x0683, 0x0688,
+ 0x0689, 0x068d, 0x068e, 0x068f, 0x06a2, 0x06a3, 0x06a9, 0x06b0,
+ 0x06b1, 0x06b4, 0x06b5, 0x0024, 0x0060, 0x0063, 0x0078, 0x0079,
+ 0x0211, 0x0244, 0x0245, 0x0247, 0x0248, 0x0249, 0x024a, 0x024b,
+ 0x026b, 0x02af, 0x02b8, 0x02bb, 0x0436, 0x0476, 0x0477, 0x047e,
+ 0x04c8, 0x04c9, 0x04ca, 0x0514, 0x0586, 0x0587, 0x0598, 0x059d,
+ 0x05d9, 0x05da, 0x0920, 0x0921, 0x093b, 0x093c, 0x093d, 0x0942,
+ 0x0943, 0x0944, 0x0945, 0x0959, 0x095e, 0x095f, 0x0982, 0x0983,
+ 0x098e, 0x098f, 0x09c4, 0x09e7, 0x09e8, 0x09e9, 0x0d02, 0x0d17,
+ 0x0d18, 0x0d19, 0x0d41, 0x0d42, 0x0d43, 0x0d50, 0x0d5f, 0x0d6d,
+ 0x0d6e, 0x0d6f, 0x0052, 0x0053, 0x0054, 0x0055, 0x0056, 0x0057,
+ 0x00c2, 0x00c3, 0x00c4, 0x00c5, 0x041e, 0x041f, 0x0420, 0x0421,
+ 0x048c, 0x048d, 0x04d3, 0x04d4, 0x04d5, 0x055c, 0x055d, 0x0572,
+ 0x0573, 0x0574, 0x0575, 0x08de, 0x08df, 0x08fe, 0x08ff, 0x0996,
+ 0x0a36, 0x0a37, 0x0b08, 0x0b09, 0x0b0a, 0x0b0b, 0x0b32, 0x0b33,
+ 0x0b34, 0x0b35, 0x0b36, 0x0b37, 0x0b38, 0x0b39, 0x0bb0, 0x0bf7,
+ 0x0bf8, 0x0bf9, 0x0bfa, 0x0bfb, 0x0bfc, 0x0bfd, 0x0bfe, 0x0bff,
+ 0x1254, 0x1255, 0x1256, 0x1257, 0x1270, 0x1271, 0x1272, 0x1273,
+ 0x1274, 0x1275, 0x12ab, 0x12ac, 0x12ad, 0x12ae, 0x12af, 0x12b0,
+ 0x12b1, 0x1315, 0x1316, 0x1317, 0x13bf, 0x13c0, 0x13c1, 0x13c2,
+ 0x13c3, 0x13c4, 0x13c5, 0x13c6, 0x13c7, 0x13c8, 0x13c9, 0x13ca,
+ 0x13cb, 0x13cc, 0x13cd, 0x1a06, 0x1a07, 0x1a28, 0x1a29, 0x1a2a,
+ 0x1a2b, 0x1a2c, 0x1a2d, 0x1a80, 0x1abb, 0x1abc, 0x1abd, 0x1ad8,
+ 0x1ad9, 0x0094, 0x0095, 0x0096, 0x0097, 0x00a0, 0x00a1, 0x00a2,
+ 0x00a3, 0x0831, 0x0832, 0x0833, 0x0834, 0x0835, 0x0836, 0x0837,
+ 0x0838, 0x0839, 0x083a, 0x083b, 0x0939, 0x093a, 0x093b, 0x093c,
+ 0x093d, 0x093e, 0x093f, 0x09a0, 0x09a1, 0x09a2, 0x09a3, 0x09a4,
+ 0x09a5, 0x11ac, 0x11ad, 0x11ae, 0x11af, 0x11b0, 0x11b1, 0x11b2,
+ 0x11b3, 0x11b4, 0x11b5, 0x11b6, 0x11b7, 0x11b8, 0x11b9, 0x11ba,
+ 0x11bb, 0x132f, 0x1454, 0x1455, 0x1456, 0x1457, 0x1458, 0x1459,
+ 0x145a, 0x145b, 0x145c, 0x145d, 0x145e, 0x145f, 0x1460, 0x1461,
+ 0x1462, 0x1463, 0x1464, 0x1465, 0x1466, 0x1467, 0x1468, 0x1469,
+ 0x146a, 0x146b, 0x17de, 0x17df, 0x17e0, 0x17e1, 0x17e2, 0x17e3,
+ 0x17e4, 0x17e5, 0x17e6, 0x17e7, 0x17e8, 0x17e9, 0x17ea, 0x17eb,
+ 0x17ec, 0x17ed, 0x2540, 0x2541, 0x2542, 0x2543, 0x2544, 0x2545,
+ 0x2546, 0x2547, 0x2548, 0x2549, 0x254a, 0x254b, 0x254c, 0x254d,
+ 0x254e, 0x254f, 0x2550, 0x2551, 0x2552, 0x2553, 0x2554, 0x2555,
+ 0x2628, 0x2766, 0x2767, 0x2768, 0x2769, 0x276a, 0x276b, 0x276c,
+ 0x276d, 0x276e, 0x276f, 0x2770, 0x2771, 0x2772, 0x2773, 0x2774,
+ 0x2775, 0x2776, 0x2777, 0x2778, 0x2779, 0x277a, 0x277b, 0x277c,
+ 0x277d, 0x3503, 0x3544, 0x3545, 0x3546, 0x3547, 0x3560, 0x3561,
+ 0x3562, 0x3563, 0x3564, 0x3565, 0x3566, 0x3567, 0x3568, 0x3569,
+ 0x356a, 0x356b, 0x356c, 0x356d, 0x356e, 0x356f, 0x3570, 0x3571,
+ 0x3572, 0x3573, 0x3574, 0x3575, 0x03f0, 0x103d, 0x103e, 0x103f,
+ 0x1040, 0x1041, 0x1042, 0x1043, 0x1044, 0x1045, 0x1046, 0x1047,
+ 0x1048, 0x1049, 0x104a, 0x104b, 0x104c, 0x104d, 0x104e, 0x104f,
+ 0x1050, 0x1051, 0x1052, 0x1053, 0x1054, 0x1055, 0x1056, 0x1057,
+ 0x1058, 0x1059, 0x105a, 0x105b, 0x105c, 0x105d, 0x105e, 0x105f,
+ 0x1060, 0x1061, 0x1270, 0x1271, 0x21b8, 0x21b9, 0x21ba, 0x21bb,
+ 0x21bc, 0x21bd, 0x21be, 0x21bf, 0x21f0, 0x21f1, 0x21f2, 0x21f3,
+ 0x21f4, 0x21f5, 0x21f6, 0x21f7, 0x21f8, 0x21f9, 0x21fa, 0x21fb,
+ 0x21fc, 0x21fd, 0x21fe, 0x21ff, 0x2340, 0x2341, 0x2342, 0x2343,
+ 0x2344, 0x2345, 0x2346, 0x2347, 0x2348, 0x2349, 0x234a, 0x234b,
+ 0x234c, 0x234d, 0x234e, 0x234f, 0x2350, 0x2351, 0x2352, 0x2353,
+ 0x2354, 0x2355, 0x2356, 0x2357, 0x265c, 0x2f88, 0x2f89, 0x2f8a,
+ 0x2f8b, 0x2f8c, 0x2f8d, 0x2f8e, 0x2f8f, 0x2f90, 0x2f91, 0x2f92,
+ 0x2f93, 0x2f94, 0x2f95, 0x2f96, 0x2f97, 0x2f98, 0x2f99, 0x2f9a,
+ 0x2f9b, 0x2f9c, 0x2f9d, 0x2f9e, 0x2f9f, 0x2fa0, 0x2fa1, 0x2fa2,
+ 0x2fa3, 0x2fa4, 0x2fa5, 0x2fa6, 0x2fa7, 0x2fa8, 0x2fa9, 0x2faa,
+ 0x2fab, 0x2fac, 0x2fad, 0x2fae, 0x2faf, 0x2fb0, 0x2fb1, 0x2fb2,
+ 0x2fb3, 0x2fb4, 0x2fb5, 0x2fb6, 0x2fb7, 0x2fb8, 0x2fb9, 0x2fba,
+ 0x2fbb, 0x4c52, 0x4c53, 0x4e28, 0x4e29, 0x4e2a, 0x4e2b, 0x4e2c,
+ 0x4e2d, 0x4e2e, 0x4e2f, 0x4e30, 0x4e31, 0x4e32, 0x4e33, 0x4e34,
+ 0x4e35, 0x4e36, 0x4e37, 0x4e38, 0x4e39, 0x4e3a, 0x4e3b, 0x4e3c,
+ 0x4e3d, 0x4e3e, 0x4e3f, 0x4e80, 0x4e81, 0x4e82, 0x4e83, 0x4e84,
+ 0x4e85, 0x4e86, 0x4e87, 0x4e88, 0x4e89, 0x4e8a, 0x4e8b, 0x4e8c,
+ 0x4e8d, 0x4e8e, 0x4e8f, 0x4e90, 0x4e91, 0x4e92, 0x4e93, 0x4e94,
+ 0x4e95, 0x4e96, 0x4e97, 0x4e98, 0x4e99, 0x4e9a, 0x4e9b, 0x4e9c,
+ 0x4e9d, 0x4e9e, 0x4e9f, 0x4ea0, 0x4ea1, 0x4ea2, 0x4ea3, 0x4ea4,
+ 0x4ea5, 0x4ea6, 0x4ea7, 0x4ea8, 0x4ea9, 0x4eaa, 0x4eab, 0x4eac,
+ 0x4ead, 0x4eae, 0x4eaf, 0x4eb0, 0x4eb1, 0x4eb2, 0x4eb3, 0x4eb4,
+ 0x4eb5, 0x4eb6, 0x4eb7, 0x4eb8, 0x4eb9, 0x4eba, 0x4ebb, 0x4ebc,
+ 0x4ebd, 0x4ebe, 0x4ebf, 0x4ec0, 0x4ec1, 0x4ec2, 0x4ec3, 0x4ec4,
+ 0x4ec5, 0x4ec6, 0x4ec7, 0x4ec8, 0x4ec9, 0x4eca, 0x4ecb, 0x6a04,
+ 0x6a05, 0x07e2, 0x07e3, 0x07e4, 0x07e5, 0x07e6, 0x07e7, 0x07e8,
+ 0x07e9, 0x07ea, 0x07eb, 0x07ec, 0x07ed, 0x07ee, 0x07ef, 0x07f0,
+ 0x07f1, 0x07f2, 0x07f3, 0x07f4, 0x07f5, 0x07f6, 0x07f7, 0x07f8,
+ 0x07f9, 0x07fa, 0x07fb, 0x07fc, 0x07fd, 0x07fe, 0x07ff, 0x2000,
+ 0x2001, 0x2002, 0x2003, 0x2004, 0x2005, 0x2006, 0x2007, 0x2008,
+ 0x2009, 0x200a, 0x200b, 0x200c, 0x200d, 0x200e, 0x200f, 0x2010,
+ 0x2011, 0x2012, 0x2013, 0x2014, 0x2015, 0x2016, 0x2017, 0x2018,
+ 0x2019, 0x201a, 0x201b, 0x201c, 0x201d, 0x201e, 0x201f, 0x2020,
+ 0x2021, 0x2022, 0x2023, 0x2024, 0x2025, 0x2026, 0x2027, 0x2028,
+ 0x2029, 0x202a, 0x202b, 0x202c, 0x202d, 0x202e, 0x202f, 0x2030,
+ 0x2031, 0x2032, 0x2033, 0x2034, 0x2035, 0x2036, 0x2037, 0x2038,
+ 0x2039, 0x203a, 0x203b, 0x203c, 0x203d, 0x203e, 0x203f, 0x2040,
+ 0x2041, 0x2042, 0x2043, 0x2044, 0x2045, 0x2046, 0x2047, 0x2048,
+ 0x2049, 0x204a, 0x204b, 0x204c, 0x204d, 0x204e, 0x204f, 0x2050,
+ 0x2051, 0x2052, 0x2053, 0x2054, 0x2055, 0x2056, 0x2057, 0x2058,
+ 0x2059, 0x205a, 0x205b, 0x205c, 0x205d, 0x205e, 0x205f, 0x2060,
+ 0x2061, 0x2062, 0x2063, 0x2064, 0x2065, 0x2066, 0x2067, 0x2068,
+ 0x2069, 0x206a, 0x206b, 0x206c, 0x206d, 0x206e, 0x206f, 0x2070,
+ 0x2071, 0x2072, 0x2073, 0x2074, 0x2075, 0x2076, 0x2077, 0x2078,
+ 0x2079, 0x4cba, 0x4cbb, 0x5d88, 0x5d89, 0x5d8a, 0x5d8b, 0x5d8c,
+ 0x5d8d, 0x5d8e, 0x5d8f, 0x5db0, 0x5db1, 0x5db2, 0x5db3, 0x5db4,
+ 0x5db5, 0x5db6, 0x5db7, 0x5db8, 0x5db9, 0x5dba, 0x5dbb, 0x5dbc,
+ 0x5dbd, 0x5dbe, 0x5dbf, 0x5e40, 0x5e41, 0x5e42, 0x5e43, 0x5e44,
+ 0x5e45, 0x5e46, 0x5e47, 0x5e48, 0x5e49, 0x5e4a, 0x5e4b, 0x5e4c,
+ 0x5e4d, 0x5e4e, 0x5e4f, 0x5e50, 0x5e51, 0x5e52, 0x5e53, 0x5e54,
+ 0x5e55, 0x5e56, 0x5e57, 0x5e58, 0x5e59, 0x5e5a, 0x5e5b, 0x5e5c,
+ 0x5e5d, 0x5e5e, 0x5e5f, 0x5e60, 0x5e61, 0x5e62, 0x5e63, 0x5e64,
+ 0x5e65, 0x5e66, 0x5e67, 0x5e68, 0x5e69, 0x5e6a, 0x5e6b, 0x5e6c,
+ 0x5e6d, 0x5e6e, 0x5e6f, 0x5e70, 0x5e71, 0x5e72, 0x5e73, 0x5e74,
+ 0x5e75, 0x5e76, 0x5e77, 0x5e78, 0x5e79, 0x5e7a, 0x5e7b, 0x5e7c,
+ 0x5e7d, 0x5e7e, 0x5e7f, 0x5e80, 0x5e81, 0x5e82, 0x5e83, 0x5e84,
+ 0x5e85, 0x5e86, 0x5e87, 0x5e88, 0x5e89, 0x5e8a, 0x5e8b, 0x5e8c,
+ 0x5e8d, 0x5e8e, 0x5e8f, 0x5e90, 0x5e91, 0x5e92, 0x5e93, 0x5e94,
+ 0x5e95, 0x5e96, 0x5e97, 0x5e98, 0x5e99, 0x5e9a, 0x5e9b, 0x5e9c,
+ 0x5e9d, 0x5e9e, 0x5e9f, 0x5ea0, 0x5ea1, 0x5ea2, 0x5ea3, 0x5ea4,
+ 0x5ea5, 0x5ea6, 0x5ea7, 0x5ea8, 0x5ea9, 0x5eaa, 0x5eab, 0x5eac,
+ 0x5ead, 0x5eae, 0x5eaf, 0x5eb0, 0x5eb1, 0x5eb2, 0x5eb3, 0x5eb4,
+ 0x5eb5, 0x5eb6, 0x5eb7, 0x5eb8, 0x5eb9, 0x5eba, 0x5ebb, 0x5ebc,
+ 0x5ebd, 0x5ebe, 0x5ebf, 0x5ec0, 0x5ec1, 0x5ec2, 0x5ec3, 0x5ec4,
+ 0x5ec5, 0x5ec6, 0x5ec7, 0x5ec8, 0x5ec9, 0x5eca, 0x5ecb, 0x5ecc,
+ 0x5ecd, 0x5ece, 0x5ecf, 0x5ed0, 0x5ed1, 0x5ed2, 0x5ed3, 0x5ed4,
+ 0x5ed5, 0x5ed6, 0x5ed7, 0x5ed8, 0x5ed9, 0x5eda, 0x5edb, 0x5edc,
+ 0x5edd, 0x5ede, 0x5edf, 0x5ee0, 0x5ee1, 0x5ee2, 0x5ee3, 0x5ee4,
+ 0x5ee5, 0x5ee6, 0x5ee7, 0x5ee8, 0x5ee9, 0x5eea, 0x5eeb, 0x5eec,
+ 0x5eed, 0x5eee, 0x5eef, 0x5ef0, 0x5ef1, 0x5ef2, 0x5ef3, 0x5ef4,
+ 0x5ef5, 0x5ef6, 0x5ef7, 0x5ef8, 0x5ef9, 0x5efa, 0x5efb, 0x5efc,
+ 0x5efd, 0x5efe, 0x5eff, 0x5f00, 0x5f01, 0x5f02, 0x5f03, 0x5f04,
+ 0x5f05, 0x5f06, 0x5f07, 0x5f08, 0x5f09, 0x5f0a, 0x5f0b, 0x5f0c,
+ 0x5f0d, 0x5f0e, 0x5f0f, 0x0000,
+};
+
+const UINT8 table0_mv_bits[1100] = {
+  1,  4,  4,  4,  5,  5,  5,  6,
+  6,  6,  7,  7,  7,  8,  8,  8,
+  8,  8,  8,  8,  8,  8,  8,  8,
+  8,  9,  9,  9,  9,  9,  9,  9,
+  9,  9,  9,  9,  9, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 16, 16, 16, 16, 16, 16, 16,
+ 16, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17, 17, 17, 17, 17, 17,
+ 17, 17, 17,  8,
+};
+
+const UINT8 table0_mvx[1099] = {
+ 32, 32, 31, 32, 33, 31, 33, 31,
+ 33, 32, 34, 32, 30, 32, 31, 34,
+ 35, 32, 34, 33, 29, 33, 30, 30,
+ 31, 31, 35, 29, 33, 35, 33, 34,
+ 31, 29, 30, 34, 30, 36, 28, 32,
+ 34, 37, 30, 27, 32, 25, 39, 32,
+ 34, 32, 35, 35, 35, 31, 35, 29,
+ 32, 29, 30, 29, 37, 27, 36, 38,
+ 37, 33, 32, 31, 29, 31, 28, 36,
+ 33, 30, 34, 33, 33, 28, 27, 25,
+ 31, 26, 39, 32, 32, 31, 33, 39,
+ 31, 38, 28, 36, 21, 23, 43, 36,
+ 34, 41, 30, 25, 28, 31, 30, 34,
+ 38, 35, 61, 34, 28, 30, 37, 37,
+ 35, 27, 36,  3, 59, 38, 37, 32,
+ 31, 29, 26, 33, 37, 33, 27, 27,
+ 35, 34, 34, 40, 42, 33, 32, 29,
+  4,  5, 28, 24, 25, 35, 39, 38,
+ 32, 23, 27, 32, 30, 35, 26, 34,
+ 60, 36, 29, 22, 26, 41,  7, 30,
+ 38, 30, 36, 29, 30, 41, 26, 25,
+ 32, 34, 24, 39,  1, 25, 39, 32,
+ 28, 29, 32, 38, 26, 36, 28, 63,
+ 28, 39, 23, 21, 26, 35, 31, 35,
+ 57, 31, 29, 29, 28, 30, 27, 35,
+  2, 38, 40, 34, 37, 29, 38, 43,
+ 26, 32, 33, 42, 24, 40, 28, 32,
+ 32, 32, 36, 32, 43, 25, 21, 31,
+ 30, 31, 41, 29, 33, 37, 26, 37,
+ 27, 59, 23, 33, 35, 31, 31, 37,
+ 38, 39, 32, 23, 32, 27, 37, 36,
+ 31, 40, 25, 27, 38, 31, 36, 28,
+ 31, 36, 25, 45,  3, 34, 38, 39,
+ 40, 38, 30, 32, 19, 24, 25, 26,
+ 45, 20, 24, 33, 33, 31, 41, 34,
+ 39, 47, 40, 58, 59, 41, 33,  3,
+ 17, 61, 42, 30, 26, 29, 36, 61,
+ 33, 37, 62, 28, 25, 38, 25, 38,
+ 17, 23, 34, 33, 21, 33, 49, 27,
+ 32, 23, 27, 22, 24, 22, 39, 43,
+ 27, 37,  6, 42, 47, 26, 30, 31,
+ 41, 39, 33, 22, 45, 36, 32, 45,
+ 19, 22, 30,  5,  5, 17, 29, 22,
+ 31, 31, 43, 37, 27, 32, 32, 32,
+ 33, 34, 43, 35, 29, 26, 22, 32,
+ 19, 32, 25, 31, 41, 49, 28, 34,
+ 28, 39, 34, 19, 37, 38, 29, 21,
+ 36, 42, 24, 48, 16, 28, 49, 22,
+ 34, 31, 38, 39, 44, 11, 35, 30,
+ 33, 33, 23, 28, 33, 46, 15, 13,
+ 24, 41, 24, 34, 34, 30, 26, 24,
+ 14, 60, 21, 29, 39, 23, 35, 37,
+ 63, 45, 33, 34, 47, 41, 22, 42,
+ 35, 35, 23, 32, 35, 43, 32,  7,
+ 31, 41, 20, 31, 16, 13, 63, 25,
+ 30, 32, 35, 30, 30, 31, 42, 47,
+ 39, 38, 40, 40, 51, 55, 56, 18,
+ 21, 39, 39, 33, 17, 41, 23, 24,
+ 43, 25, 31, 20, 19, 45,  1, 34,
+ 31, 22, 35, 15, 46, 46, 35, 31,
+ 28, 29, 29, 23, 41, 27, 14, 53,
+ 53, 27, 24, 32, 57, 32, 17, 42,
+ 37, 29, 33,  1, 25, 32, 32, 63,
+ 26, 40, 44, 36, 31, 39, 20, 20,
+ 44, 23, 33, 34, 35, 33, 33, 28,
+ 41, 23, 41, 41, 29, 25, 26, 49,
+ 29, 24, 37, 49, 50, 51, 51, 26,
+ 39, 25, 26, 15, 39, 18, 42, 17,
+  4, 31, 32, 32, 60,  1, 42, 32,
+  0, 12, 19, 35, 21, 41, 17, 26,
+ 20, 45, 46, 32, 37, 22, 47, 29,
+ 31, 27, 29, 30, 21, 33, 35, 18,
+ 25, 33, 50, 51, 42,  2, 15, 51,
+ 53, 33, 25, 29, 55, 37, 38, 33,
+ 38, 59, 38, 33, 39, 13, 32, 40,
+ 61, 61, 32,  9, 44,  3, 31, 29,
+ 25, 31, 27, 23,  9, 25,  9, 29,
+ 20, 30, 30, 42, 18, 28, 25, 28,
+ 28, 21, 29, 43, 29, 43, 26, 44,
+ 44, 21, 38, 21, 24, 45, 45, 35,
+ 39, 22, 35, 36, 34, 34, 45, 34,
+ 29, 31, 46, 25, 46, 16, 17, 31,
+ 20, 32, 47, 47, 47, 32, 49, 49,
+ 49, 31,  1, 27, 28, 39, 39, 21,
+ 36, 23, 51,  2, 40, 51, 32, 53,
+ 24, 30, 24, 30, 21, 40, 57, 57,
+ 31, 41, 58, 32, 12,  4, 32, 34,
+ 59, 31, 32, 13,  9, 35, 26, 35,
+ 37, 61, 37, 63, 26, 29, 41, 38,
+ 23, 20, 41, 26, 41, 42, 42, 42,
+ 26, 26, 26, 26,  1, 26, 37, 37,
+ 37, 23, 34, 42, 27, 43, 34, 27,
+ 31, 24, 33, 16,  3, 31, 24, 33,
+ 24,  4, 44, 44, 11, 44, 31, 13,
+ 13, 44, 45, 13, 25, 22, 38, 26,
+ 38, 38, 39, 32, 30, 39, 30, 22,
+ 32, 26, 30, 47, 47, 47, 19, 47,
+ 30, 31, 35,  8, 23, 47, 47, 27,
+ 35, 47, 31, 48, 35, 19, 36, 49,
+ 49, 33, 31, 39, 27, 39, 49, 49,
+ 50, 50, 50, 39, 31, 51, 51, 39,
+ 28, 33, 33, 21, 40, 31, 52, 53,
+ 40, 53,  9, 33, 31, 53, 54, 54,
+ 54, 55, 55, 34, 15, 56, 25, 56,
+ 21, 21, 40, 40, 25, 40, 58, 36,
+  5, 41, 41, 12, 60, 41, 41, 37,
+ 22, 61, 18, 29, 29, 30, 61, 30,
+ 61, 62, 62, 30, 30, 63, 18, 13,
+ 30, 23, 19, 20, 20, 41, 13,  2,
+  5,  5,  1,  5, 32,  6, 32, 35,
+ 20, 35, 27, 35, 35, 36, 36, 13,
+ 36, 41, 41, 41,  3, 30, 42, 27,
+ 20, 30, 27, 28, 30, 21, 33, 33,
+ 14, 24, 30, 42, 24, 33, 25, 42,
+ 43, 14, 43, 43, 14, 43,  7, 36,
+ 37, 37, 37, 37,  7, 14, 25, 43,
+ 43, 44, 15, 37,  7,  7,  3,  1,
+  8, 15, 15,  8, 44, 44, 44, 45,
+ 45, 45, 45,  8,  8, 45, 21, 45,
+ 28, 28, 28, 21, 28, 28, 22, 37,
+ 46, 46, 37,  8, 29, 37, 29, 22,
+ 46, 37, 22, 29, 47, 47, 38, 38,
+ 16, 38, 38, 33, 38, 22, 47, 47,
+ 29, 25, 16,  0, 48,  1, 34, 48,
+ 48, 34, 25, 26, 26, 49, 49, 26,
+  1, 49,  4, 26,  4, 49,  1,  9,
+ 49, 49, 49, 10, 49, 17, 38, 17,
+ 17, 50, 38, 50, 50, 22, 38, 51,
+ 38, 38, 51, 39, 39, 18, 22, 39,
+ 51, 22, 52, 52, 52, 39, 53, 53,
+ 10, 23, 18, 29, 10, 53, 29, 54,
+ 11, 54, 11, 11, 55,  1, 18, 55,
+ 55, 55, 55, 55, 55, 29, 34, 18,
+ 29, 56, 56, 34, 57, 34, 34, 29,
+ 29, 57, 57, 35, 35, 35, 35, 35,
+ 39, 35, 59, 59, 18, 59, 39, 30,
+ 18, 40, 60, 60, 61, 30, 18, 61,
+ 61, 19, 19,
+};
+
+const UINT8 table0_mvy[1099] = {
+ 32, 31, 32, 33, 32, 31, 31, 33,
+ 33, 34, 32, 30, 32, 35, 34, 31,
+ 32, 29, 33, 30, 32, 34, 33, 31,
+ 30, 35, 31, 31, 29, 33, 35, 30,
+ 29, 33, 34, 34, 30, 32, 32, 36,
+ 29, 32, 35, 32, 28, 32, 32, 27,
+ 35, 37, 34, 29, 30, 36, 35, 34,
+ 25, 30, 29, 35, 33, 31, 31, 32,
+ 31, 28, 39, 28, 29, 37, 31, 33,
+ 27, 36, 28, 36, 37, 33, 33, 31,
+ 27, 32, 31, 38, 26, 25, 25, 33,
+ 39, 31, 34, 30, 32, 32, 32, 34,
+ 36, 32, 28, 33, 30, 38, 37, 27,
+ 33, 28, 32, 37, 35, 38, 29, 34,
+ 27, 29, 29, 32, 32, 34, 35,  3,
+ 26, 36, 31, 38, 30, 26, 35, 34,
+ 37, 26, 25, 32, 32, 39, 23, 37,
+ 32, 32, 29, 32, 29, 36, 29, 30,
+ 41, 31, 30, 21, 39, 25, 34, 38,
+ 32, 35, 39, 32, 33, 33, 32, 27,
+ 29, 25, 28, 27, 26, 31, 30, 35,
+ 24, 24, 31, 34, 32, 30, 35, 40,
+ 28, 38,  5, 35, 29, 36, 36, 32,
+ 38, 30, 33, 31, 35, 26, 23, 38,
+ 32, 41, 28, 25, 37, 40, 37, 39,
+ 32, 36, 33, 39, 25, 26, 28, 31,
+ 28, 42, 23, 31, 33, 31, 39,  1,
+ 59, 22, 27,  4, 33, 34, 33, 24,
+ 41,  3, 35, 41, 41, 28, 36, 36,
+ 28, 33, 35, 21, 23, 21, 22, 37,
+ 27, 27, 43, 29, 60, 39, 27, 25,
+ 59, 34, 27, 27, 26, 40, 37, 27,
+ 61, 26, 39, 33, 31, 22, 37, 25,
+ 30, 25, 24, 61, 31, 34, 25, 38,
+ 32, 32, 30,  3, 61, 43, 29, 23,
+ 28, 32, 28, 32, 31, 34,  5, 33,
+ 32, 33, 33, 42, 37, 23, 38, 31,
+ 40, 26, 32, 26, 37, 38, 36, 24,
+ 29, 30, 20, 22, 29, 24, 32, 41,
+  2, 34, 25, 33, 29, 31, 39, 35,
+ 36, 24, 32, 30, 33, 27, 44, 60,
+ 30, 36, 19, 34, 31, 24, 16, 35,
+ 32, 38, 21, 33, 31, 31, 21, 35,
+  5, 17, 29, 38, 38, 18, 58, 19,
+ 43, 41, 30, 41, 43, 39, 29,  7,
+ 29, 17, 28, 19, 28, 31, 25, 19,
+ 40, 26, 21, 33, 39, 23, 40, 30,
+ 39, 34, 35, 32, 32, 24, 33, 30,
+ 40, 47, 39, 37, 32, 33, 24, 23,
+ 45, 47, 27, 23, 42, 32, 32, 33,
+ 36, 37, 37, 17, 18, 22, 40, 38,
+ 32, 31, 35, 24, 17, 25, 17, 23,
+ 33, 34, 51, 42, 31, 36, 36, 29,
+ 21, 22, 37, 44, 43, 25, 47, 33,
+ 45, 27, 31, 58, 31, 32, 31, 38,
+ 43, 20, 47, 45, 54,  1, 26, 34,
+ 38, 14, 22, 24, 33, 34, 32, 32,
+ 37, 21, 23, 49, 35, 23, 28, 39,
+ 39, 23, 55, 33, 30, 30, 63, 16,
+ 42, 28, 13, 33, 33, 35, 19, 46,
+ 43, 17, 19, 36, 39, 24, 31, 32,
+ 33, 26, 28, 62, 33, 63, 33, 39,
+ 19, 49, 17, 31, 43, 13, 15, 29,
+ 25, 35, 33, 23, 49, 41, 28, 29,
+ 34, 38,  7, 61, 11, 50, 13, 41,
+ 19, 47, 25, 26, 15, 42, 41, 29,
+ 45, 27, 17, 35, 32, 29, 32, 24,
+ 13, 26, 26, 31, 24, 33, 28, 30,
+ 31, 11, 45, 46, 33, 33, 35, 57,
+ 32, 32, 35, 45, 34, 11, 37, 42,
+ 39, 37, 31, 49, 21, 27, 29, 47,
+ 53, 40, 51, 16, 26,  1, 40, 30,
+ 41, 44, 34, 25, 27, 31, 35, 35,
+ 31, 15, 49,  1, 35, 40,  5, 58,
+ 21, 29, 22, 59, 45, 31,  9, 26,
+  9, 29, 11, 32, 30,  3, 13, 20,
+ 18, 20, 11,  3, 29, 40, 31, 53,
+ 30, 17, 20, 37, 31, 42, 47, 47,
+ 54, 38,  9, 34, 13, 37, 21, 25,
+ 27, 43, 42, 45, 40, 25, 27, 46,
+ 22, 25, 53, 20,  2, 14, 39, 15,
+ 22, 44, 34, 21, 38, 33, 27, 48,
+ 34, 52, 35, 47, 49, 54,  2, 13,
+ 23, 52, 29, 45, 22, 49, 54, 21,
+ 40, 42, 31, 30, 29, 34,  0, 25,
+ 23, 51, 24, 59, 28, 38, 29, 31,
+  2, 13, 31,  8, 31, 33, 12, 45,
+ 41,  7, 14, 30, 25, 18, 43, 20,
+ 43, 35, 44,  1, 49, 42, 42, 18,
+ 41, 38, 41, 44, 53, 11, 20, 25,
+ 45, 46, 47, 48, 39, 52, 46, 49,
+ 63, 55, 44, 38, 13, 13, 57, 22,
+ 51, 16, 12, 28, 35, 57, 25, 20,
+ 26, 28, 28, 29, 32, 31, 62, 34,
+ 35, 35, 19, 49, 48, 39, 40, 18,
+ 43, 46, 11,  6, 48, 19, 49, 41,
+ 10, 23, 58, 17, 21, 23, 34, 30,
+ 60,  0, 44, 34, 26, 37, 46, 43,
+ 49, 59,  4, 34, 59, 37, 22, 25,
+ 28, 46,  6, 40, 59, 42, 36, 61,
+ 28, 30, 31, 43, 10, 22, 23, 47,
+ 20, 52, 55, 36, 25, 16,  1, 11,
+ 27, 29,  5, 63, 18, 41, 31, 34,
+ 38,  1,  5, 13, 28, 31, 17, 38,
+ 39, 41, 36, 37, 22, 39, 33, 43,
+ 43, 15, 17, 49, 30, 21, 22, 20,
+ 10, 17, 25, 54, 57,  3, 34,  8,
+ 36, 25, 31, 14, 15, 19, 29, 25,
+ 18, 39, 53, 22, 27, 20, 29, 33,
+ 41, 42, 35, 62, 50, 29, 53, 50,
+ 35, 55, 42, 61, 63,  4,  7, 42,
+ 21, 46, 47, 49, 27, 46, 17, 55,
+ 41, 50, 63,  4, 56, 18,  8, 10,
+ 18, 51, 63, 36, 55, 18,  5, 55,
+  9, 29, 17, 21, 30, 27,  1, 59,
+  7, 11, 12, 15,  5, 42, 24, 41,
+ 43,  7, 27, 22, 25, 31, 30, 37,
+ 22, 39, 53, 29, 36, 37, 48,  0,
+  5, 13, 17, 31, 32, 26, 46, 28,
+ 44, 45, 46, 53, 49, 51,  3, 41,
+  3, 22, 42, 33,  5, 45,  7, 22,
+ 40, 53, 24, 14, 25, 27, 10, 12,
+ 34, 16, 17, 53, 20, 26, 39, 45,
+ 18, 45, 35, 33, 31, 49,  4, 39,
+ 42, 11, 51,  5, 13, 26, 27, 17,
+ 52, 30,  0, 22, 12, 34, 62, 36,
+ 38, 41, 47, 30, 63, 38, 41, 43,
+ 59, 33, 45, 37, 38, 40, 47, 24,
+ 48, 49, 30,  1, 10, 22, 49, 15,
+ 39, 59, 31, 32, 33, 18, 13, 15,
+ 31, 21, 27, 44, 42, 39, 46, 17,
+ 26, 32, 30, 31,  0, 30, 34,  9,
+ 12, 13, 25, 31, 32, 55, 43, 35,
+ 61, 33, 35, 46, 25, 47, 48, 62,
+ 63, 38, 61,  1,  2,  5,  7,  9,
+ 46, 10, 34, 35, 36, 55, 51,  7,
+ 40, 23, 34, 37,  5, 13, 42, 18,
+ 25, 27, 28,
+};
+
+/* motion vector table 1 */
+const UINT16 table1_mv_code[1100] = {
+ 0x0000, 0x0007, 0x0009, 0x000f, 0x000a, 0x0011, 0x001a, 0x001c,
+ 0x0011, 0x0031, 0x0025, 0x002d, 0x002f, 0x006f, 0x0075, 0x0041,
+ 0x004c, 0x004e, 0x005c, 0x0060, 0x0062, 0x0066, 0x0068, 0x0069,
+ 0x006b, 0x00a6, 0x00c1, 0x00cb, 0x00cc, 0x00ce, 0x00da, 0x00e8,
+ 0x00ee, 0x0087, 0x0090, 0x009e, 0x009f, 0x00ba, 0x00ca, 0x00d8,
+ 0x00db, 0x00df, 0x0104, 0x0109, 0x010c, 0x0143, 0x0145, 0x014a,
+ 0x0156, 0x015c, 0x01b3, 0x01d3, 0x01da, 0x0103, 0x0109, 0x010b,
+ 0x0122, 0x0127, 0x0134, 0x0161, 0x0164, 0x0176, 0x0184, 0x018d,
+ 0x018e, 0x018f, 0x0190, 0x0193, 0x0196, 0x019d, 0x019e, 0x019f,
+ 0x01a9, 0x01b2, 0x01b4, 0x01ba, 0x01bb, 0x01bc, 0x0201, 0x0202,
+ 0x0205, 0x0207, 0x020d, 0x0210, 0x0211, 0x0215, 0x021b, 0x021f,
+ 0x0281, 0x0285, 0x0290, 0x029c, 0x029d, 0x02a2, 0x02a7, 0x02a8,
+ 0x02aa, 0x02b0, 0x02b1, 0x02b4, 0x02bc, 0x02bf, 0x0320, 0x0326,
+ 0x0327, 0x0329, 0x032a, 0x0336, 0x0360, 0x0362, 0x0363, 0x0372,
+ 0x03b2, 0x03bc, 0x03bd, 0x0203, 0x0205, 0x021a, 0x0249, 0x024a,
+ 0x024c, 0x02c7, 0x02ca, 0x02ce, 0x02ef, 0x030d, 0x0322, 0x0325,
+ 0x0338, 0x0373, 0x037a, 0x0409, 0x0415, 0x0416, 0x0418, 0x0428,
+ 0x042d, 0x042f, 0x0434, 0x0508, 0x0509, 0x0510, 0x0511, 0x051c,
+ 0x051e, 0x0524, 0x0541, 0x0543, 0x0546, 0x0547, 0x054d, 0x0557,
+ 0x055f, 0x056a, 0x056c, 0x056d, 0x056f, 0x0576, 0x0577, 0x057a,
+ 0x057b, 0x057c, 0x057d, 0x0600, 0x0601, 0x0603, 0x0614, 0x0616,
+ 0x0617, 0x061c, 0x061f, 0x0642, 0x0648, 0x0649, 0x064a, 0x064b,
+ 0x0657, 0x0668, 0x0669, 0x066b, 0x066e, 0x067f, 0x06c2, 0x06c8,
+ 0x06cb, 0x06de, 0x06df, 0x06e2, 0x06e3, 0x06ef, 0x0748, 0x074b,
+ 0x076e, 0x076f, 0x077c, 0x0409, 0x0423, 0x0428, 0x0429, 0x042a,
+ 0x042b, 0x0432, 0x0433, 0x0496, 0x049a, 0x04d5, 0x04db, 0x0581,
+ 0x0582, 0x058b, 0x058c, 0x058d, 0x0598, 0x0599, 0x059a, 0x059e,
+ 0x05dd, 0x0619, 0x0632, 0x0633, 0x0648, 0x0672, 0x06a1, 0x06a2,
+ 0x06a3, 0x06af, 0x06e2, 0x06e3, 0x06e4, 0x0800, 0x0801, 0x0802,
+ 0x0803, 0x081a, 0x081b, 0x0829, 0x082f, 0x0832, 0x083e, 0x083f,
+ 0x0852, 0x0853, 0x0858, 0x086b, 0x0877, 0x0878, 0x0879, 0x087a,
+ 0x087b, 0x0a00, 0x0a01, 0x0a0d, 0x0a0e, 0x0a0f, 0x0a24, 0x0a37,
+ 0x0a3a, 0x0a3b, 0x0a3e, 0x0a46, 0x0a47, 0x0a4a, 0x0a4b, 0x0a5f,
+ 0x0a79, 0x0a7a, 0x0a7b, 0x0a80, 0x0a81, 0x0a84, 0x0a85, 0x0a99,
+ 0x0aa5, 0x0aa6, 0x0ab8, 0x0aba, 0x0abb, 0x0abc, 0x0abd, 0x0ac8,
+ 0x0ace, 0x0acf, 0x0ad7, 0x0adc, 0x0aeb, 0x0c04, 0x0c25, 0x0c26,
+ 0x0c27, 0x0c2a, 0x0c2b, 0x0c3a, 0x0c3b, 0x0c3c, 0x0c3d, 0x0ca0,
+ 0x0cad, 0x0cd4, 0x0cd5, 0x0cfc, 0x0cfd, 0x0d86, 0x0d92, 0x0d93,
+ 0x0d94, 0x0d95, 0x0db0, 0x0db8, 0x0db9, 0x0dba, 0x0dbb, 0x0dc0,
+ 0x0dc2, 0x0dc3, 0x0dda, 0x0ddb, 0x0ddc, 0x0ddd, 0x0e92, 0x0e93,
+ 0x0e94, 0x0e95, 0x0ec7, 0x0ecc, 0x0ece, 0x0ecf, 0x0ed8, 0x0ed9,
+ 0x0eda, 0x0edb, 0x0808, 0x0809, 0x080a, 0x0810, 0x0811, 0x0844,
+ 0x0845, 0x0861, 0x0862, 0x0863, 0x086c, 0x0922, 0x0923, 0x092e,
+ 0x092f, 0x0936, 0x0937, 0x09b1, 0x09b2, 0x09b3, 0x09b4, 0x09b5,
+ 0x09b8, 0x09b9, 0x09ba, 0x09bb, 0x09bc, 0x09bd, 0x09be, 0x09bf,
+ 0x0b00, 0x0b15, 0x0b2c, 0x0b2d, 0x0b2e, 0x0b2f, 0x0b36, 0x0bb9,
+ 0x0c28, 0x0c2a, 0x0c2b, 0x0c2c, 0x0c2d, 0x0c2e, 0x0c2f, 0x0c30,
+ 0x0c31, 0x0c38, 0x0c60, 0x0c61, 0x0c62, 0x0c63, 0x0c8d, 0x0c8e,
+ 0x0c8f, 0x0c92, 0x0cbe, 0x0cbf, 0x0ce6, 0x0ce7, 0x0d40, 0x0d41,
+ 0x0d57, 0x0d58, 0x0d59, 0x0d5a, 0x0d5b, 0x0d5c, 0x0d5d, 0x0d98,
+ 0x0d99, 0x0d9a, 0x0d9b, 0x0d9c, 0x0d9d, 0x0dad, 0x0dae, 0x0daf,
+ 0x0dc0, 0x0dc1, 0x0dc2, 0x0dc3, 0x0dca, 0x0dcb, 0x0dec, 0x0ded,
+ 0x0dee, 0x0def, 0x1018, 0x1022, 0x1023, 0x1030, 0x1031, 0x1032,
+ 0x1033, 0x1050, 0x1051, 0x105c, 0x1074, 0x1075, 0x1076, 0x1077,
+ 0x1078, 0x1079, 0x107a, 0x107b, 0x10b2, 0x10b3, 0x10b8, 0x10b9,
+ 0x10ba, 0x10bb, 0x10d4, 0x10ea, 0x10eb, 0x10ec, 0x10ed, 0x1404,
+ 0x1405, 0x1406, 0x1407, 0x1410, 0x1411, 0x1412, 0x1413, 0x1414,
+ 0x1415, 0x1416, 0x1417, 0x1418, 0x1419, 0x1466, 0x1467, 0x1468,
+ 0x1469, 0x146a, 0x146b, 0x146c, 0x146d, 0x147e, 0x147f, 0x1488,
+ 0x1489, 0x148a, 0x148b, 0x14b6, 0x14b7, 0x14b8, 0x14b9, 0x14ba,
+ 0x14bb, 0x14bc, 0x14bd, 0x14f0, 0x14f1, 0x14f8, 0x14f9, 0x14fa,
+ 0x14fb, 0x14fc, 0x14fd, 0x14fe, 0x14ff, 0x152a, 0x152b, 0x152c,
+ 0x152d, 0x152e, 0x152f, 0x1530, 0x1531, 0x1548, 0x1549, 0x154e,
+ 0x154f, 0x1558, 0x1559, 0x155a, 0x155b, 0x1572, 0x159a, 0x159b,
+ 0x15ac, 0x15ba, 0x15bb, 0x15d0, 0x15d1, 0x15d2, 0x15d3, 0x15d4,
+ 0x15d5, 0x181d, 0x181e, 0x181f, 0x1840, 0x1841, 0x1842, 0x1843,
+ 0x1844, 0x1845, 0x1846, 0x1847, 0x1848, 0x1849, 0x1861, 0x1862,
+ 0x1863, 0x1864, 0x1865, 0x1866, 0x1867, 0x1868, 0x1869, 0x186a,
+ 0x186b, 0x186c, 0x186d, 0x186e, 0x191b, 0x191c, 0x191d, 0x191e,
+ 0x191f, 0x1942, 0x1943, 0x1944, 0x1945, 0x1946, 0x1947, 0x1958,
+ 0x1959, 0x19ed, 0x19ee, 0x19ef, 0x19f0, 0x19f1, 0x19f2, 0x19f3,
+ 0x19f4, 0x19f5, 0x19f6, 0x19f7, 0x1b0e, 0x1b0f, 0x1b62, 0x1b63,
+ 0x1b64, 0x1b65, 0x1b66, 0x1b67, 0x1b68, 0x1b69, 0x1b6a, 0x1b6b,
+ 0x1b6c, 0x1b6d, 0x1b6e, 0x1b6f, 0x1b82, 0x1ba8, 0x1ba9, 0x1baa,
+ 0x1bab, 0x1bac, 0x1bad, 0x1bae, 0x1baf, 0x1bb0, 0x1bb1, 0x1bb2,
+ 0x1bb3, 0x1d80, 0x1d81, 0x1d82, 0x1d83, 0x1d84, 0x1d85, 0x1d86,
+ 0x1d87, 0x1d88, 0x1d89, 0x1d8a, 0x1d8b, 0x1d8c, 0x1d8d, 0x1007,
+ 0x1008, 0x1009, 0x100a, 0x100b, 0x100c, 0x100d, 0x100e, 0x100f,
+ 0x1016, 0x1080, 0x1081, 0x1082, 0x1083, 0x1084, 0x1085, 0x1086,
+ 0x1087, 0x10c0, 0x123a, 0x123b, 0x123c, 0x123d, 0x123e, 0x123f,
+ 0x1240, 0x1241, 0x1242, 0x1243, 0x1350, 0x1352, 0x1353, 0x1358,
+ 0x1359, 0x135a, 0x135b, 0x135c, 0x135d, 0x135e, 0x135f, 0x1360,
+ 0x1361, 0x1602, 0x1603, 0x160c, 0x160d, 0x160e, 0x160f, 0x1620,
+ 0x1621, 0x1622, 0x1623, 0x1624, 0x1625, 0x1626, 0x1627, 0x1628,
+ 0x1629, 0x166e, 0x166f, 0x167c, 0x167d, 0x167e, 0x167f, 0x1770,
+ 0x1771, 0x1852, 0x1853, 0x1872, 0x1873, 0x1874, 0x1875, 0x1876,
+ 0x1877, 0x1878, 0x1879, 0x187a, 0x187b, 0x187c, 0x187d, 0x187e,
+ 0x187f, 0x1918, 0x1919, 0x1926, 0x1927, 0x1970, 0x1971, 0x1972,
+ 0x1973, 0x1974, 0x1975, 0x1976, 0x1977, 0x1978, 0x1979, 0x197a,
+ 0x197b, 0x1aa0, 0x1aa1, 0x1aa2, 0x1aa3, 0x1aa4, 0x1aa5, 0x1aa6,
+ 0x1aa7, 0x1aa8, 0x1aa9, 0x1aaa, 0x1aab, 0x1aac, 0x1aad, 0x1b3c,
+ 0x1b3d, 0x1b3e, 0x1b3f, 0x1b50, 0x1b51, 0x1b52, 0x1b53, 0x1b54,
+ 0x1b55, 0x1b56, 0x1b57, 0x1b58, 0x1b59, 0x2032, 0x2033, 0x2034,
+ 0x2035, 0x2036, 0x2037, 0x2038, 0x2039, 0x203a, 0x203b, 0x203c,
+ 0x203d, 0x203e, 0x203f, 0x2040, 0x2041, 0x2042, 0x2043, 0x20ba,
+ 0x20bb, 0x20cc, 0x20cd, 0x20ce, 0x20cf, 0x20e0, 0x20e1, 0x20e2,
+ 0x20e3, 0x20e4, 0x20e5, 0x20e6, 0x20e7, 0x21aa, 0x21ab, 0x21c0,
+ 0x21c1, 0x21c2, 0x21c3, 0x21c4, 0x21c5, 0x21c6, 0x21c7, 0x21c8,
+ 0x21c9, 0x21ca, 0x21cb, 0x21cc, 0x21cd, 0x21ce, 0x21cf, 0x21d0,
+ 0x21d1, 0x21d2, 0x21d3, 0x2894, 0x2895, 0x2896, 0x2897, 0x2898,
+ 0x2899, 0x289a, 0x289b, 0x289c, 0x289d, 0x289e, 0x289f, 0x28c0,
+ 0x28c1, 0x28c2, 0x28c3, 0x28c4, 0x28c5, 0x28c6, 0x28c7, 0x28c8,
+ 0x28c9, 0x28ca, 0x28cb, 0x2930, 0x2931, 0x2932, 0x2933, 0x2934,
+ 0x2935, 0x2936, 0x2937, 0x2938, 0x2939, 0x293a, 0x293b, 0x293c,
+ 0x293d, 0x293e, 0x293f, 0x2960, 0x2961, 0x2962, 0x2963, 0x2964,
+ 0x2965, 0x2966, 0x2967, 0x2968, 0x2969, 0x296a, 0x296b, 0x2a40,
+ 0x2a41, 0x2a42, 0x2a43, 0x2a44, 0x2a45, 0x2a46, 0x2a47, 0x2a48,
+ 0x2a49, 0x2a4a, 0x2a4b, 0x2a4c, 0x2a4d, 0x2a4e, 0x2a4f, 0x2a50,
+ 0x2a51, 0x2a52, 0x2a53, 0x2ae6, 0x2ae7, 0x2b24, 0x2b25, 0x2b26,
+ 0x2b27, 0x2b28, 0x2b29, 0x2b2a, 0x2b2b, 0x2b2c, 0x2b2d, 0x2b2e,
+ 0x2b2f, 0x2b30, 0x2b31, 0x2b32, 0x2b33, 0x2b5a, 0x2b5b, 0x3014,
+ 0x3015, 0x3016, 0x3017, 0x3020, 0x3021, 0x3022, 0x3023, 0x3024,
+ 0x3025, 0x3026, 0x3027, 0x3028, 0x3029, 0x302a, 0x302b, 0x302c,
+ 0x302d, 0x302e, 0x302f, 0x3030, 0x3031, 0x3032, 0x3033, 0x3034,
+ 0x3035, 0x3036, 0x3037, 0x3038, 0x3039, 0x30c0, 0x30c1, 0x30de,
+ 0x30df, 0x3218, 0x3219, 0x321a, 0x321b, 0x321c, 0x321d, 0x321e,
+ 0x321f, 0x3220, 0x3221, 0x3222, 0x3223, 0x3224, 0x3225, 0x3226,
+ 0x3227, 0x3228, 0x3229, 0x322a, 0x322b, 0x322c, 0x322d, 0x322e,
+ 0x322f, 0x3230, 0x3231, 0x3232, 0x3233, 0x3234, 0x3235, 0x3378,
+ 0x3379, 0x337a, 0x337b, 0x337c, 0x337d, 0x337e, 0x337f, 0x33c0,
+ 0x33c1, 0x33c2, 0x33c3, 0x33c4, 0x33c5, 0x33c6, 0x33c7, 0x33c8,
+ 0x33c9, 0x33ca, 0x33cb, 0x33cc, 0x33cd, 0x33ce, 0x33cf, 0x33d0,
+ 0x33d1, 0x33d2, 0x33d3, 0x33d4, 0x33d5, 0x33d6, 0x33d7, 0x33d8,
+ 0x33d9, 0x3706, 0x3707, 0x3730, 0x3731, 0x3732, 0x3733, 0x3734,
+ 0x3735, 0x3736, 0x3737, 0x3738, 0x3739, 0x373a, 0x373b, 0x373c,
+ 0x373d, 0x373e, 0x373f, 0x3740, 0x3741, 0x3742, 0x3743, 0x3744,
+ 0x3745, 0x3746, 0x3747, 0x3748, 0x3749, 0x374a, 0x374b, 0x374c,
+ 0x374d, 0x374e, 0x374f, 0x3b34, 0x3b35, 0x3b36, 0x3b37, 0x3be8,
+ 0x3be9, 0x3bea, 0x3beb, 0x3bec, 0x3bed, 0x3bee, 0x3bef, 0x3bf0,
+ 0x3bf1, 0x3bf2, 0x3bf3, 0x3bf4, 0x3bf5, 0x3bf6, 0x3bf7, 0x3bf8,
+ 0x3bf9, 0x3bfa, 0x3bfb, 0x3bfc, 0x3bfd, 0x3bfe, 0x3bff, 0x2000,
+ 0x2001, 0x2002, 0x2003, 0x2004, 0x2005, 0x2006, 0x2007, 0x2008,
+ 0x2009, 0x200a, 0x200b, 0x200c, 0x200d, 0x202e, 0x202f, 0x2182,
+ 0x2183, 0x21b4, 0x21b5, 0x21b6, 0x21b7, 0x21b8, 0x21b9, 0x21ba,
+ 0x21bb, 0x21bc, 0x21bd, 0x21be, 0x21bf, 0x2460, 0x2461, 0x2462,
+ 0x2463, 0x2464, 0x2465, 0x2466, 0x2467, 0x2468, 0x2469, 0x246a,
+ 0x246b, 0x246c, 0x246d, 0x246e, 0x246f, 0x2470, 0x2471, 0x2472,
+ 0x2473, 0x26a2, 0x26a3, 0x000b,
+};
+
+const UINT8 table1_mv_bits[1100] = {
+  2,  4,  4,  4,  5,  5,  5,  5,
+  6,  6,  7,  7,  7,  7,  7,  8,
+  8,  8,  8,  8,  8,  8,  8,  8,
+  8,  8,  8,  8,  8,  8,  8,  8,
+  8,  9,  9,  9,  9,  9,  9,  9,
+  9,  9,  9,  9,  9,  9,  9,  9,
+  9,  9,  9,  9,  9, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 13,
+ 13, 13, 13, 13, 13, 13, 13, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15, 15, 15, 15, 15, 15,
+ 15, 15, 15,  4,
+};
+
+const UINT8 table1_mvx[1099] = {
+ 32, 31, 32, 31, 33, 32, 33, 33,
+ 31, 34, 30, 32, 32, 34, 35, 32,
+ 34, 33, 29, 30, 30, 32, 31, 31,
+ 33, 35, 35, 33, 31, 29, 29, 33,
+ 34, 30, 31, 28, 36, 30, 34, 32,
+ 32, 37, 32, 32, 25, 27, 39, 32,
+ 32, 32, 38, 35, 36, 32, 37, 61,
+ 26, 32, 34, 35,  3, 35, 27, 28,
+ 29, 34, 28, 37, 31, 36, 32, 27,
+ 31, 30, 29, 39, 33, 29, 33, 35,
+ 25, 25, 29, 33, 31, 31, 31, 33,
+ 32, 30, 32, 32, 41, 39, 33, 36,
+ 32, 28, 34, 36, 38, 24, 60, 31,
+ 23, 28, 32, 33, 59, 32, 40, 30,
+  5, 34, 32, 38, 32, 30, 43,  4,
+ 32, 32, 42, 31, 31, 32, 26, 38,
+ 26, 22, 21, 37, 61, 63, 37, 31,
+ 32, 33,  2,  1, 23, 33, 41, 27,
+ 35, 30, 38, 23, 33,  3, 28, 34,
+ 34, 27, 41, 29, 39, 35, 36, 29,
+ 32, 27, 30, 32, 24, 61, 37, 26,
+ 59, 25, 35, 27, 36, 37, 30, 31,
+ 34, 40,  3, 28, 34, 39, 32, 31,
+ 32, 30, 24, 28, 35, 36, 26, 32,
+ 31, 33, 29, 33, 39, 25, 30, 24,
+ 35, 59, 29, 34, 25, 30, 21, 35,
+ 43, 40, 32, 29,  5, 28, 31, 62,
+ 33, 33, 25, 31, 21, 31, 43, 31,
+ 34, 33, 20, 40, 39, 31, 31, 57,
+ 38, 32, 42, 33, 32, 31, 32, 29,
+ 30, 44,  5, 31, 22, 34, 36, 17,
+ 38, 58, 38, 35, 32, 60, 35, 24,
+ 32, 38, 16, 45, 42, 32, 31, 29,
+  4, 30, 17, 40, 46, 48, 63, 32,
+ 42, 19, 41, 22, 28, 36, 45, 33,
+ 33, 32, 29,  7, 41, 42, 18, 33,
+ 33, 32, 22, 37,  1, 26, 22, 23,
+ 49, 28, 26, 27, 32, 33, 27, 23,
+ 28, 36, 15,  6, 34, 27, 31, 26,
+ 23,  2, 33, 32, 34, 41, 28, 32,
+ 41,  0, 36, 38, 34, 31, 47, 32,
+ 17, 31, 39, 33, 37, 51, 30, 47,
+ 32, 50, 32, 19, 63, 30, 25, 27,
+ 33, 62, 24, 31, 27, 30, 37, 31,
+ 45, 32, 39, 20, 46, 47, 35, 19,
+ 34,  1, 49, 21, 21, 14, 51, 26,
+ 23, 31, 36, 35, 58, 29, 29, 21,
+ 20, 42, 13, 28, 12, 40, 31, 33,
+ 39, 60, 32, 44, 33, 31, 28, 37,
+ 29, 32, 30, 49, 43, 28, 39, 25,
+ 32, 48,  2, 15, 20, 25, 31, 28,
+ 21, 24, 25, 15, 31, 17, 37, 43,
+ 18, 32, 33, 24, 33, 36, 13, 33,
+ 31, 39, 11, 31, 33, 32, 39, 37,
+ 32, 32, 29, 17, 44, 46, 36, 35,
+ 26, 37, 58, 32, 34, 38,  8, 38,
+ 38, 22, 29, 25, 16, 35, 32, 35,
+ 33, 43, 18, 46, 38, 50, 33, 18,
+ 53, 60, 13, 32, 36, 33, 51, 36,
+ 43, 45, 27, 42, 29, 24, 30, 25,
+ 31, 52, 31, 35, 38,  9, 22, 34,
+  4, 17, 28, 55, 42, 25, 17, 20,
+ 47, 34, 33, 16, 40, 25, 16, 30,
+ 53, 29, 10, 11, 14, 26, 33,  4,
+ 35, 44, 26, 16, 31, 26, 34, 38,
+ 29, 31, 30, 24, 22, 61, 32,  9,
+ 45, 34, 31, 19,  9, 31, 46, 31,
+ 35, 54, 29, 57, 30, 50,  3, 31,
+ 63, 34, 47, 41, 51, 18, 31, 14,
+ 37, 38, 31, 24, 32, 31, 50, 33,
+ 31, 54, 27,  9, 33, 23, 19, 32,
+ 29, 29, 33, 28, 47, 49, 30, 47,
+ 33, 27, 25, 54, 44, 45, 50, 58,
+ 51, 48, 33, 59, 33, 34, 57, 13,
+ 26, 33, 13, 48, 30, 11,  7, 56,
+ 34, 55, 26,  0, 26, 35,  1, 51,
+ 33, 53, 31, 45, 12, 29, 29, 51,
+ 31, 48,  2,  6, 34, 30, 28, 33,
+ 60, 40, 27, 46, 31,  9, 35, 29,
+ 31, 39, 55, 46, 19, 37, 62, 34,
+ 30, 16, 19, 49, 41, 41, 39, 37,
+ 14,  5, 13, 35, 55, 30, 40, 40,
+ 42,  8, 20, 25, 45, 35, 33, 36,
+ 54, 38, 27, 37, 62, 40, 15, 59,
+ 49, 31, 29, 34, 34, 39, 24, 29,
+ 25, 29, 21, 29, 10, 61, 33, 49,
+ 35, 34,  3, 38, 39, 29,  7, 41,
+  1, 35,  4, 23, 15, 23, 11, 37,
+ 28, 35, 30, 30, 24,  1, 43, 56,
+  8, 34, 42, 24, 45, 30, 20, 23,
+  8, 38, 22, 33, 17, 52, 34, 22,
+ 53, 43, 44,  1, 27, 31, 41, 43,
+ 41, 30, 31, 36, 30,  5, 55, 31,
+ 33, 30, 40, 23, 15, 29, 34, 34,
+ 59, 34, 30, 11, 13, 38,  5,  0,
+ 30, 42,  5, 30, 29, 34, 10, 44,
+ 30, 63, 35, 12,  3, 26, 15, 17,
+ 25, 34, 43, 39, 34, 56, 29, 23,
+ 30, 12, 30, 10, 35,  9, 24, 58,
+ 10, 12, 54, 33, 37, 20, 41, 35,
+ 29, 18, 61, 30, 40, 24, 39, 53,
+ 62, 26, 29, 33, 34, 53, 49, 21,
+ 27, 11, 63, 20, 26, 23,  7, 13,
+  6, 47, 29, 30,  9, 51, 22, 34,
+ 21, 25, 33, 56, 57, 30, 38, 51,
+ 51, 38, 63, 28, 40, 35, 33, 18,
+ 33, 33, 24, 58, 58, 34, 49, 29,
+ 43,  4,  1,  4, 42, 35, 35, 30,
+ 17,  5, 56, 61, 25, 37, 36, 55,
+ 28, 35, 29, 50, 48, 52,  2, 42,
+ 34, 40, 46, 46, 43, 35, 29, 48,
+ 20, 29, 31, 41,  7, 30, 35, 19,
+ 14, 21,  8, 39, 39, 40, 46, 55,
+ 34,  6, 30, 34, 37, 25, 37, 33,
+ 22, 44, 52, 17, 35, 29, 36, 35,
+ 40, 37, 28, 30, 50, 14, 28, 55,
+  6, 23, 19, 14, 30,  3, 30, 28,
+ 28, 61, 61, 47, 45, 48, 40, 40,
+ 34, 34, 25, 30, 29, 35,  4, 26,
+ 53, 50, 26, 41, 27, 59, 27, 38,
+ 39,  3, 50, 43, 47, 23, 33, 55,
+ 35, 21, 23, 35, 61, 33, 46, 52,
+ 35, 34, 24, 30, 43, 16, 37, 21,
+  2, 24, 45, 34, 30, 55, 55,  1,
+ 29, 29, 26, 28, 25, 31, 36, 22,
+ 17, 30, 52,  2, 44, 44, 57, 26,
+ 62, 41, 39, 57, 26, 46, 49, 11,
+ 16, 19,  5, 59, 38, 39, 58, 38,
+ 25, 49, 50, 22, 28, 59,  9, 59,
+  7, 28, 55, 17,  4, 35, 50, 21,
+ 29, 44, 47, 18, 24, 19, 25, 42,
+ 35,  3, 51, 35, 16, 35, 30, 63,
+ 57, 39, 39, 25, 35, 38,  9, 16,
+ 36, 45, 31, 60, 14, 34, 42, 24,
+  0, 37, 18, 61, 57, 37, 28, 53,
+ 20, 46, 14, 47, 38, 38, 38,  9,
+ 34, 39, 43, 17, 39, 59,  5, 27,
+  0, 12, 27,
+};
+
+const UINT8 table1_mvy[1099] = {
+ 32, 32, 31, 31, 32, 33, 31, 33,
+ 33, 32, 32, 30, 34, 31, 32, 29,
+ 33, 30, 32, 33, 31, 35, 34, 30,
+ 34, 31, 33, 29, 29, 31, 33, 35,
+ 30, 30, 35, 32, 32, 34, 34, 28,
+ 25, 32, 36, 27, 32, 32, 32, 37,
+ 39,  3, 32, 30, 31, 26, 31, 32,
+ 32, 38, 29, 29, 32, 34, 31, 31,
+ 34, 35, 33, 33, 28, 33,  1, 33,
+ 27, 29, 30, 31, 28, 29, 37, 35,
+ 31, 33, 35, 27, 36, 37, 25, 25,
+ 61, 35,  4,  5, 32, 33, 36, 30,
+ 23, 30, 28, 34, 31, 32, 32, 39,
+ 32, 34, 21, 39, 32, 59, 32, 28,
+ 32, 36, 60, 33, 24, 36, 32, 32,
+ 41,  2, 32, 38, 26, 22, 33, 30,
+ 31, 32, 32, 30, 31, 32, 29,  3,
+ 40, 38, 32, 32, 33, 26, 31, 34,
+ 28, 38, 34, 31,  3, 31, 35, 38,
+ 27, 35, 33, 28, 29, 27, 29, 27,
+ 43, 29, 37, 63, 31, 33, 34, 30,
+ 31, 30, 37, 30, 35, 35, 26, 41,
+ 37, 31, 33, 28, 26, 30, 42, 24,
+  7, 27, 33, 29, 36, 28, 34, 57,
+ 23, 41, 36, 23, 35, 34, 25, 30,
+ 25, 33, 25, 25, 29, 24, 33, 39,
+ 33, 33,  0, 37, 31, 36, 21, 32,
+ 61, 24, 35, 61, 31,  5, 31, 59,
+ 39, 21, 32, 30, 34, 22, 40, 32,
+ 29, 16, 31,  5, 62,  2, 20, 39,
+ 39, 32, 33,  1, 31, 24, 36, 32,
+ 36, 32, 28, 26,  6, 31, 38, 34,
+ 58, 35, 32, 33, 33, 17, 43, 26,
+ 31, 40, 31, 34, 32, 32, 31, 19,
+ 30, 32, 29, 33, 38, 38, 32, 59,
+ 40, 18, 38, 32, 35, 34, 32, 17,
+  1, 15, 30, 28, 31, 28, 34, 29,
+ 32, 27, 35, 27, 49, 22, 37, 34,
+ 37, 26, 32, 32, 22, 28, 45, 29,
+ 30, 31, 43, 46, 41, 30, 26, 13,
+ 34, 32, 27, 38, 42, 42, 33, 47,
+ 33, 60, 27, 42, 25, 32, 22, 32,
+ 48, 32, 45, 33, 33, 41, 27, 25,
+ 19, 31, 35, 19, 36, 42, 27, 17,
+ 31, 44, 28, 33, 33, 31, 23, 31,
+ 40, 33, 31, 34, 30, 32, 33, 36,
+ 35, 47, 37, 41, 31, 23, 41, 29,
+ 30, 35, 32, 25, 32, 28, 58,  2,
+ 37, 33, 14, 33, 49, 20, 39, 36,
+ 21,  9, 23, 33, 35, 24, 39, 37,
+ 11, 33, 30, 31, 31, 28, 51, 40,
+ 35, 29, 25, 33, 46, 35, 37, 30,
+ 30,  8, 63, 28, 15, 40, 33, 45,
+ 49, 25, 32,  4, 47, 51, 36, 39,
+ 53, 10, 24, 29, 30, 31, 25, 40,
+ 38, 38, 33, 56, 23, 27, 32, 37,
+ 26, 29, 43, 36, 33, 24, 55, 43,
+  9, 29, 34, 34, 24, 33, 18, 33,
+ 33, 30, 31, 50, 24, 60, 30, 39,
+ 34, 30, 39, 28, 22, 38,  2, 26,
+ 63, 32, 57, 21, 39, 33, 28, 18,
+ 30, 34, 22, 33, 29, 41, 30, 34,
+ 35, 21, 13, 34, 35, 39, 30, 46,
+ 32, 42, 32, 31, 33, 26, 11, 33,
+ 22, 31, 25, 31, 53, 27, 43, 25,
+ 40, 50, 21, 36, 38, 30, 12, 31,
+ 34, 20, 15, 29, 32, 62, 30, 13,
+ 17, 32, 19, 31, 20, 31, 30,  7,
+  1, 17, 34, 37, 31, 31, 44, 34,
+ 26, 40, 16, 37, 52, 48, 30, 20,
+ 18, 33, 38, 29,  7, 25, 30, 54,
+ 45, 47, 46, 41, 29, 29, 16, 30,
+ 14, 26, 38, 34, 34, 29, 34, 30,
+ 29, 30, 57, 30,  4, 46, 33, 29,
+ 39, 44, 30, 31, 50, 33, 31, 32,
+ 19, 32, 40, 31, 37, 47,  1, 35,
+ 16, 31,  0, 35, 33,  1, 17, 34,
+  9, 34, 33, 31, 49, 43, 42, 51,
+ 34, 29, 23, 29, 14, 30, 45, 49,
+ 11, 24, 31, 28, 35, 41, 30, 44,
+ 18, 29, 34, 35, 36, 25, 26, 21,
+ 31, 30, 34, 19, 34, 44, 36, 38,
+ 25, 31, 28, 23, 37,  3, 55, 41,
+ 30, 22, 41, 24, 33, 26, 35, 35,
+ 30, 55, 51, 47, 48, 38, 24, 15,
+ 21, 50, 25, 46, 30, 29, 10, 34,
+ 42, 45, 29, 42, 22,  3, 33, 27,
+ 34,  1, 34, 28, 34, 36, 35, 23,
+ 23, 13, 58,  3, 26, 63, 25, 31,
+ 34, 61, 38, 39, 25, 61, 29, 37,
+ 30, 41, 26, 48, 28, 33, 50, 35,
+ 30, 37, 29, 29, 40,  6, 39, 28,
+ 28, 19,  8, 22, 45, 34, 35, 10,
+ 58, 17, 37, 39, 30, 18, 54, 14,
+ 29, 16, 59, 30, 35, 23, 35, 30,
+ 47, 36, 29, 55, 20, 12, 31, 35,
+ 14, 29, 18, 34, 34, 24, 29, 26,
+ 22,  2, 27, 23,  8, 30, 55, 38,
+ 60, 31,  4, 34, 49, 34, 27, 34,
+ 33, 30, 31, 54, 42, 35, 38, 46,
+ 44, 26, 27,  9, 39, 25, 21, 29,
+ 28, 42, 13,  0,  5, 34, 37, 28,
+ 24, 29, 63, 26, 22, 27, 29, 25,
+ 33, 25, 61,  0, 35, 25, 36, 15,
+ 27, 40, 53, 33,  3, 10, 16, 37,
+ 38, 18, 30, 46, 27,  9,  6, 29,
+ 62,  8, 42, 28, 29,  3, 25, 16,
+ 26, 29, 35, 28, 27, 51, 61, 48,
+ 37,  9, 34,  7, 49, 45, 20, 29,
+ 21,  5,  5, 29, 28, 34, 29, 24,
+ 10, 24, 35, 36, 38, 55, 11, 36,
+ 38, 53, 54, 26, 30, 49, 20, 27,
+ 30, 39, 33, 41, 49, 22, 38, 38,
+  4, 30,  8,  9,  3, 24, 22, 50,
+ 37, 36, 31, 27,  2,  9, 42, 63,
+ 25, 19, 44,  1, 28, 28, 48, 30,
+ 34, 41, 41, 38, 12, 27, 15,  0,
+ 16, 34, 35, 38, 28, 29, 40, 42,
+ 51, 52, 45, 54, 59, 59, 42, 44,
+ 37, 26, 46, 24, 15, 39, 22, 46,
+ 19, 35, 38, 17, 37, 23, 52, 55,
+ 50, 37, 26, 11, 37, 12, 24, 30,
+ 16, 13, 22, 13, 36, 35, 40, 41,
+ 34, 41, 26, 53, 51,  5, 21, 30,
+  2, 63, 41, 20,  1, 56, 21, 24,
+ 25,  5, 28, 35, 26, 28, 30, 18,
+ 29, 23, 40, 34, 20, 42, 39, 34,
+ 28, 61, 38, 27, 62,  9, 36, 17,
+  9, 49, 24, 25, 54, 34, 39, 37,
+  3,  1, 25, 38, 38, 44, 35, 36,
+ 12, 60, 36, 38, 40, 25, 43, 39,
+ 53, 28, 39, 57, 46, 10, 52, 27,
+ 35, 42, 45, 59, 15, 60, 38, 24,
+ 23, 39, 12, 29, 24,  0, 20, 16,
+ 28, 43, 35, 28,  1, 49,  4, 21,
+ 42, 39, 29,  3, 44, 21, 53, 55,
+ 11,  5,  3, 39, 53, 28, 25, 19,
+ 34, 28, 21,
+};
+
+static MVTable mv_tables[2] = {
+    {
+        1099,
+        table0_mv_code,
+        table0_mv_bits,
+        table0_mvx,
+        table0_mvy,
+    },
+    {
+        1099,
+        table1_mv_code,
+        table1_mv_bits,
+        table1_mvx,
+        table1_mvy,
+    }
+};
diff --git a/libavcodec/resample.c b/libavcodec/resample.c
new file mode 100644
index 0000000000..29445964f6
--- /dev/null
+++ b/libavcodec/resample.c
@@ -0,0 +1,301 @@
+/*
+ * Sample rate convertion for both audio and video
+ * Copyright (c) 2000 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <math.h>
+#include "avcodec.h"
+
+#define NDEBUG
+#include <assert.h>
+
+typedef struct {
+    /* fractional resampling */
+    UINT32 incr; /* fractional increment */
+    UINT32 frac;
+    int last_sample;
+    /* integer down sample */
+    int iratio;  /* integer divison ratio */
+    int icount, isum;
+    int inv;
+} ReSampleChannelContext;
+
+struct ReSampleContext {
+    ReSampleChannelContext channel_ctx[2];
+    float ratio;
+    /* channel convert */
+    int input_channels, output_channels, filter_channels;
+};
+
+
+#define FRAC_BITS 16
+#define FRAC (1 << FRAC_BITS)
+
+static void init_mono_resample(ReSampleChannelContext *s, float ratio)
+{
+    ratio = 1.0 / ratio;
+    s->iratio = (int)floor(ratio);
+    if (s->iratio == 0)
+        s->iratio = 1;
+    s->incr = (int)((ratio / s->iratio) * FRAC);
+    s->frac = 0;
+    s->last_sample = 0;
+    s->icount = s->iratio;
+    s->isum = 0;
+    s->inv = (FRAC / s->iratio);
+}
+
+/* fractional audio resampling */
+static int fractional_resample(ReSampleChannelContext *s, short *output, short *input, int nb_samples)
+{
+    unsigned int frac, incr;
+    int l0, l1;
+    short *q, *p, *pend;
+
+    l0 = s->last_sample;
+    incr = s->incr;
+    frac = s->frac;
+
+    p = input;
+    pend = input + nb_samples;
+    q = output;
+
+    l1 = *p++;
+    for(;;) {
+        /* interpolate */
+        *q++ = (l0 * (FRAC - frac) + l1 * frac) >> FRAC_BITS;
+        frac = frac + s->incr;
+        while (frac >= FRAC) {
+            if (p >= pend)
+                goto the_end;
+            frac -= FRAC;
+            l0 = l1;
+            l1 = *p++;
+        }
+    }
+ the_end:
+    s->last_sample = l1;
+    s->frac = frac;
+    return q - output;
+}
+
+static int integer_downsample(ReSampleChannelContext *s, short *output, short *input, int nb_samples)
+{
+    short *q, *p, *pend;
+    int c, sum;
+
+    p = input;
+    pend = input + nb_samples;
+    q = output;
+
+    c = s->icount;
+    sum = s->isum;
+
+    for(;;) {
+        sum += *p++;
+        if (--c == 0) {
+            *q++ = (sum * s->inv) >> FRAC_BITS;
+            c = s->iratio;
+            sum = 0;
+        }
+        if (p >= pend)
+            break;
+    }
+    s->isum = sum;
+    s->icount = c;
+    return q - output;
+}
+
+/* n1: number of samples */
+static void stereo_to_mono(short *output, short *input, int n1)
+{
+    short *p, *q;
+    int n = n1;
+
+    p = input;
+    q = output;
+    while (n >= 4) {
+        q[0] = (p[0] + p[1]) >> 1;
+        q[1] = (p[2] + p[3]) >> 1;
+        q[2] = (p[4] + p[5]) >> 1;
+        q[3] = (p[6] + p[7]) >> 1;
+        q += 4;
+        p += 8;
+        n -= 4;
+    }
+    while (n > 0) {
+        q[0] = (p[0] + p[1]) >> 1;
+        q++;
+        p += 2;
+        n--;
+    }
+}
+
+/* n1: number of samples */
+static void mono_to_stereo(short *output, short *input, int n1)
+{
+    short *p, *q;
+    int n = n1;
+    int v;
+
+    p = input;
+    q = output;
+    while (n >= 4) {
+        v = p[0]; q[0] = v; q[1] = v;
+        v = p[1]; q[2] = v; q[3] = v;
+        v = p[2]; q[4] = v; q[5] = v;
+        v = p[3]; q[6] = v; q[7] = v;
+        q += 8;
+        p += 4;
+        n -= 4;
+    }
+    while (n > 0) {
+        v = p[0]; q[0] = v; q[1] = v;
+        q += 2;
+        p += 1;
+        n--;
+    }
+}
+
+/* XXX: should use more abstract 'N' channels system */
+static void stereo_split(short *output1, short *output2, short *input, int n)
+{
+    int i;
+
+    for(i=0;i<n;i++) {
+        *output1++ = *input++;
+        *output2++ = *input++;
+    }
+}
+
+static void stereo_mux(short *output, short *input1, short *input2, int n)
+{
+    int i;
+
+    for(i=0;i<n;i++) {
+        *output++ = *input1++;
+        *output++ = *input2++;
+    }
+}
+
+static int mono_resample(ReSampleChannelContext *s, short *output, short *input, int nb_samples)
+{
+    short buf1[nb_samples];
+    short *buftmp;
+
+    /* first downsample by an integer factor with averaging filter */
+    if (s->iratio > 1) {
+        buftmp = buf1;
+        nb_samples = integer_downsample(s, buftmp, input, nb_samples);
+    } else {
+        buftmp = input;
+    }
+
+    /* then do a fractional resampling with linear interpolation */
+    if (s->incr != FRAC) {
+        nb_samples = fractional_resample(s, output, buftmp, nb_samples);
+    } else {
+        memcpy(output, buftmp, nb_samples * sizeof(short));
+    }
+    return nb_samples;
+}
+
+ReSampleContext *audio_resample_init(int output_channels, int input_channels, 
+                                      int output_rate, int input_rate)
+{
+    ReSampleContext *s;
+    int i;
+    
+    if (output_channels > 2 || input_channels > 2)
+        return NULL;
+
+    s = av_mallocz(sizeof(ReSampleContext));
+    if (!s)
+        return NULL;
+
+    s->ratio = (float)output_rate / (float)input_rate;
+    
+    s->input_channels = input_channels;
+    s->output_channels = output_channels;
+    
+    s->filter_channels = s->input_channels;
+    if (s->output_channels < s->filter_channels)
+        s->filter_channels = s->output_channels;
+
+    for(i=0;i<s->filter_channels;i++) {
+        init_mono_resample(&s->channel_ctx[i], s->ratio);
+    }
+    return s;
+}
+
+/* resample audio. 'nb_samples' is the number of input samples */
+/* XXX: optimize it ! */
+/* XXX: do it with polyphase filters, since the quality here is
+   HORRIBLE. Return the number of samples available in output */
+int audio_resample(ReSampleContext *s, short *output, short *input, int nb_samples)
+{
+    int i, nb_samples1;
+    short bufin[2][nb_samples];
+    short bufout[2][(int)(nb_samples * s->ratio) + 16]; /* make some zoom to avoid round pb */
+    short *buftmp2[2], *buftmp3[2];
+
+    if (s->input_channels == s->output_channels && s->ratio == 1.0) {
+        /* nothing to do */
+        memcpy(output, input, nb_samples * s->input_channels * sizeof(short));
+        return nb_samples;
+    }
+
+    if (s->input_channels == 2 &&
+        s->output_channels == 1) {
+        buftmp2[0] = bufin[0];
+        buftmp3[0] = output;
+        stereo_to_mono(buftmp2[0], input, nb_samples);
+    } else if (s->output_channels == 2 && s->input_channels == 1) {
+        buftmp2[0] = input;
+        buftmp3[0] = bufout[0];
+    } else if (s->output_channels == 2) {
+        buftmp2[0] = bufin[0];
+        buftmp2[1] = bufin[1];
+        buftmp3[0] = bufout[0];
+        buftmp3[1] = bufout[1];
+        stereo_split(buftmp2[0], buftmp2[1], input, nb_samples);
+    } else {
+        buftmp2[0] = input;
+        buftmp3[0] = output;
+    }
+
+    /* resample each channel */
+    nb_samples1 = 0; /* avoid warning */
+    for(i=0;i<s->filter_channels;i++) {
+        nb_samples1 = mono_resample(&s->channel_ctx[i], buftmp3[i], buftmp2[i], nb_samples);
+    }
+
+    if (s->output_channels == 2 && s->input_channels == 1) {
+        mono_to_stereo(output, buftmp3[0], nb_samples1);
+    } else if (s->output_channels == 2) {
+        stereo_mux(output, buftmp3[0], buftmp3[1], nb_samples1);
+    }
+
+    return nb_samples1;
+}
+
+void audio_resample_close(ReSampleContext *s)
+{
+    free(s);
+}
diff --git a/libavcodec/rv10.c b/libavcodec/rv10.c
new file mode 100644
index 0000000000..b03cd02937
--- /dev/null
+++ b/libavcodec/rv10.c
@@ -0,0 +1,487 @@
+/*
+ * RV10 codec
+ * Copyright (c) 2000,2001 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include "common.h"
+#include "dsputil.h"
+#include "avcodec.h"
+#include "mpegvideo.h"
+
+//#define DEBUG
+
+static const UINT16 rv_lum_code[256] =
+{
+ 0x3e7f, 0x0f00, 0x0f01, 0x0f02, 0x0f03, 0x0f04, 0x0f05, 0x0f06,
+ 0x0f07, 0x0f08, 0x0f09, 0x0f0a, 0x0f0b, 0x0f0c, 0x0f0d, 0x0f0e,
+ 0x0f0f, 0x0f10, 0x0f11, 0x0f12, 0x0f13, 0x0f14, 0x0f15, 0x0f16,
+ 0x0f17, 0x0f18, 0x0f19, 0x0f1a, 0x0f1b, 0x0f1c, 0x0f1d, 0x0f1e,
+ 0x0f1f, 0x0f20, 0x0f21, 0x0f22, 0x0f23, 0x0f24, 0x0f25, 0x0f26,
+ 0x0f27, 0x0f28, 0x0f29, 0x0f2a, 0x0f2b, 0x0f2c, 0x0f2d, 0x0f2e,
+ 0x0f2f, 0x0f30, 0x0f31, 0x0f32, 0x0f33, 0x0f34, 0x0f35, 0x0f36,
+ 0x0f37, 0x0f38, 0x0f39, 0x0f3a, 0x0f3b, 0x0f3c, 0x0f3d, 0x0f3e,
+ 0x0f3f, 0x0380, 0x0381, 0x0382, 0x0383, 0x0384, 0x0385, 0x0386,
+ 0x0387, 0x0388, 0x0389, 0x038a, 0x038b, 0x038c, 0x038d, 0x038e,
+ 0x038f, 0x0390, 0x0391, 0x0392, 0x0393, 0x0394, 0x0395, 0x0396,
+ 0x0397, 0x0398, 0x0399, 0x039a, 0x039b, 0x039c, 0x039d, 0x039e,
+ 0x039f, 0x00c0, 0x00c1, 0x00c2, 0x00c3, 0x00c4, 0x00c5, 0x00c6,
+ 0x00c7, 0x00c8, 0x00c9, 0x00ca, 0x00cb, 0x00cc, 0x00cd, 0x00ce,
+ 0x00cf, 0x0050, 0x0051, 0x0052, 0x0053, 0x0054, 0x0055, 0x0056,
+ 0x0057, 0x0020, 0x0021, 0x0022, 0x0023, 0x000c, 0x000d, 0x0004,
+ 0x0000, 0x0005, 0x000e, 0x000f, 0x0024, 0x0025, 0x0026, 0x0027,
+ 0x0058, 0x0059, 0x005a, 0x005b, 0x005c, 0x005d, 0x005e, 0x005f,
+ 0x00d0, 0x00d1, 0x00d2, 0x00d3, 0x00d4, 0x00d5, 0x00d6, 0x00d7,
+ 0x00d8, 0x00d9, 0x00da, 0x00db, 0x00dc, 0x00dd, 0x00de, 0x00df,
+ 0x03a0, 0x03a1, 0x03a2, 0x03a3, 0x03a4, 0x03a5, 0x03a6, 0x03a7,
+ 0x03a8, 0x03a9, 0x03aa, 0x03ab, 0x03ac, 0x03ad, 0x03ae, 0x03af,
+ 0x03b0, 0x03b1, 0x03b2, 0x03b3, 0x03b4, 0x03b5, 0x03b6, 0x03b7,
+ 0x03b8, 0x03b9, 0x03ba, 0x03bb, 0x03bc, 0x03bd, 0x03be, 0x03bf,
+ 0x0f40, 0x0f41, 0x0f42, 0x0f43, 0x0f44, 0x0f45, 0x0f46, 0x0f47,
+ 0x0f48, 0x0f49, 0x0f4a, 0x0f4b, 0x0f4c, 0x0f4d, 0x0f4e, 0x0f4f,
+ 0x0f50, 0x0f51, 0x0f52, 0x0f53, 0x0f54, 0x0f55, 0x0f56, 0x0f57,
+ 0x0f58, 0x0f59, 0x0f5a, 0x0f5b, 0x0f5c, 0x0f5d, 0x0f5e, 0x0f5f,
+ 0x0f60, 0x0f61, 0x0f62, 0x0f63, 0x0f64, 0x0f65, 0x0f66, 0x0f67,
+ 0x0f68, 0x0f69, 0x0f6a, 0x0f6b, 0x0f6c, 0x0f6d, 0x0f6e, 0x0f6f,
+ 0x0f70, 0x0f71, 0x0f72, 0x0f73, 0x0f74, 0x0f75, 0x0f76, 0x0f77,
+ 0x0f78, 0x0f79, 0x0f7a, 0x0f7b, 0x0f7c, 0x0f7d, 0x0f7e, 0x0f7f,
+};
+
+static const UINT8 rv_lum_bits[256] = 
+{
+ 14, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 10,  8,  8,  8,  8,  8,  8,  8,
+  8,  8,  8,  8,  8,  8,  8,  8,
+  8,  7,  7,  7,  7,  7,  7,  7,
+  7,  6,  6,  6,  6,  5,  5,  4,
+  2,  4,  5,  5,  6,  6,  6,  6,
+  7,  7,  7,  7,  7,  7,  7,  7,
+  8,  8,  8,  8,  8,  8,  8,  8,
+  8,  8,  8,  8,  8,  8,  8,  8,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+};
+
+static const UINT16 rv_chrom_code[256] =
+{
+ 0xfe7f, 0x3f00, 0x3f01, 0x3f02, 0x3f03, 0x3f04, 0x3f05, 0x3f06,
+ 0x3f07, 0x3f08, 0x3f09, 0x3f0a, 0x3f0b, 0x3f0c, 0x3f0d, 0x3f0e,
+ 0x3f0f, 0x3f10, 0x3f11, 0x3f12, 0x3f13, 0x3f14, 0x3f15, 0x3f16,
+ 0x3f17, 0x3f18, 0x3f19, 0x3f1a, 0x3f1b, 0x3f1c, 0x3f1d, 0x3f1e,
+ 0x3f1f, 0x3f20, 0x3f21, 0x3f22, 0x3f23, 0x3f24, 0x3f25, 0x3f26,
+ 0x3f27, 0x3f28, 0x3f29, 0x3f2a, 0x3f2b, 0x3f2c, 0x3f2d, 0x3f2e,
+ 0x3f2f, 0x3f30, 0x3f31, 0x3f32, 0x3f33, 0x3f34, 0x3f35, 0x3f36,
+ 0x3f37, 0x3f38, 0x3f39, 0x3f3a, 0x3f3b, 0x3f3c, 0x3f3d, 0x3f3e,
+ 0x3f3f, 0x0f80, 0x0f81, 0x0f82, 0x0f83, 0x0f84, 0x0f85, 0x0f86,
+ 0x0f87, 0x0f88, 0x0f89, 0x0f8a, 0x0f8b, 0x0f8c, 0x0f8d, 0x0f8e,
+ 0x0f8f, 0x0f90, 0x0f91, 0x0f92, 0x0f93, 0x0f94, 0x0f95, 0x0f96,
+ 0x0f97, 0x0f98, 0x0f99, 0x0f9a, 0x0f9b, 0x0f9c, 0x0f9d, 0x0f9e,
+ 0x0f9f, 0x03c0, 0x03c1, 0x03c2, 0x03c3, 0x03c4, 0x03c5, 0x03c6,
+ 0x03c7, 0x03c8, 0x03c9, 0x03ca, 0x03cb, 0x03cc, 0x03cd, 0x03ce,
+ 0x03cf, 0x00e0, 0x00e1, 0x00e2, 0x00e3, 0x00e4, 0x00e5, 0x00e6,
+ 0x00e7, 0x0030, 0x0031, 0x0032, 0x0033, 0x0008, 0x0009, 0x0002,
+ 0x0000, 0x0003, 0x000a, 0x000b, 0x0034, 0x0035, 0x0036, 0x0037,
+ 0x00e8, 0x00e9, 0x00ea, 0x00eb, 0x00ec, 0x00ed, 0x00ee, 0x00ef,
+ 0x03d0, 0x03d1, 0x03d2, 0x03d3, 0x03d4, 0x03d5, 0x03d6, 0x03d7,
+ 0x03d8, 0x03d9, 0x03da, 0x03db, 0x03dc, 0x03dd, 0x03de, 0x03df,
+ 0x0fa0, 0x0fa1, 0x0fa2, 0x0fa3, 0x0fa4, 0x0fa5, 0x0fa6, 0x0fa7,
+ 0x0fa8, 0x0fa9, 0x0faa, 0x0fab, 0x0fac, 0x0fad, 0x0fae, 0x0faf,
+ 0x0fb0, 0x0fb1, 0x0fb2, 0x0fb3, 0x0fb4, 0x0fb5, 0x0fb6, 0x0fb7,
+ 0x0fb8, 0x0fb9, 0x0fba, 0x0fbb, 0x0fbc, 0x0fbd, 0x0fbe, 0x0fbf,
+ 0x3f40, 0x3f41, 0x3f42, 0x3f43, 0x3f44, 0x3f45, 0x3f46, 0x3f47,
+ 0x3f48, 0x3f49, 0x3f4a, 0x3f4b, 0x3f4c, 0x3f4d, 0x3f4e, 0x3f4f,
+ 0x3f50, 0x3f51, 0x3f52, 0x3f53, 0x3f54, 0x3f55, 0x3f56, 0x3f57,
+ 0x3f58, 0x3f59, 0x3f5a, 0x3f5b, 0x3f5c, 0x3f5d, 0x3f5e, 0x3f5f,
+ 0x3f60, 0x3f61, 0x3f62, 0x3f63, 0x3f64, 0x3f65, 0x3f66, 0x3f67,
+ 0x3f68, 0x3f69, 0x3f6a, 0x3f6b, 0x3f6c, 0x3f6d, 0x3f6e, 0x3f6f,
+ 0x3f70, 0x3f71, 0x3f72, 0x3f73, 0x3f74, 0x3f75, 0x3f76, 0x3f77,
+ 0x3f78, 0x3f79, 0x3f7a, 0x3f7b, 0x3f7c, 0x3f7d, 0x3f7e, 0x3f7f,
+};
+
+static const UINT8 rv_chrom_bits[256] =
+{
+ 16, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 10,  8,  8,  8,  8,  8,  8,  8,
+  8,  6,  6,  6,  6,  4,  4,  3,
+  2,  3,  4,  4,  6,  6,  6,  6,
+  8,  8,  8,  8,  8,  8,  8,  8,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 12, 12, 12, 12, 12, 12, 12, 12,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+ 14, 14, 14, 14, 14, 14, 14, 14,
+};
+
+static VLC rv_dc_lum, rv_dc_chrom;
+
+int rv_decode_dc(MpegEncContext *s, int n)
+{
+    int code;
+
+    if (n < 4) {
+        code = get_vlc(&s->gb, &rv_dc_lum);
+        if (code < 0) {
+            /* XXX: I don't understand why they use LONGER codes than
+               necessary. The following code would be completely useless
+               if they had thought about it !!! */
+            code = get_bits(&s->gb, 7);
+            if (code == 0x7c) {
+                code = (INT8)(get_bits(&s->gb, 7) + 1);
+            } else if (code == 0x7d) {
+                code = -128 + get_bits(&s->gb, 7);
+            } else if (code == 0x7e) {
+                if (get_bits(&s->gb, 1) == 0)
+                    code = (INT8)(get_bits(&s->gb, 8) + 1);
+                else
+                    code = (INT8)(get_bits(&s->gb, 8));
+            } else if (code == 0x7f) {
+                get_bits(&s->gb, 11);
+                code = 1;
+            }
+        } else {
+            code -= 128;
+        }
+    } else {
+        code = get_vlc(&s->gb, &rv_dc_chrom);
+        /* same remark */
+        if (code < 0) {
+            code = get_bits(&s->gb, 9);
+            if (code == 0x1fc) {
+                code = (INT8)(get_bits(&s->gb, 7) + 1);
+            } else if (code == 0x1fd) {
+                code = -128 + get_bits(&s->gb, 7);
+            } else if (code == 0x1fe) {
+                get_bits(&s->gb, 9);
+                code = 1;
+            } else {
+                return 0xffff;
+            }
+        } else {
+            code -= 128;
+        }
+    }
+    return -code;
+}
+
+/* write RV 1.0 compatible frame header */
+void rv10_encode_picture_header(MpegEncContext *s, int picture_number)
+{
+    align_put_bits(&s->pb);
+
+    put_bits(&s->pb, 1, 1);	/* marker */
+
+    put_bits(&s->pb, 1, (s->pict_type == P_TYPE));
+
+    put_bits(&s->pb, 1, 0);	/* not PB frame */
+
+    put_bits(&s->pb, 5, s->qscale);
+
+    if (s->pict_type == I_TYPE) {
+	/* specific MPEG like DC coding not used */
+    }
+    /* if multiple packets per frame are sent, the position at which
+       to display the macro blocks is coded here */
+    put_bits(&s->pb, 6, 0);	/* mb_x */
+    put_bits(&s->pb, 6, 0);	/* mb_y */
+    put_bits(&s->pb, 12, s->mb_width * s->mb_height);
+
+    put_bits(&s->pb, 3, 0);	/* ignored */
+}
+
+static int get_num(GetBitContext *gb)
+{
+    int n, n1;
+
+    n = get_bits(gb, 16);
+    if (n >= 0x4000) {
+        return n - 0x4000;
+    } else {
+        n1 = get_bits(gb, 16);
+        return (n << 16) | n1;
+    }
+}
+
+/* read RV 1.0 compatible frame header */
+static int rv10_decode_picture_header(MpegEncContext *s)
+{
+    int mb_count, pb_frame, marker, h, full_frame;
+    
+    /* skip packet header */
+    h = get_bits(&s->gb, 8);
+    if ((h & 0xc0) == 0xc0) {
+        int len, pos;
+        full_frame = 1;
+        len = get_num(&s->gb);
+        pos = get_num(&s->gb);
+    } else {
+        int seq, frame_size, pos;
+        full_frame = 0;
+        seq = get_bits(&s->gb, 8);
+        frame_size = get_num(&s->gb);
+        pos = get_num(&s->gb);
+    }
+    /* picture number */
+    get_bits(&s->gb, 8);
+
+    marker = get_bits(&s->gb, 1);
+
+    if (get_bits(&s->gb, 1))
+        s->pict_type = P_TYPE;
+    else
+        s->pict_type = I_TYPE;
+
+    pb_frame = get_bits(&s->gb, 1);
+
+#ifdef DEBUG
+    printf("pict_type=%d pb_frame=%d\n", s->pict_type, pb_frame);
+#endif
+    
+    if (pb_frame)
+        return -1;
+
+    s->qscale = get_bits(&s->gb, 5);
+
+    if (s->pict_type == I_TYPE) {
+        if (s->rv10_version == 3) {
+            /* specific MPEG like DC coding not used */
+            s->last_dc[0] = get_bits(&s->gb, 8);
+            s->last_dc[1] = get_bits(&s->gb, 8);
+            s->last_dc[2] = get_bits(&s->gb, 8);
+#ifdef DEBUG
+            printf("DC:%d %d %d\n",
+                   s->last_dc[0],
+                   s->last_dc[1],
+                   s->last_dc[2]);
+#endif
+        }
+    }
+    /* if multiple packets per frame are sent, the position at which
+       to display the macro blocks is coded here */
+    if (!full_frame) {
+        s->mb_x = get_bits(&s->gb, 6);	/* mb_x */
+        s->mb_y = get_bits(&s->gb, 6);	/* mb_y */
+        mb_count = get_bits(&s->gb, 12);
+    } else {
+        s->mb_x = 0;
+        s->mb_y = 0;
+        mb_count = s->mb_width * s->mb_height;
+    }
+
+    get_bits(&s->gb, 3);	/* ignored */
+    s->f_code = 1;
+    s->unrestricted_mv = 1;
+#if 0
+    s->h263_long_vectors = 1;
+#endif
+    return mb_count;
+}
+
+static int rv10_decode_init(AVCodecContext *avctx)
+{
+    MpegEncContext *s = avctx->priv_data;
+    static int done;
+
+    s->out_format = FMT_H263;
+
+    s->width = avctx->width;
+    s->height = avctx->height;
+
+    s->h263_rv10 = 1;
+    s->rv10_version = avctx->sub_id;
+
+    if (MPV_common_init(s) < 0)
+        return -1;
+
+    h263_decode_init_vlc(s);
+
+    /* init rv vlc */
+    if (!done) {
+        init_vlc(&rv_dc_lum, 9, 256, 
+                 rv_lum_bits, 1, 1,
+                 rv_lum_code, 2, 2);
+        init_vlc(&rv_dc_chrom, 9, 256, 
+                 rv_chrom_bits, 1, 1,
+                 rv_chrom_code, 2, 2);
+        done = 1;
+    }
+
+    return 0;
+}
+
+static int rv10_decode_end(AVCodecContext *avctx)
+{
+    MpegEncContext *s = avctx->priv_data;
+
+    MPV_common_end(s);
+    return 0;
+}
+
+static int rv10_decode_frame(AVCodecContext *avctx, 
+                             void *data, int *data_size,
+                             UINT8 *buf, int buf_size)
+{
+    MpegEncContext *s = avctx->priv_data;
+    int i, mb_count, mb_pos, left;
+    DCTELEM block[6][64];
+    AVPicture *pict = data; 
+
+#ifdef DEBUG
+    printf("*****frame %d size=%d\n", avctx->frame_number, buf_size);
+#endif
+
+    /* no supplementary picture */
+    if (buf_size == 0) {
+        *data_size = 0;
+        return 0;
+    }
+
+    init_get_bits(&s->gb, buf, buf_size);
+
+    mb_count = rv10_decode_picture_header(s);
+    if (mb_count < 0) {
+#ifdef DEBUG
+        printf("HEADER ERROR\n");
+#endif
+        return -1;
+    }
+    
+    if (s->mb_x >= s->mb_width ||
+        s->mb_y >= s->mb_height) {
+#ifdef DEBUG
+        printf("POS ERROR %d %d\n", s->mb_x, s->mb_y);
+#endif
+        return -1;
+    }
+    mb_pos = s->mb_y * s->mb_width + s->mb_x;
+    left = s->mb_width * s->mb_height - mb_pos;
+    if (mb_count > left) {
+#ifdef DEBUG
+        printf("COUNT ERROR\n");
+#endif
+        return -1;
+    }
+
+    if (s->mb_x == 0 && s->mb_y == 0) {
+        MPV_frame_start(s);
+    }
+
+#ifdef DEBUG
+    printf("qscale=%d\n", s->qscale);
+#endif
+
+    /* default quantization values */
+    s->y_dc_scale = 8;
+    s->c_dc_scale = 8;
+    s->rv10_first_dc_coded[0] = 0;
+    s->rv10_first_dc_coded[1] = 0;
+    s->rv10_first_dc_coded[2] = 0;
+    
+    /* decode each macroblock */
+    for(i=0;i<mb_count;i++) {
+#ifdef DEBUG
+        printf("**mb x=%d y=%d\n", s->mb_x, s->mb_y);
+#endif
+        
+        memset(block, 0, sizeof(block));
+        s->mv_dir = MV_DIR_FORWARD;
+        s->mv_type = MV_TYPE_16X16; 
+        if (h263_decode_mb(s, block) < 0) {
+#ifdef DEBUG
+            printf("ERROR\n");
+#endif
+            return -1;
+        }
+        MPV_decode_mb(s, block);
+        if (++s->mb_x == s->mb_width) {
+            s->mb_x = 0;
+            s->mb_y++;
+        }
+    }
+
+    if (s->mb_x == 0 &&
+        s->mb_y == s->mb_height) {
+        MPV_frame_end(s);
+        
+        pict->data[0] = s->current_picture[0];
+        pict->data[1] = s->current_picture[1];
+        pict->data[2] = s->current_picture[2];
+        pict->linesize[0] = s->linesize;
+        pict->linesize[1] = s->linesize / 2;
+        pict->linesize[2] = s->linesize / 2;
+        
+        avctx->quality = s->qscale;
+        *data_size = sizeof(AVPicture);
+    } else {
+        *data_size = 0;
+    }
+    return buf_size;
+}
+
+AVCodec rv10_decoder = {
+    "rv10",
+    CODEC_TYPE_VIDEO,
+    CODEC_ID_RV10,
+    sizeof(MpegEncContext),
+    rv10_decode_init,
+    NULL,
+    rv10_decode_end,
+    rv10_decode_frame,
+};
diff --git a/libavcodec/utils.c b/libavcodec/utils.c
new file mode 100644
index 0000000000..9497f90117
--- /dev/null
+++ b/libavcodec/utils.c
@@ -0,0 +1,306 @@
+/*
+ * utils for libavcodec
+ * Copyright (c) 2001 Gerard Lantau.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <errno.h>
+#include "common.h"
+#include "dsputil.h"
+#include "avcodec.h"
+
+/* memory alloc */
+void *av_mallocz(int size)
+{
+    void *ptr;
+    ptr = malloc(size);
+    if (!ptr)
+        return NULL;
+    memset(ptr, 0, size);
+    return ptr;
+}
+
+/* encoder management */
+AVCodec *first_avcodec;
+
+void register_avcodec(AVCodec *format)
+{
+    AVCodec **p;
+    p = &first_avcodec;
+    while (*p != NULL) p = &(*p)->next;
+    *p = format;
+    format->next = NULL;
+}
+
+int avcodec_open(AVCodecContext *avctx, AVCodec *codec)
+{
+    int ret;
+
+    avctx->codec = codec;
+    avctx->frame_number = 0;
+    avctx->priv_data = av_mallocz(codec->priv_data_size);
+    if (!avctx->priv_data) 
+        return -ENOMEM;
+    ret = avctx->codec->init(avctx);
+    if (ret < 0) {
+        free(avctx->priv_data);
+        avctx->priv_data = NULL;
+        return ret;
+    }
+    return 0;
+}
+
+int avcodec_encode_audio(AVCodecContext *avctx, UINT8 *buf, int buf_size, 
+                         const short *samples)
+{
+    int ret;
+
+    ret = avctx->codec->encode(avctx, buf, buf_size, (void *)samples);
+    avctx->frame_number++;
+    return ret;
+}
+
+int avcodec_encode_video(AVCodecContext *avctx, UINT8 *buf, int buf_size, 
+                         const AVPicture *pict)
+{
+    int ret;
+
+    ret = avctx->codec->encode(avctx, buf, buf_size, (void *)pict);
+    avctx->frame_number++;
+    return ret;
+}
+
+/* decode a frame. return -1 if error, otherwise return the number of
+   bytes used. If no frame could be decompressed, *got_picture_ptr is
+   zero. Otherwise, it is non zero */
+int avcodec_decode_video(AVCodecContext *avctx, AVPicture *picture, 
+                         int *got_picture_ptr,
+                         UINT8 *buf, int buf_size)
+{
+    int ret;
+
+    ret = avctx->codec->decode(avctx, picture, got_picture_ptr, 
+                               buf, buf_size);
+    avctx->frame_number++;
+    return ret;
+}
+
+/* decode an audio frame. return -1 if error, otherwise return the
+   *number of bytes used. If no frame could be decompressed,
+   *frame_size_ptr is zero. Otherwise, it is the decompressed frame
+   *size in BYTES. */
+int avcodec_decode_audio(AVCodecContext *avctx, INT16 *samples, 
+                         int *frame_size_ptr,
+                         UINT8 *buf, int buf_size)
+{
+    int ret;
+
+    ret = avctx->codec->decode(avctx, samples, frame_size_ptr, 
+                               buf, buf_size);
+    avctx->frame_number++;
+    return ret;
+}
+
+int avcodec_close(AVCodecContext *avctx)
+{
+    if (avctx->codec->close)
+        avctx->codec->close(avctx);
+    free(avctx->priv_data);
+    avctx->priv_data = NULL;
+    avctx->codec = NULL;
+    return 0;
+}
+
+AVCodec *avcodec_find_encoder(enum CodecID id)
+{
+    AVCodec *p;
+    p = first_avcodec;
+    while (p) {
+        if (p->encode != NULL && p->id == id)
+            return p;
+        p = p->next;
+    }
+    return NULL;
+}
+
+AVCodec *avcodec_find_decoder(enum CodecID id)
+{
+    AVCodec *p;
+    p = first_avcodec;
+    while (p) {
+        if (p->decode != NULL && p->id == id)
+            return p;
+        p = p->next;
+    }
+    return NULL;
+}
+
+AVCodec *avcodec_find_decoder_by_name(const char *name)
+{
+    AVCodec *p;
+    p = first_avcodec;
+    while (p) {
+        if (p->decode != NULL && strcmp(name,p->name) == 0)
+            return p;
+        p = p->next;
+    }
+    return NULL;
+}
+
+AVCodec *avcodec_find(enum CodecID id)
+{
+    AVCodec *p;
+    p = first_avcodec;
+    while (p) {
+        if (p->id == id)
+            return p;
+        p = p->next;
+    }
+    return NULL;
+}
+
+void avcodec_string(char *buf, int buf_size, AVCodecContext *enc, int encode)
+{
+    const char *codec_name;
+    AVCodec *p;
+    char buf1[32];
+
+    if (encode)
+        p = avcodec_find_encoder(enc->codec_id);
+    else
+        p = avcodec_find_decoder(enc->codec_id);
+
+    if (p) {
+        codec_name = p->name;
+    } else if (enc->codec_name[0] != '\0') {
+        codec_name = enc->codec_name;
+    } else {
+        /* output avi tags */
+        if (enc->codec_type == CODEC_TYPE_VIDEO) {
+            snprintf(buf1, sizeof(buf1), "%c%c%c%c", 
+                     enc->codec_tag & 0xff,
+                     (enc->codec_tag >> 8) & 0xff,
+                     (enc->codec_tag >> 16) & 0xff,
+                     (enc->codec_tag >> 24) & 0xff);
+        } else {
+            snprintf(buf1, sizeof(buf1), "0x%04x", enc->codec_tag);
+        }
+        codec_name = buf1;
+    }
+
+    switch(enc->codec_type) {
+    case CODEC_TYPE_VIDEO:
+        snprintf(buf, buf_size,
+                 "Video: %s%s",
+                 codec_name, enc->flags & CODEC_FLAG_HQ ? " (hq)" : "");
+        if (enc->width) {
+            snprintf(buf + strlen(buf), buf_size - strlen(buf),
+                     ", %dx%d, %0.2f fps",
+                     enc->width, enc->height, 
+                     (float)enc->frame_rate / FRAME_RATE_BASE);
+        }
+        break;
+    case CODEC_TYPE_AUDIO:
+        snprintf(buf, buf_size,
+                 "Audio: %s",
+                 codec_name);
+        if (enc->sample_rate) {
+            snprintf(buf + strlen(buf), buf_size - strlen(buf),
+                     ", %d Hz, %s",
+                     enc->sample_rate,
+                     enc->channels == 2 ? "stereo" : "mono");
+        }
+        break;
+    default:
+        abort();
+    }
+    if (enc->bit_rate != 0) {
+        snprintf(buf + strlen(buf), buf_size - strlen(buf), 
+                 ", %d kb/s", enc->bit_rate / 1000);
+    }
+}
+
+/* must be called before any other functions */
+void avcodec_init(void)
+{
+    dsputil_init();
+}
+
+/* simple call to use all the codecs */
+void avcodec_register_all(void)
+{
+    register_avcodec(&ac3_encoder);
+    register_avcodec(&mp2_encoder);
+    register_avcodec(&mpeg1video_encoder);
+    register_avcodec(&h263_encoder);
+    register_avcodec(&h263p_encoder);
+    register_avcodec(&rv10_encoder);
+    register_avcodec(&mjpeg_encoder);
+    register_avcodec(&opendivx_encoder);
+    register_avcodec(&msmpeg4_encoder);
+    register_avcodec(&pcm_codec);
+    register_avcodec(&rawvideo_codec);
+    /* decoders */
+    register_avcodec(&h263_decoder);
+    register_avcodec(&opendivx_decoder);
+    register_avcodec(&msmpeg4_decoder);
+    register_avcodec(&mpeg_decoder);
+    register_avcodec(&h263i_decoder);
+    register_avcodec(&rv10_decoder);
+}
+
+static int encode_init(AVCodecContext *s)
+{
+    return 0;
+}
+
+static int decode_frame(AVCodecContext *avctx, 
+                        void *data, int *data_size,
+                        UINT8 *buf, int buf_size)
+{
+    return -1;
+}
+
+static int encode_frame(AVCodecContext *avctx,
+                        unsigned char *frame, int buf_size, void *data)
+{
+    return -1;
+}
+
+/* dummy pcm codec */
+AVCodec pcm_codec = {
+    "pcm",
+    CODEC_TYPE_AUDIO,
+    CODEC_ID_PCM,
+    0,
+    encode_init,
+    encode_frame,
+    NULL,
+    decode_frame,
+};
+
+AVCodec rawvideo_codec = {
+    "rawvideo",
+    CODEC_TYPE_VIDEO,
+    CODEC_ID_RAWVIDEO,
+    0,
+    encode_init,
+    encode_frame,
+    NULL,
+    decode_frame,
+};