Merge remote-tracking branch 'qatar/master'

* qatar/master:
  fate: Add tests for more AAC features.
  aacps: Add missing newline in error message.
  fate: Add tests for vc1/wmapro in ism.
  aacdec: Add a fate test for 5.1 channel SBR.
  aacdec: Turn off PS for multichannel files that use PCE based configs.
  cabac: remove put_cabac_u/ueg from cabac-test.
  swscale: RGB4444 and BGR444 input
  FATE: add test for xWMA demuxer.
  FATE: add test for SMJPEG demuxer and associated IMA ADPCM audio decoder.
  mpegaudiodec: optimized iMDCT transform
  mpegaudiodec: change imdct window arrangment for better pointer alignment
  mpegaudiodec: move imdct and windowing function to mpegaudiodsp
  mpegaudiodec: interleave iMDCT buffer to simplify future SIMD implementations
  swscale: convert yuy2/uyvy/nv12/nv21ToY/UV from inline asm to yasm.
  FATE: test to exercise WTV demuxer.
  mjpegdec: K&R formatting cosmetics
  swscale: K&R formatting cosmetics for code examples
  swscale: K&R reformatting cosmetics for header files
  FATE test: cvid-grayscale; ensures that the grayscale Cinepak variant is exercised.

Conflicts:
	libavcodec/cabac.c
	libavcodec/mjpegdec.c
	libavcodec/mpegaudiodec.c
	libavcodec/mpegaudiodsp.c
	libavcodec/mpegaudiodsp.h
	libavcodec/mpegaudiodsp_template.c
	libavcodec/x86/Makefile
	libavcodec/x86/imdct36_sse.asm
	libavcodec/x86/mpegaudiodec_mmx.c
	libswscale/swscale-test.c
	libswscale/swscale.c
	libswscale/swscale_internal.h
	libswscale/x86/swscale_template.c
	tests/fate/demux.mak
	tests/fate/microsoft.mak
	tests/fate/video.mak
	tests/fate/wma.mak
	tests/ref/lavfi/pixfmts_scale

Merged-by: Michael Niedermayer <michaelni@gmx.at>

38 files changed, 2217 insertions, 723 deletions

diff --git a/libavcodec/aac.h b/libavcodec/aac.h
index 631fdd405f..6fdeedea21 100644
--- a/libavcodec/aac.h
+++ b/libavcodec/aac.h
@@ -84,6 +84,7 @@ enum BandType {
 #define IS_CODEBOOK_UNSIGNED(x) ((x - 1) & 10)
 
 enum ChannelPosition {
+    AAC_CHANNEL_OFF   = 0,
     AAC_CHANNEL_FRONT = 1,
     AAC_CHANNEL_SIDE  = 2,
     AAC_CHANNEL_BACK  = 3,
diff --git a/libavcodec/aacdec.c b/libavcodec/aacdec.c
index c0798429af..b7f76a61da 100644
--- a/libavcodec/aacdec.c
+++ b/libavcodec/aacdec.c
@@ -163,6 +163,19 @@ static ChannelElement *get_che(AACContext *ac, int type, int elem_id)
     }
 }
 
+static int count_channels(enum ChannelPosition che_pos[4][MAX_ELEM_ID])
+{
+    int i, type, sum = 0;
+    for (i = 0; i < MAX_ELEM_ID; i++) {
+        for (type = 0; type < 4; type++) {
+            sum += (1 + (type == TYPE_CPE)) *
+                (che_pos[type][i] != AAC_CHANNEL_OFF &&
+                 che_pos[type][i] != AAC_CHANNEL_CC);
+        }
+    }
+    return sum;
+}
+
 /**
  * Check for the channel element in the current channel position configuration.
  * If it exists, make sure the appropriate element is allocated and map the
@@ -437,6 +450,12 @@ static int decode_ga_specific_config(AACContext *ac, AVCodecContext *avctx,
         if ((ret = set_default_channel_config(avctx, new_che_pos, channel_config)))
             return ret;
     }
+
+    if (count_channels(new_che_pos) > 1) {
+        m4ac->ps = 0;
+    } else if (m4ac->sbr == 1 && m4ac->ps == -1)
+        m4ac->ps = 1;
+
     if (ac && (ret = output_configure(ac, ac->che_pos, new_che_pos, channel_config, OC_GLOBAL_HDR)))
         return ret;
 
@@ -495,8 +514,6 @@ static int decode_audio_specific_config(AACContext *ac,
         av_log(avctx, AV_LOG_ERROR, "invalid sampling rate index %d\n", m4ac->sampling_index);
         return -1;
     }
-    if (m4ac->sbr == 1 && m4ac->ps == -1)
-        m4ac->ps = 1;
 
     skip_bits_long(&gb, i);
 
diff --git a/libavcodec/aacps.c b/libavcodec/aacps.c
index 6ef8347ee5..3b5aa58109 100644
--- a/libavcodec/aacps.c
+++ b/libavcodec/aacps.c
@@ -223,7 +223,7 @@ int ff_ps_read_data(AVCodecContext *avctx, GetBitContext *gb_host, PSContext *ps
             cnt -= 2 + ps_read_extension_data(gb, ps, ps_extension_id);
         }
         if (cnt < 0) {
-            av_log(avctx, AV_LOG_ERROR, "ps extension overflow %d", cnt);
+            av_log(avctx, AV_LOG_ERROR, "ps extension overflow %d\n", cnt);
             goto err;
         }
         skip_bits(gb, cnt);
diff --git a/libavcodec/cabac.c b/libavcodec/cabac.c
index 983614581d..6cabfb9e73 100644
--- a/libavcodec/cabac.c
+++ b/libavcodec/cabac.c
@@ -252,67 +252,6 @@ static int put_cabac_terminate(CABACContext *c, int bit){
     return (put_bits_count(&c->pb)+7)>>3;
 }
 
-/**
- * put (truncated) unary binarization.
- */
-static void put_cabac_u(CABACContext *c, uint8_t * state, int v, int max, int max_index, int truncated){
-    int i;
-
-    assert(v <= max);
-
-    for(i=0; i<v; i++){
-        put_cabac(c, state, 1);
-        if(i < max_index) state++;
-    }
-    if(truncated==0 || v<max)
-        put_cabac(c, state, 0);
-}
-
-/**
- * put unary exp golomb k-th order binarization.
- */
-static void put_cabac_ueg(CABACContext *c, uint8_t * state, int v, int max, int is_signed, int k, int max_index){
-    int i;
-
-    if(v==0)
-        put_cabac(c, state, 0);
-    else{
-        const int sign= v < 0;
-
-        if(is_signed) v= FFABS(v);
-
-        if(v<max){
-            for(i=0; i<v; i++){
-                put_cabac(c, state, 1);
-                if(i < max_index) state++;
-            }
-
-            put_cabac(c, state, 0);
-        }else{
-            int m= 1<<k;
-
-            for(i=0; i<max; i++){
-                put_cabac(c, state, 1);
-                if(i < max_index) state++;
-            }
-
-            v -= max;
-            while(v >= m){ //FIXME optimize
-                put_cabac_bypass(c, 1);
-                v-= m;
-                m+= m;
-            }
-            put_cabac_bypass(c, 0);
-            while(m>>=1){
-                put_cabac_bypass(c, v&m);
-            }
-        }
-
-        if(is_signed)
-            put_cabac_bypass(c, sign);
-    }
-}
-
 int main(void){
     CABACContext c;
     uint8_t b[9*SIZE];
@@ -342,19 +281,6 @@ START_TIMER
 STOP_TIMER("put_cabac")
     }
 
-#if 0
-    for(i=0; i<SIZE; i++){
-START_TIMER
-        put_cabac_u(&c, state, r[i], 6, 3, i&1);
-STOP_TIMER("put_cabac_u")
-    }
-
-    for(i=0; i<SIZE; i++){
-START_TIMER
-        put_cabac_ueg(&c, state, r[i], 3, 0, 1, 2);
-STOP_TIMER("put_cabac_ueg")
-    }
-#endif
     put_cabac_terminate(&c, 1);
 
     ff_init_cabac_decoder(&c, b, SIZE);
diff --git a/libavcodec/mpegaudiodec.c b/libavcodec/mpegaudiodec.c
index eae030ce44..18ef40e011 100644
--- a/libavcodec/mpegaudiodec.c
+++ b/libavcodec/mpegaudiodec.c
@@ -132,10 +132,6 @@ static uint16_t band_index_long[9][23];
 static INTFLOAT is_table[2][16];
 static INTFLOAT is_table_lsf[2][2][16];
 static INTFLOAT csa_table[8][4];
-/** Window for MDCT. Note that only the component [0,17] and [20,37] are used,
-    the components 18 and 19 are there only to assure 128-bit alignment for asm
- */
-DECLARE_ALIGNED(16, static INTFLOAT, mdct_win)[8][40];
 
 static int16_t division_tab3[1<<6 ];
 static int16_t division_tab5[1<<8 ];
@@ -422,45 +418,6 @@ static av_cold void decode_init_static(void)
         csa_table[i][3] = ca - cs;
 #endif
     }
-
-    /* compute mdct windows */
-    for (i = 0; i < 36; i++) {
-        for (j = 0; j < 4; j++) {
-            double d;
-
-            if (j == 2 && i % 3 != 1)
-                continue;
-
-            d = sin(M_PI * (i + 0.5) / 36.0);
-            if (j == 1) {
-                if      (i >= 30) d = 0;
-                else if (i >= 24) d = sin(M_PI * (i - 18 + 0.5) / 12.0);
-                else if (i >= 18) d = 1;
-            } else if (j == 3) {
-                if      (i <   6) d = 0;
-                else if (i <  12) d = sin(M_PI * (i -  6 + 0.5) / 12.0);
-                else if (i <  18) d = 1;
-            }
-            //merge last stage of imdct into the window coefficients
-            d *= 0.5 / cos(M_PI * (2 * i + 19) / 72);
-
-            if (j == 2)
-                mdct_win[j][i/3] = FIXHR((d / (1<<5)));
-            else {
-                int idx = i < 18 ? i : i + 2;
-                mdct_win[j][idx] = FIXHR((d / (1<<5)));
-            }
-        }
-    }
-
-    /* NOTE: we do frequency inversion adter the MDCT by changing
-        the sign of the right window coefs */
-    for (j = 0; j < 4; j++) {
-        for (i = 0; i < 40; i += 2) {
-            mdct_win[j + 4][i    ] =  mdct_win[j][i    ];
-            mdct_win[j + 4][i + 1] = -mdct_win[j][i + 1];
-        }
-    }
 }
 
 static av_cold int decode_init(AVCodecContext * avctx)
@@ -1284,59 +1241,53 @@ static void compute_imdct(MPADecodeContext *s, GranuleDef *g,
         mdct_long_end = sblimit;
     }
 
-    buf = mdct_buf;
-    ptr = g->sb_hybrid;
-    for (j = 0; j < mdct_long_end; j++) {
-        int win_idx = (g->switch_point && j < 2) ? 0 : g->block_type;
-        /* apply window & overlap with previous buffer */
-        out_ptr = sb_samples + j;
-        /* select window */
-        win = mdct_win[win_idx + (4 & -(j & 1))];
-        s->mpadsp.RENAME(imdct36)(out_ptr, buf, ptr, win);
-        out_ptr += 18 * SBLIMIT;
-        ptr     += 18;
-        buf     += 18;
-    }
+    s->mpadsp.RENAME(imdct36_blocks)(sb_samples, mdct_buf, g->sb_hybrid,
+                                     mdct_long_end, g->switch_point,
+                                     g->block_type);
+
+    buf = mdct_buf + 4*18*(mdct_long_end >> 2) + (mdct_long_end & 3);
+    ptr = g->sb_hybrid + 18 * mdct_long_end;
+
     for (j = mdct_long_end; j < sblimit; j++) {
         /* select frequency inversion */
-        win     = mdct_win[2 + (4  & -(j & 1))];
+        win     = RENAME(ff_mdct_win)[2 + (4  & -(j & 1))];
         out_ptr = sb_samples + j;
 
         for (i = 0; i < 6; i++) {
-            *out_ptr = buf[i];
+            *out_ptr = buf[4*i];
             out_ptr += SBLIMIT;
         }
         imdct12(out2, ptr + 0);
         for (i = 0; i < 6; i++) {
-            *out_ptr     = MULH3(out2[i    ], win[i    ], 1) + buf[i + 6*1];
-            buf[i + 6*2] = MULH3(out2[i + 6], win[i + 6], 1);
+            *out_ptr     = MULH3(out2[i    ], win[i    ], 1) + buf[4*(i + 6*1)];
+            buf[4*(i + 6*2)] = MULH3(out2[i + 6], win[i + 6], 1);
             out_ptr += SBLIMIT;
         }
         imdct12(out2, ptr + 1);
         for (i = 0; i < 6; i++) {
-            *out_ptr     = MULH3(out2[i    ], win[i    ], 1) + buf[i + 6*2];
-            buf[i + 6*0] = MULH3(out2[i + 6], win[i + 6], 1);
+            *out_ptr     = MULH3(out2[i    ], win[i    ], 1) + buf[4*(i + 6*2)];
+            buf[4*(i + 6*0)] = MULH3(out2[i + 6], win[i + 6], 1);
             out_ptr += SBLIMIT;
         }
         imdct12(out2, ptr + 2);
         for (i = 0; i < 6; i++) {
-            buf[i + 6*0] = MULH3(out2[i    ], win[i    ], 1) + buf[i + 6*0];
-            buf[i + 6*1] = MULH3(out2[i + 6], win[i + 6], 1);
-            buf[i + 6*2] = 0;
+            buf[4*(i + 6*0)] = MULH3(out2[i    ], win[i    ], 1) + buf[4*(i + 6*0)];
+            buf[4*(i + 6*1)] = MULH3(out2[i + 6], win[i + 6], 1);
+            buf[4*(i + 6*2)] = 0;
         }
         ptr += 18;
-        buf += 18;
+        buf += (j&3) != 3 ? 1 : (4*18-3);
     }
     /* zero bands */
     for (j = sblimit; j < SBLIMIT; j++) {
         /* overlap */
         out_ptr = sb_samples + j;
         for (i = 0; i < 18; i++) {
-            *out_ptr = buf[i];
-            buf[i]   = 0;
+            *out_ptr = buf[4*i];
+            buf[4*i]   = 0;
             out_ptr += SBLIMIT;
         }
-        buf += 18;
+        buf += (j&3) != 3 ? 1 : (4*18-3);
     }
 }
 
diff --git a/libavcodec/mpegaudiodsp.c b/libavcodec/mpegaudiodsp.c
index 212610d887..cc12dd9cee 100644
--- a/libavcodec/mpegaudiodsp.c
+++ b/libavcodec/mpegaudiodsp.c
@@ -28,6 +28,8 @@ void ff_mpadsp_init(MPADSPContext *s)
     DCTContext dct;
 
     ff_dct_init(&dct, 5, DCT_II);
+    ff_init_mpadsp_tabs_float();
+    ff_init_mpadsp_tabs_fixed();
 
     s->apply_window_float = ff_mpadsp_apply_window_float;
     s->apply_window_fixed = ff_mpadsp_apply_window_fixed;
@@ -35,8 +37,8 @@ void ff_mpadsp_init(MPADSPContext *s)
     s->dct32_float = dct.dct32;
     s->dct32_fixed = ff_dct32_fixed;
 
-    s->imdct36_float = ff_imdct36_float;
-    s->imdct36_fixed = ff_imdct36_fixed;
+    s->imdct36_blocks_float = ff_imdct36_blocks_float;
+    s->imdct36_blocks_fixed = ff_imdct36_blocks_fixed;
 
     if (ARCH_ARM)     ff_mpadsp_init_arm(s);
     if (HAVE_MMX)     ff_mpadsp_init_mmx(s);
diff --git a/libavcodec/mpegaudiodsp.h b/libavcodec/mpegaudiodsp.h
index 9e84784403..da53fe56e8 100644
--- a/libavcodec/mpegaudiodsp.h
+++ b/libavcodec/mpegaudiodsp.h
@@ -20,6 +20,7 @@
 #define AVCODEC_MPEGAUDIODSP_H
 
 #include <stdint.h>
+#include "libavutil/common.h"
 
 typedef struct MPADSPContext {
     void (*apply_window_float)(float *synth_buf, float *window,
@@ -28,8 +29,11 @@ typedef struct MPADSPContext {
                                int *dither_state, int16_t *samples, int incr);
     void (*dct32_float)(float *dst, const float *src);
     void (*dct32_fixed)(int *dst, const int *src);
-    void (*imdct36_float)(float *out, float *buf, float *in, float *win);
-    void (*imdct36_fixed)(int *out, int *buf, int *in, int *win);
+
+    void (*imdct36_blocks_float)(float *out, float *buf, float *in,
+                                 int count, int switch_point, int block_type);
+    void (*imdct36_blocks_fixed)(int *out, int *buf, int *in,
+                                 int count, int switch_point, int block_type);
 } MPADSPContext;
 
 void ff_mpadsp_init(MPADSPContext *s);
@@ -63,7 +67,19 @@ void ff_mpadsp_apply_window_fixed(int32_t *synth_buf, int32_t *window,
                                   int *dither_state, int16_t *samples,
                                   int incr);
 
-void ff_imdct36_fixed(int *out, int *buf, int *in, int *win);
-void ff_imdct36_float(float *out, float *buf, float *in, float *win);
+void ff_imdct36_blocks_float(float *out, float *buf, float *in,
+                             int count, int switch_point, int block_type);
+
+void ff_imdct36_blocks_fixed(int *out, int *buf, int *in,
+                             int count, int switch_point, int block_type);
+
+void ff_init_mpadsp_tabs_float(void);
+void ff_init_mpadsp_tabs_fixed(void);
+
+/** For SSE implementation, MDCT_BUF_SIZE/2 should be 128-bit aligned */
+#define MDCT_BUF_SIZE FFALIGN(36, 2*4)
+
+extern int ff_mdct_win_fixed[8][MDCT_BUF_SIZE];
+extern float ff_mdct_win_float[8][MDCT_BUF_SIZE];
 
 #endif /* AVCODEC_MPEGAUDIODSP_H */
diff --git a/libavcodec/mpegaudiodsp_template.c b/libavcodec/mpegaudiodsp_template.c
index 40eca76b81..53b6139da3 100644
--- a/libavcodec/mpegaudiodsp_template.c
+++ b/libavcodec/mpegaudiodsp_template.c
@@ -69,6 +69,12 @@ static inline int round_sample(int64_t *sum)
 #   define FIXHR(a)       ((int)((a) * (1LL<<32) + 0.5))
 #endif
 
+/** Window for MDCT. Actually only the elements in [0,17] and
+    [MDCT_BUF_SIZE/2, MDCT_BUF_SIZE/2 + 17] are actually used. The rest
+    is just to preserve alignment for SIMD implementations.
+*/
+DECLARE_ALIGNED(16, INTFLOAT, RENAME(ff_mdct_win))[8][MDCT_BUF_SIZE];
+
 DECLARE_ALIGNED(16, MPA_INT, RENAME(ff_mpa_synth_window))[512+256];
 
 #define SUM8(op, sum, w, p)               \
@@ -204,6 +210,7 @@ void av_cold RENAME(ff_mpa_synth_init)(MPA_INT *window)
             window[512 - i] = v;
     }
 
+
     // Needed for avoiding shuffles in ASM implementations
     for(i=0; i < 8; i++)
         for(j=0; j < 16; j++)
@@ -214,6 +221,48 @@ void av_cold RENAME(ff_mpa_synth_init)(MPA_INT *window)
             window[512+128+16*i+j] = window[64*i+48-j];
 }
 
+void RENAME(ff_init_mpadsp_tabs)(void)
+{
+    int i, j;
+    /* compute mdct windows */
+    for (i = 0; i < 36; i++) {
+        for (j = 0; j < 4; j++) {
+            double d;
+
+            if (j == 2 && i % 3 != 1)
+                continue;
+
+            d = sin(M_PI * (i + 0.5) / 36.0);
+            if (j == 1) {
+                if      (i >= 30) d = 0;
+                else if (i >= 24) d = sin(M_PI * (i - 18 + 0.5) / 12.0);
+                else if (i >= 18) d = 1;
+            } else if (j == 3) {
+                if      (i <   6) d = 0;
+                else if (i <  12) d = sin(M_PI * (i -  6 + 0.5) / 12.0);
+                else if (i <  18) d = 1;
+            }
+            //merge last stage of imdct into the window coefficients
+            d *= 0.5 / cos(M_PI * (2 * i + 19) / 72);
+
+            if (j == 2)
+                RENAME(ff_mdct_win)[j][i/3] = FIXHR((d / (1<<5)));
+            else {
+                int idx = i < 18 ? i : i + (MDCT_BUF_SIZE/2 - 18);
+                RENAME(ff_mdct_win)[j][idx] = FIXHR((d / (1<<5)));
+            }
+        }
+    }
+
+    /* NOTE: we do frequency inversion adter the MDCT by changing
+        the sign of the right window coefs */
+    for (j = 0; j < 4; j++) {
+        for (i = 0; i < MDCT_BUF_SIZE; i += 2) {
+            RENAME(ff_mdct_win)[j + 4][i    ] =  RENAME(ff_mdct_win)[j][i    ];
+            RENAME(ff_mdct_win)[j + 4][i + 1] = -RENAME(ff_mdct_win)[j][i + 1];
+        }
+    }
+}
 /* cos(pi*i/18) */
 #define C1 FIXHR(0.98480775301220805936/2)
 #define C2 FIXHR(0.93969262078590838405/2)
@@ -227,43 +276,42 @@ void av_cold RENAME(ff_mpa_synth_init)(MPA_INT *window)
 /* 0.5 / cos(pi*(2*i+1)/36) */
 static const INTFLOAT icos36[9] = {
     FIXR(0.50190991877167369479),
-    FIXR(0.51763809020504152469),
+    FIXR(0.51763809020504152469), //0
     FIXR(0.55168895948124587824),
     FIXR(0.61038729438072803416),
-    FIXR(0.70710678118654752439),
+    FIXR(0.70710678118654752439), //1
     FIXR(0.87172339781054900991),
     FIXR(1.18310079157624925896),
-    FIXR(1.93185165257813657349),
+    FIXR(1.93185165257813657349), //2
     FIXR(5.73685662283492756461),
 };
 
 /* 0.5 / cos(pi*(2*i+1)/36) */
 static const INTFLOAT icos36h[9] = {
     FIXHR(0.50190991877167369479/2),
-    FIXHR(0.51763809020504152469/2),
+    FIXHR(0.51763809020504152469/2), //0
     FIXHR(0.55168895948124587824/2),
     FIXHR(0.61038729438072803416/2),
-    FIXHR(0.70710678118654752439/2),
+    FIXHR(0.70710678118654752439/2), //1
     FIXHR(0.87172339781054900991/2),
     FIXHR(1.18310079157624925896/4),
-    FIXHR(1.93185165257813657349/4),
+    FIXHR(1.93185165257813657349/4), //2
+//    FIXHR(5.73685662283492756461),
 };
 
-
 /* using Lee like decomposition followed by hand coded 9 points DCT */
-void RENAME(ff_imdct36)(INTFLOAT *out, INTFLOAT *buf, INTFLOAT *in,
-                        INTFLOAT *win)
+static void imdct36(INTFLOAT *out, INTFLOAT *buf, INTFLOAT *in, INTFLOAT *win)
 {
     int i, j;
     INTFLOAT t0, t1, t2, t3, s0, s1, s2, s3;
     INTFLOAT tmp[18], *tmp1, *in1;
 
-    for(i=17;i>=1;i--)
+    for (i = 17; i >= 1; i--)
         in[i] += in[i-1];
-    for(i=17;i>=3;i-=2)
+    for (i = 17; i >= 3; i -= 2)
         in[i] += in[i-2];
 
-    for(j=0;j<2;j++) {
+    for (j = 0; j < 2; j++) {
         tmp1 = tmp + j;
         in1 = in + j;
 
@@ -295,7 +343,7 @@ void RENAME(ff_imdct36)(INTFLOAT *out, INTFLOAT *buf, INTFLOAT *in,
     }
 
     i = 0;
-    for(j=0;j<4;j++) {
+    for (j = 0; j < 4; j++) {
         t0 = tmp[i];
         t1 = tmp[i + 2];
         s0 = t1 + t0;
@@ -303,22 +351,22 @@ void RENAME(ff_imdct36)(INTFLOAT *out, INTFLOAT *buf, INTFLOAT *in,
 
         t2 = tmp[i + 1];
         t3 = tmp[i + 3];
-        s1 = MULH3(t3 + t2, icos36h[j], 2);
-        s3 = MULLx(t3 - t2, icos36[8 - j], FRAC_BITS);
+        s1 = MULH3(t3 + t2, icos36h[    j], 2);
+        s3 = MULLx(t3 - t2, icos36 [8 - j], FRAC_BITS);
 
         t0 = s0 + s1;
         t1 = s0 - s1;
-        out[(9 + j)*SBLIMIT] =  MULH3(t1, win[9 + j], 1) + buf[9 + j];
-        out[(8 - j)*SBLIMIT] =  MULH3(t1, win[8 - j], 1) + buf[8 - j];
-        buf[9 + j] = MULH3(t0, win[20 + 9 + j], 1);
-        buf[8 - j] = MULH3(t0, win[20 + 8 - j], 1);
+        out[(9 + j) * SBLIMIT] = MULH3(t1, win[     9 + j], 1) + buf[4*(9 + j)];
+        out[(8 - j) * SBLIMIT] = MULH3(t1, win[     8 - j], 1) + buf[4*(8 - j)];
+        buf[4 * ( 9 + j     )] = MULH3(t0, win[MDCT_BUF_SIZE/2 + 9 + j], 1);
+        buf[4 * ( 8 - j     )] = MULH3(t0, win[MDCT_BUF_SIZE/2 + 8 - j], 1);
 
         t0 = s2 + s3;
         t1 = s2 - s3;
-        out[(9 + 8 - j)*SBLIMIT] =  MULH3(t1, win[9 + 8 - j], 1) + buf[9 + 8 - j];
-        out[(        j)*SBLIMIT] =  MULH3(t1, win[        j], 1) + buf[        j];
-        buf[9 + 8 - j] = MULH3(t0, win[20 + 9 + 8 - j], 1);
-        buf[      + j] = MULH3(t0, win[20         + j], 1);
+        out[(9 + 8 - j) * SBLIMIT] = MULH3(t1, win[     9 + 8 - j], 1) + buf[4*(9 + 8 - j)];
+        out[         j  * SBLIMIT] = MULH3(t1, win[             j], 1) + buf[4*(        j)];
+        buf[4 * ( 9 + 8 - j     )] = MULH3(t0, win[MDCT_BUF_SIZE/2 + 9 + 8 - j], 1);
+        buf[4 * (         j     )] = MULH3(t0, win[MDCT_BUF_SIZE/2         + j], 1);
         i += 4;
     }
 
@@ -326,9 +374,28 @@ void RENAME(ff_imdct36)(INTFLOAT *out, INTFLOAT *buf, INTFLOAT *in,
     s1 = MULH3(tmp[17], icos36h[4], 2);
     t0 = s0 + s1;
     t1 = s0 - s1;
-    out[(9 + 4)*SBLIMIT] =  MULH3(t1, win[9 + 4], 1) + buf[9 + 4];
-    out[(8 - 4)*SBLIMIT] =  MULH3(t1, win[8 - 4], 1) + buf[8 - 4];
-    buf[9 + 4] = MULH3(t0, win[20 + 9 + 4], 1);
-    buf[8 - 4] = MULH3(t0, win[20 + 8 - 4], 1);
+    out[(9 + 4) * SBLIMIT] = MULH3(t1, win[     9 + 4], 1) + buf[4*(9 + 4)];
+    out[(8 - 4) * SBLIMIT] = MULH3(t1, win[     8 - 4], 1) + buf[4*(8 - 4)];
+    buf[4 * ( 9 + 4     )] = MULH3(t0, win[MDCT_BUF_SIZE/2 + 9 + 4], 1);
+    buf[4 * ( 8 - 4     )] = MULH3(t0, win[MDCT_BUF_SIZE/2 + 8 - 4], 1);
+}
+
+void RENAME(ff_imdct36_blocks)(INTFLOAT *out, INTFLOAT *buf, INTFLOAT *in,
+                               int count, int switch_point, int block_type)
+{
+    int j;
+    for (j=0 ; j < count; j++) {
+        /* apply window & overlap with previous buffer */
+
+        /* select window */
+        int win_idx = (switch_point && j < 2) ? 0 : block_type;
+        INTFLOAT *win = RENAME(ff_mdct_win)[win_idx + (4 & -(j & 1))];
+
+        imdct36(out, buf, in, win);
+
+        in  += 18;
+        buf += ((j&3) != 3 ? 1 : (72-3));
+        out++;
+    }
 }
 
diff --git a/libavcodec/x86/imdct36_sse.asm b/libavcodec/x86/imdct36_sse.asm
index 2b8fe57e0b..2908459db7 100644
--- a/libavcodec/x86/imdct36_sse.asm
+++ b/libavcodec/x86/imdct36_sse.asm
@@ -53,92 +53,118 @@ ps_cosh_sse3:  dd 1.0, -0.50190991877167369479,  1.0, -5.73685662283492756461
                dd 1.0, -0.61038729438072803416, -1.0,  0.87172339781054900991
                dd 1.0,  0.70710678118654752439,  0.0,  0.0
 
+costabs:  times 4 dd  0.98480773
+          times 4 dd  0.93969262
+          times 4 dd  0.86602539
+          times 4 dd -0.76604444
+          times 4 dd -0.64278764
+          times 4 dd  0.50000000
+          times 4 dd -0.50000000
+          times 4 dd -0.34202015
+          times 4 dd -0.17364818
+          times 4 dd  0.50190992
+          times 4 dd  0.51763808
+          times 4 dd  0.55168896
+          times 4 dd  0.61038726
+          times 4 dd  0.70710677
+          times 4 dd  0.87172341
+          times 4 dd  1.18310082
+          times 4 dd  1.93185163
+          times 4 dd  5.73685646
+
 %define SBLIMIT 32
 SECTION_TEXT
 
-%macro PSHUFD_SSE_AVX 3
-    shufps %1, %2, %2, %3
-%endmacro
-%macro PSHUFD_SSE2 3
+%macro PSHUFD 3
+%if cpuflag(sse2) && notcpuflag(avx)
     pshufd %1, %2, %3
+%else
+    shufps %1, %2, %2, %3
+%endif
 %endmacro
 
-; input  %1={x1,x2,x3,x4}, %2={y1,y2,y3,y4}
-; output %3={x3,x4,y1,y2}
-%macro BUILDINVHIGHLOW_SSE 3
-    movlhps %3, %2
-    movhlps %3, %1
-%endmacro
-%macro BUILDINVHIGHLOW_AVX 3
-    shufps %3, %1, %2, 0x4e
+; input  %2={x1,x2,x3,x4}, %3={y1,y2,y3,y4}
+; output %1={x3,x4,y1,y2}
+%macro BUILDINVHIGHLOW 3
+%if cpuflag(avx)
+    shufps %1, %2, %3, 0x4e
+%else
+    movlhps %1, %3
+    movhlps %1, %2
+%endif
 %endmacro
 
-; input  %1={x1,x2,x3,x4}, %2={y1,y2,y3,y4}
-; output %3={x4,y1,y2,y3}
-%macro ROTLEFT_SSE 3
+; input  %2={x1,x2,x3,x4}, %3={y1,y2,y3,y4}
+; output %1={x4,y1,y2,y3}
+%macro ROTLEFT 3
+%if cpuflag(ssse3)
+    palignr  %1, %3, %2, 12
+%else
     BUILDINVHIGHLOW %1, %2, %3
-    shufps  %3, %3, %2, 0x99
-%endmacro
-
-%macro ROTLEFT_SSSE3 3
-    palignr  %3, %2, %1, 12
+    shufps  %1, %1, %3, 0x99
+%endif
 %endmacro
 
-%macro INVERTHL_SSE1 2
+%macro INVERTHL 2
+%if cpuflag(sse2)
+    PSHUFD  %1, %2, 0x4e
+%else
     movhlps %1, %2
     movlhps %1, %2
+%endif
 %endmacro
 
-%macro INVERTHL_SSE2 2
-    PSHUFD  %1, %2, 0x4e
-%endmacro
-
-%macro BUTTERF_SSE12 3
+%macro BUTTERF 3
     INVERTHL %2, %1
     xorps    %1, [ps_p1p1m1m1]
     addps    %1, %2
+%if cpuflag(sse3)
+    mulps    %1, %1, [ps_cosh_sse3 + %3]
+    PSHUFD   %2, %1, 0xb1
+    addsubps %1, %1, %2
+%else
     mulps    %1, [ps_cosh + %3]
     PSHUFD   %2, %1, 0xb1
     xorps    %1, [ps_p1m1p1m1]
     addps    %1, %2
-%endmacro
-%macro BUTTERF_SSE3 3
-    INVERTHL %2, %1
-    xorps    %1, %1, [ps_p1p1m1m1]
-    addps    %1, %1, %2
-    mulps    %1, %1, [ps_cosh_sse3 + %3]
-    PSHUFD   %2, %1, 0xb1
-    addsubps %1, %1, %2
+%endif
 %endmacro
 
-%macro STORE 3
+%macro STORE 4
     movhlps %2, %1
-    movss   [%3             ], %1
-    movss   [%3 +  8*SBLIMIT], %2
+    movss   [%3       ], %1
+    movss   [%3 + 2*%4], %2
     shufps  %1, %1, 0xb1
-    movss   [%3 +  4*SBLIMIT], %1
+    movss   [%3 +   %4], %1
     movhlps %2, %1
-    movss   [%3 + 12*SBLIMIT], %2
+    movss   [%3 + 3*%4], %2
+%endmacro
+
+%macro LOAD 4
+    movlps  %1, [%3       ]
+    movhps  %1, [%3 +   %4]
+    movlps  %2, [%3 + 2*%4]
+    movhps  %2, [%3 + 3*%4]
+    shufps  %1, %2, 0x88
 %endmacro
 
 %macro LOADA64 2
+%if cpuflag(avx)
+   movu     %1, [%2]
+%else
    movlps   %1, [%2]
    movhps   %1, [%2 + 8]
+%endif
 %endmacro
 
-%macro STOREA64 2
-   movlps   [%1    ], %2
-   movhps   [%1 + 8], %2
-%endmacro
-
-%macro DEFINE_IMDCT 1
-cglobal imdct36_float_%1, 4,4,9, out, buf, in, win
+%macro DEFINE_IMDCT 0
+cglobal imdct36_float, 4,4,9, out, buf, in, win
 
     ; for(i=17;i>=1;i--) in[i] += in[i-1];
     LOADA64 m0, inq
     LOADA64 m1, inq + 16
 
-    ROTLEFT m0, m1, m5
+    ROTLEFT m5, m0, m1
 
     PSHUFD  m6, m0, 0x93
     andps   m6, m6, [ps_mask]
@@ -146,16 +172,16 @@ cglobal imdct36_float_%1, 4,4,9, out, buf, in, win
 
     LOADA64 m2, inq + 32
 
-    ROTLEFT m1, m2, m7
+    ROTLEFT m7, m1, m2
 
     addps   m1, m1, m5
     LOADA64 m3, inq + 48
 
-    ROTLEFT m2, m3, m5
+    ROTLEFT m5, m2, m3
 
     xorps   m4, m4, m4
     movlps  m4, [inq+64]
-    BUILDINVHIGHLOW m3, m4, m6
+    BUILDINVHIGHLOW m6, m3, m4
     shufps  m6, m6, m4, 0xa9
 
     addps   m4, m4, m6
@@ -166,17 +192,17 @@ cglobal imdct36_float_%1, 4,4,9, out, buf, in, win
     movlhps m5, m5, m0
     andps   m5, m5, [ps_mask3]
 
-    BUILDINVHIGHLOW m0, m1, m7
+    BUILDINVHIGHLOW m7, m0, m1
     andps   m7, m7, [ps_mask2]
 
     addps   m0, m0, m5
 
-    BUILDINVHIGHLOW m1, m2, m6
+    BUILDINVHIGHLOW m6, m1, m2
     andps   m6, m6, [ps_mask2]
 
     addps  m1, m1, m7
 
-    BUILDINVHIGHLOW m2, m3, m7
+    BUILDINVHIGHLOW m7, m2, m3
     andps   m7, m7, [ps_mask2]
 
     addps   m2, m2, m6
@@ -245,7 +271,7 @@ cglobal imdct36_float_%1, 4,4,9, out, buf, in, win
     subps  m0, m0, m1
     addps  m0, m0, m6
 
-    BUILDINVHIGHLOW m2, m3, m4
+    BUILDINVHIGHLOW m4, m2, m3
     shufps  m3, m3, m2, 0x4e
 
     ; we have tmp = {SwAPLH(m0), SwAPLH(m7), m3, m4, m5}
@@ -280,7 +306,9 @@ cglobal imdct36_float_%1, 4,4,9, out, buf, in, win
     ; permutation done
 
     PSHUFD  m6, m2, 0xb1
-    movlps  m7, [bufq + 64]
+    movss   m4, [bufq + 4*68]
+    movss   m7, [bufq + 4*64]
+    unpcklps  m7, m7, m4
     mulps   m6, m6, [winq + 16*4]
     addps   m6, m6, m7
     movss   [outq + 64*SBLIMIT], m6
@@ -288,74 +316,406 @@ cglobal imdct36_float_%1, 4,4,9, out, buf, in, win
     movss   [outq + 68*SBLIMIT], m6
 
     mulps   m6, m3, [winq + 4*4]
-    LOADA64 m4, bufq + 16
+    LOAD    m4, m7, bufq + 4*16, 16
     addps   m6, m6, m4
-    STORE   m6, m7, outq + 16*SBLIMIT
+    STORE   m6, m7, outq + 16*SBLIMIT, 4*SBLIMIT
 
     shufps  m4, m0, m3, 0xb5
     mulps   m4, m4, [winq + 8*4]
-    LOADA64 m7, bufq + 32
+    LOAD    m7, m6, bufq + 4*32, 16
     addps   m4, m4, m7
-    STORE   m4, m6, outq + 32*SBLIMIT
+    STORE   m4, m6, outq + 32*SBLIMIT, 4*SBLIMIT
 
     shufps  m3, m3, m2, 0xb1
     mulps   m3, m3, [winq + 12*4]
-    LOADA64 m7, bufq + 48
+    LOAD    m7, m6, bufq + 4*48, 16
     addps   m3, m3, m7
-    STORE   m3, m7, outq + 48*SBLIMIT
+    STORE   m3, m7, outq + 48*SBLIMIT, 4*SBLIMIT
 
     mulps   m2, m2, [winq]
-    LOADA64 m6, bufq
+    LOAD    m6, m7, bufq, 16
     addps   m2, m2, m6
-    STORE   m2, m7, outq
+    STORE   m2, m7, outq, 4*SBLIMIT
 
     mulps    m4, m1, [winq + 20*4]
-    STOREA64 bufq, m4
+    STORE    m4, m7, bufq, 16
 
     mulps    m3, m5, [winq + 24*4]
-    STOREA64 bufq + 16, m3
+    STORE    m3, m7, bufq + 4*16, 16
 
     shufps   m0, m0, m5, 0xb0
     mulps    m0, m0, [winq + 28*4]
-    STOREA64 bufq + 32, m0
+    STORE    m0, m7, bufq + 4*32, 16
 
     shufps   m5, m5, m1, 0xb1
     mulps    m5, m5, [winq + 32*4]
-    STOREA64 bufq + 48, m5
+    STORE    m5, m7, bufq + 4*48, 16
 
     shufps   m1, m1, m1, 0xb1
     mulps    m1, m1, [winq + 36*4]
-    movlps  [bufq + 64], m1
+    movss    [bufq + 4*64], m1
+    shufps   m1, m1, 0xb1
+    movss    [bufq + 4*68], m1
     RET
 %endmacro
 
-%define PSHUFD PSHUFD_SSE_AVX
-%define INVERTHL INVERTHL_SSE1
-%define BUTTERF  BUTTERF_SSE12
-%define BUTTERF0 BUTTERF0_SSE12
-%define BUILDINVHIGHLOW BUILDINVHIGHLOW_SSE
-%define ROTLEFT ROTLEFT_SSE
+INIT_XMM sse
+DEFINE_IMDCT
 
-INIT_XMM
+INIT_XMM sse2
+DEFINE_IMDCT
 
-DEFINE_IMDCT sse
+INIT_XMM sse3
+DEFINE_IMDCT
 
-%define PSHUFD PSHUFD_SSE2
-%define INVERTHL INVERTHL_SSE2
+INIT_XMM ssse3
+DEFINE_IMDCT
 
-DEFINE_IMDCT sse2
+INIT_XMM avx
+DEFINE_IMDCT
 
-%define BUTTERF  BUTTERF_SSE3
-%define BUTTERF0 BUTTERF0_SSE3
+INIT_XMM sse
 
-DEFINE_IMDCT sse3
-
-%define ROTLEFT ROTLEFT_SSSE3
+%ifdef ARCH_X86_64
+%define SPILL SWAP
+%define UNSPILL SWAP
+%define SPILLED(x) m %+ x
+%else
+%define SPILLED(x) [tmpq+(x-8)*16 + 32*4]
+%macro SPILL 2 ; xmm#, mempos
+    movaps SPILLED(%2), m%1
+%endmacro
+%macro UNSPILL 2
+    movaps m%1, SPILLED(%2)
+%endmacro
+%endif
 
-DEFINE_IMDCT ssse3
+%macro DEFINE_FOUR_IMDCT 0
+cglobal four_imdct36_float, 5,5,8, out, buf, in, win, tmp
+    movlps  m0, [inq+64]
+    movhps  m0, [inq+64 +   72]
+    movlps  m3, [inq+64 + 2*72]
+    movhps  m3, [inq+64 + 3*72]
+
+    shufps  m5, m0, m3, 0xdd
+    shufps  m0, m0, m3, 0x88
+
+    mova     m1, [inq+48]
+    movu     m6, [inq+48 +   72]
+    mova     m7, [inq+48 + 2*72]
+    movu     m3, [inq+48 + 3*72]
+
+    TRANSPOSE4x4PS 1, 6, 7, 3, 4
+
+    addps   m4, m6, m7
+    mova    [tmpq+4*28], m4
+
+    addps    m7, m3
+    addps    m6, m1
+    addps    m3, m0
+    addps    m0, m5
+    addps    m0, m7
+    addps    m7, m6
+    mova    [tmpq+4*12], m7
+    SPILL   3, 12
+
+    mova     m4, [inq+32]
+    movu     m5, [inq+32 +   72]
+    mova     m2, [inq+32 + 2*72]
+    movu     m7, [inq+32 + 3*72]
+
+    TRANSPOSE4x4PS 4, 5, 2, 7, 3
+
+    addps   m1, m7
+    SPILL   1, 11
+
+    addps   m3, m5, m2
+    SPILL   3, 13
+
+    addps    m7, m2
+    addps    m5, m4
+    addps    m6, m7
+    mova    [tmpq], m6
+    addps   m7, m5
+    mova    [tmpq+4*16], m7
+
+    mova    m2, [inq+16]
+    movu    m7, [inq+16 +   72]
+    mova    m1, [inq+16 + 2*72]
+    movu    m6, [inq+16 + 3*72]
+
+    TRANSPOSE4x4PS 2, 7, 1, 6, 3
+
+    addps   m4, m6
+    addps   m6, m1
+    addps   m1, m7
+    addps   m7, m2
+    addps   m5, m6
+    SPILL   5, 15
+    addps   m6, m7
+    mulps   m6, [costabs + 16*2]
+    mova    [tmpq+4*8], m6
+    SPILL   1, 10
+    SPILL   0, 14
+
+    mova    m1, [inq]
+    movu    m6, [inq +   72]
+    mova    m3, [inq + 2*72]
+    movu    m5, [inq + 3*72]
+
+    TRANSPOSE4x4PS 1, 6, 3, 5, 0
+
+    addps    m2, m5
+    addps    m5, m3
+    addps    m7, m5
+    addps    m3, m6
+    addps    m6, m1
+    SPILL    7, 8
+    addps    m5, m6
+    SPILL    6, 9
+    addps    m6, m4, SPILLED(12)
+    subps    m6, m2
+    UNSPILL  7, 11
+    SPILL    5, 11
+    subps    m5, m1, m7
+    mulps    m7, [costabs + 16*5]
+    addps    m7, m1
+    mulps    m0, m6, [costabs + 16*6]
+    addps    m0, m5
+    mova     [tmpq+4*24], m0
+    addps    m6, m5
+    mova     [tmpq+4*4], m6
+    addps    m6, m4, m2
+    mulps    m6, [costabs + 16*1]
+    subps    m4, SPILLED(12)
+    mulps    m4, [costabs + 16*8]
+    addps    m2, SPILLED(12)
+    mulps    m2, [costabs + 16*3]
+    subps    m5, m7, m6
+    subps    m5, m2
+    addps    m6, m7
+    addps    m6, m4
+    addps    m7, m2
+    subps    m7, m4
+    mova     [tmpq+4*20], m7
+    mova     m2, [tmpq+4*28]
+    mova     [tmpq+4*28], m5
+    UNSPILL  7, 13
+    subps    m5, m7, m2
+    mulps    m5, [costabs + 16*7]
+    UNSPILL  1, 10
+    mulps    m1, [costabs + 16*2]
+    addps    m4, m3, m2
+    mulps    m4, [costabs + 16*4]
+    addps    m2, m7
+    addps    m7, m3
+    mulps    m7, [costabs]
+    subps    m3, m2
+    mulps    m3, [costabs + 16*2]
+    addps    m2, m7, m5
+    addps    m2, m1
+    SPILL    2, 10
+    addps    m7, m4
+    subps    m7, m1
+    SPILL    7, 12
+    subps    m5, m4
+    subps    m5, m1
+    UNSPILL  0, 14
+    SPILL    5, 13
+    addps    m1, m0, SPILLED(15)
+    subps    m1, SPILLED(8)
+    mova     m4, [costabs + 16*5]
+    mulps    m4, [tmpq]
+    UNSPILL  2, 9
+    addps    m4, m2
+    subps    m2, [tmpq]
+    mulps    m5, m1, [costabs + 16*6]
+    addps    m5, m2
+    SPILL    5, 9
+    addps    m2, m1
+    SPILL    2, 14
+    UNSPILL  5, 15
+    subps    m7, m5, m0
+    addps    m5, SPILLED(8)
+    mulps    m5, [costabs + 16*1]
+    mulps    m7, [costabs + 16*8]
+    addps    m0, SPILLED(8)
+    mulps    m0, [costabs + 16*3]
+    subps    m2, m4, m5
+    subps    m2, m0
+    SPILL    2, 15
+    addps    m5, m4
+    addps    m5, m7
+    addps    m4, m0
+    subps    m4, m7
+    SPILL    4, 8
+    mova     m7, [tmpq+4*16]
+    mova     m2, [tmpq+4*12]
+    addps    m0, m7, m2
+    subps    m0, SPILLED(11)
+    mulps    m0, [costabs + 16*2]
+    addps    m4, m7, SPILLED(11)
+    mulps    m4, [costabs]
+    subps    m7, m2
+    mulps    m7, [costabs + 16*7]
+    addps    m2, SPILLED(11)
+    mulps    m2, [costabs + 16*4]
+    addps    m1, m7, [tmpq+4*8]
+    addps    m1, m4
+    addps    m4, m2
+    subps    m4, [tmpq+4*8]
+    SPILL    4, 11
+    subps    m7, m2
+    subps    m7, [tmpq+4*8]
+    addps    m4, m6, SPILLED(10)
+    subps    m6, SPILLED(10)
+    addps    m2, m5, m1
+    mulps    m2, [costabs + 16*9]
+    subps    m5, m1
+    mulps    m5, [costabs + 16*17]
+    subps    m1, m4, m2
+    addps    m4, m2
+    mulps    m2, m1, [winq+4*36]
+    addps    m2, [bufq+4*36]
+    mova     [outq+1152], m2
+    mulps    m1, [winq+4*32]
+    addps    m1, [bufq+4*32]
+    mova     [outq+1024], m1
+    mulps    m1, m4, [winq+4*116]
+    mova     [bufq+4*36], m1
+    mulps    m4, [winq+4*112]
+    mova     [bufq+4*32], m4
+    addps    m2, m6, m5
+    subps    m6, m5
+    mulps    m1, m6, [winq+4*68]
+    addps    m1, [bufq+4*68]
+    mova     [outq+2176], m1
+    mulps    m6, [winq]
+    addps    m6, [bufq]
+    mova     [outq], m6
+    mulps    m1, m2, [winq+4*148]
+    mova     [bufq+4*68], m1
+    mulps    m2, [winq+4*80]
+    mova     [bufq], m2
+    addps    m5, m3, [tmpq+4*24]
+    mova     m2, [tmpq+4*24]
+    subps    m2, m3
+    mova     m1, SPILLED(9)
+    subps    m1, m0
+    mulps    m1, [costabs + 16*10]
+    addps    m0, SPILLED(9)
+    mulps    m0, [costabs + 16*16]
+    addps    m6, m5, m1
+    subps    m5, m1
+    mulps    m3, m5, [winq+4*40]
+    addps    m3, [bufq+4*40]
+    mova     [outq+1280], m3
+    mulps    m5, [winq+4*28]
+    addps    m5, [bufq+4*28]
+    mova     [outq+896], m5
+    mulps    m1, m6, [winq+4*120]
+    mova     [bufq+4*40], m1
+    mulps    m6, [winq+4*108]
+    mova     [bufq+4*28], m6
+    addps    m1, m2, m0
+    subps    m2, m0
+    mulps    m5, m2, [winq+4*64]
+    addps    m5, [bufq+4*64]
+    mova     [outq+2048], m5
+    mulps    m2, [winq+4*4]
+    addps    m2, [bufq+4*4]
+    mova     [outq+128], m2
+    mulps    m0, m1, [winq+4*144]
+    mova     [bufq+4*64], m0
+    mulps    m1, [winq+4*84]
+    mova     [bufq+4*4], m1
+    mova     m1, [tmpq+4*28]
+    mova     m5, m1
+    addps    m1, SPILLED(13)
+    subps    m5, SPILLED(13)
+    UNSPILL  3, 15
+    addps    m2, m7, m3
+    mulps    m2, [costabs + 16*11]
+    subps    m3, m7
+    mulps    m3, [costabs + 16*15]
+    addps    m0, m2, m1
+    subps    m1, m2
+    SWAP     m0, m2
+    mulps    m6, m1, [winq+4*44]
+    addps    m6, [bufq+4*44]
+    mova     [outq+1408], m6
+    mulps    m1, [winq+4*24]
+    addps    m1, [bufq+4*24]
+    mova     [outq+768], m1
+    mulps    m0, m2, [winq+4*124]
+    mova     [bufq+4*44], m0
+    mulps    m2, [winq+4*104]
+    mova     [bufq+4*24], m2
+    addps    m0, m5, m3
+    subps    m5, m3
+    mulps    m1, m5, [winq+4*60]
+    addps    m1, [bufq+4*60]
+    mova     [outq+1920], m1
+    mulps    m5, [winq+4*8]
+    addps    m5, [bufq+4*8]
+    mova     [outq+256], m5
+    mulps    m1, m0, [winq+4*140]
+    mova     [bufq+4*60], m1
+    mulps    m0, [winq+4*88]
+    mova     [bufq+4*8], m0
+    mova     m1, [tmpq+4*20]
+    addps    m1, SPILLED(12)
+    mova     m2, [tmpq+4*20]
+    subps    m2, SPILLED(12)
+    UNSPILL  7, 8
+    subps    m0, m7, SPILLED(11)
+    addps    m7, SPILLED(11)
+    mulps    m4, m7, [costabs + 16*12]
+    mulps    m0, [costabs + 16*14]
+    addps    m5, m1, m4
+    subps    m1, m4
+    mulps    m7, m1, [winq+4*48]
+    addps    m7, [bufq+4*48]
+    mova     [outq+1536], m7
+    mulps    m1, [winq+4*20]
+    addps    m1, [bufq+4*20]
+    mova     [outq+640], m1
+    mulps    m1, m5, [winq+4*128]
+    mova     [bufq+4*48], m1
+    mulps    m5, [winq+4*100]
+    mova     [bufq+4*20], m5
+    addps    m6, m2, m0
+    subps    m2, m0
+    mulps    m1, m2, [winq+4*56]
+    addps    m1, [bufq+4*56]
+    mova     [outq+1792], m1
+    mulps    m2, [winq+4*12]
+    addps    m2, [bufq+4*12]
+    mova     [outq+384], m2
+    mulps    m0, m6, [winq+4*136]
+    mova    [bufq+4*56], m0
+    mulps    m6, [winq+4*92]
+    mova     [bufq+4*12], m6
+    UNSPILL  0, 14
+    mulps    m0, [costabs + 16*13]
+    mova     m3, [tmpq+4*4]
+    addps    m2, m0, m3
+    subps    m3, m0
+    mulps    m0, m3, [winq+4*52]
+    addps    m0, [bufq+4*52]
+    mova     [outq+1664], m0
+    mulps    m3, [winq+4*16]
+    addps    m3, [bufq+4*16]
+    mova     [outq+512], m3
+    mulps    m0, m2, [winq+4*132]
+    mova     [bufq+4*52], m0
+    mulps    m2, [winq+4*96]
+    mova     [bufq+4*16], m2
+    RET
+%endmacro
 
-%define BUILDINVHIGHLOW BUILDINVHIGHLOW_AVX
-%define PSHUFD PSHUFD_SSE_AVX
+INIT_XMM sse
+DEFINE_FOUR_IMDCT
 
-INIT_AVX
-DEFINE_IMDCT avx
+INIT_XMM avx
+DEFINE_FOUR_IMDCT
diff --git a/libavcodec/x86/mpegaudiodec_mmx.c b/libavcodec/x86/mpegaudiodec_mmx.c
index 980faf9cde..b1fa6573b8 100644
--- a/libavcodec/x86/mpegaudiodec_mmx.c
+++ b/libavcodec/x86/mpegaudiodec_mmx.c
@@ -29,6 +29,12 @@ void ff_imdct36_float_sse2(float *out, float *buf, float *in, float *win);
 void ff_imdct36_float_sse3(float *out, float *buf, float *in, float *win);
 void ff_imdct36_float_ssse3(float *out, float *buf, float *in, float *win);
 void ff_imdct36_float_avx(float *out, float *buf, float *in, float *win);
+void ff_four_imdct36_float_sse(float *out, float *buf, float *in, float *win,
+                               float *tmpbuf);
+void ff_four_imdct36_float_avx(float *out, float *buf, float *in, float *win,
+                               float *tmpbuf);
+
+DECLARE_ALIGNED(16, static float, mdct_win_sse)[2][4][4*40];
 
 #define MACS(rt, ra, rb) rt+=(ra)*(rb)
 #define MLSS(rt, ra, rb) rt-=(ra)*(rb)
@@ -153,26 +159,79 @@ static void apply_window_mp3(float *in, float *win, int *unused, float *out,
     *out = sum;
 }
 
+
+#define DECL_IMDCT_BLOCKS(CPU1, CPU2)                                       \
+static void imdct36_blocks_ ## CPU1(float *out, float *buf, float *in,      \
+                               int count, int switch_point, int block_type) \
+{                                                                           \
+    int align_end = count - (count & 3);                                \
+    int j;                                                              \
+    for (j = 0; j < align_end; j+= 4) {                                 \
+        LOCAL_ALIGNED_16(float, tmpbuf, [1024]);                        \
+        float *win = mdct_win_sse[switch_point && j < 4][block_type];   \
+        /* apply window & overlap with previous buffer */               \
+                                                                        \
+        /* select window */                                             \
+        ff_four_imdct36_float_ ## CPU2(out, buf, in, win, tmpbuf);      \
+        in      += 4*18;                                                \
+        buf     += 4*18;                                                \
+        out     += 4;                                                   \
+    }                                                                   \
+    for (; j < count; j++) {                                            \
+        /* apply window & overlap with previous buffer */               \
+                                                                        \
+        /* select window */                                             \
+        int win_idx = (switch_point && j < 2) ? 0 : block_type;         \
+        float *win = ff_mdct_win_float[win_idx + (4 & -(j & 1))];       \
+                                                                        \
+        ff_imdct36_float_ ## CPU1(out, buf, in, win);                   \
+                                                                        \
+        in  += 18;                                                      \
+        buf++;                                                          \
+        out++;                                                          \
+    }                                                                   \
+}
+
+DECL_IMDCT_BLOCKS(sse,sse)
+DECL_IMDCT_BLOCKS(sse2,sse)
+DECL_IMDCT_BLOCKS(sse3,sse)
+DECL_IMDCT_BLOCKS(ssse3,sse)
+DECL_IMDCT_BLOCKS(avx,avx)
+
 void ff_mpadsp_init_mmx(MPADSPContext *s)
 {
     int mm_flags = av_get_cpu_flags();
 
+    int i, j;
+    for (j = 0; j < 4; j++) {
+        for (i = 0; i < 40; i ++) {
+            mdct_win_sse[0][j][4*i    ] = ff_mdct_win_float[j    ][i];
+            mdct_win_sse[0][j][4*i + 1] = ff_mdct_win_float[j + 4][i];
+            mdct_win_sse[0][j][4*i + 2] = ff_mdct_win_float[j    ][i];
+            mdct_win_sse[0][j][4*i + 3] = ff_mdct_win_float[j + 4][i];
+            mdct_win_sse[1][j][4*i    ] = ff_mdct_win_float[0    ][i];
+            mdct_win_sse[1][j][4*i + 1] = ff_mdct_win_float[4    ][i];
+            mdct_win_sse[1][j][4*i + 2] = ff_mdct_win_float[j    ][i];
+            mdct_win_sse[1][j][4*i + 3] = ff_mdct_win_float[j + 4][i];
+        }
+    }
+
     if (mm_flags & AV_CPU_FLAG_SSE2) {
         s->apply_window_float = apply_window_mp3;
     }
-    if (HAVE_YASM && mm_flags & AV_CPU_FLAG_AVX && HAVE_AVX) {
-        s->imdct36_float = ff_imdct36_float_avx;
-    }
-    else if (HAVE_YASM && mm_flags & AV_CPU_FLAG_SSSE3 && HAVE_SSE) {
-        s->imdct36_float = ff_imdct36_float_ssse3;
-    }
-    else if (HAVE_YASM && mm_flags & AV_CPU_FLAG_SSE3 && HAVE_SSE) {
-        s->imdct36_float = ff_imdct36_float_sse3;
-    }
-    else if (HAVE_YASM && mm_flags & AV_CPU_FLAG_SSE2 && HAVE_SSE) {
-        s->imdct36_float = ff_imdct36_float_sse2;
-    }
-    else if (HAVE_YASM && mm_flags & AV_CPU_FLAG_SSE && HAVE_SSE) {
-        s->imdct36_float = ff_imdct36_float_sse;
+#if HAVE_YASM
+    if (mm_flags & AV_CPU_FLAG_AVX && HAVE_AVX) {
+        s->imdct36_blocks_float = imdct36_blocks_avx;
+#if HAVE_SSE
+    } else if (mm_flags & AV_CPU_FLAG_SSSE3) {
+        s->imdct36_blocks_float = imdct36_blocks_ssse3;
+    } else if (mm_flags & AV_CPU_FLAG_SSE3) {
+        s->imdct36_blocks_float = imdct36_blocks_sse3;
+    } else if (mm_flags & AV_CPU_FLAG_SSE2) {
+        s->imdct36_blocks_float = imdct36_blocks_sse2;
+    } else if (mm_flags & AV_CPU_FLAG_SSE) {
+        s->imdct36_blocks_float = imdct36_blocks_sse;
+#endif /* HAVE_SSE */
     }
+#endif /* HAVE_YASM */
 }
diff --git a/libavutil/x86/x86inc.asm b/libavutil/x86/x86inc.asm
index b1cc6b752e..a28c606d71 100644
--- a/libavutil/x86/x86inc.asm
+++ b/libavutil/x86/x86inc.asm
@@ -931,6 +931,8 @@ AVX_INSTR minpd, 1, 0, 1
 AVX_INSTR minps, 1, 0, 1
 AVX_INSTR minsd, 1, 0, 1
 AVX_INSTR minss, 1, 0, 1
+AVX_INSTR movhlps, 1, 0, 0
+AVX_INSTR movlhps, 1, 0, 0
 AVX_INSTR movsd, 1, 0, 0
 AVX_INSTR movss, 1, 0, 0
 AVX_INSTR mpsadbw, 0, 1, 0
diff --git a/libswscale/Makefile b/libswscale/Makefile
index 78d0112c8e..77d896a76b 100644
--- a/libswscale/Makefile
+++ b/libswscale/Makefile
@@ -19,7 +19,8 @@ OBJS-$(HAVE_MMX)           +=  x86/rgb2rgb.o            \
                                x86/swscale_mmx.o        \
                                x86/yuv2rgb_mmx.o
 OBJS-$(HAVE_VIS)           +=  sparc/yuv2rgb_vis.o
-MMX-OBJS-$(HAVE_YASM)      +=  x86/output.o             \
+MMX-OBJS-$(HAVE_YASM)      +=  x86/input.o              \
+                               x86/output.o             \
                                x86/scale.o
 
 $(SUBDIR)x86/swscale_mmx.o: CFLAGS += $(NOREDZONE_FLAGS)
diff --git a/libswscale/colorspace-test.c b/libswscale/colorspace-test.c
index a5709e482e..89713a8a0c 100644
--- a/libswscale/colorspace-test.c
+++ b/libswscale/colorspace-test.c
@@ -27,19 +27,19 @@
 #include "swscale.h"
 #include "rgb2rgb.h"
 
-#define SIZE 1000
+#define SIZE    1000
 #define srcByte 0x55
 #define dstByte 0xBB
 
-#define FUNC(s,d,n) {s,d,#n,n}
+#define FUNC(s, d, n) { s, d, #n, n }
 
 int main(int argc, char **argv)
 {
     int i, funcNum;
     uint8_t *srcBuffer = av_malloc(SIZE);
     uint8_t *dstBuffer = av_malloc(SIZE);
-    int failedNum=0;
-    int passedNum=0;
+    int failedNum      = 0;
+    int passedNum      = 0;
 
     if (!srcBuffer || !dstBuffer)
         return -1;
@@ -47,7 +47,7 @@ int main(int argc, char **argv)
     av_log(NULL, AV_LOG_INFO, "memory corruption test ...\n");
     sws_rgb2rgb_init();
 
-    for(funcNum=0; ; funcNum++) {
+    for (funcNum = 0; ; funcNum++) {
         struct func_info_s {
             int src_bpp;
             int dst_bpp;
@@ -85,67 +85,78 @@ int main(int argc, char **argv)
             FUNC(0, 0, NULL)
         };
         int width;
-        int failed=0;
-        int srcBpp=0;
-        int dstBpp=0;
+        int failed = 0;
+        int srcBpp = 0;
+        int dstBpp = 0;
 
-        if (!func_info[funcNum].func) break;
+        if (!func_info[funcNum].func)
+            break;
 
-        av_log(NULL, AV_LOG_INFO,".");
+        av_log(NULL, AV_LOG_INFO, ".");
         memset(srcBuffer, srcByte, SIZE);
 
-        for(width=63; width>0; width--) {
+        for (width = 63; width > 0; width--) {
             int dstOffset;
-            for(dstOffset=128; dstOffset<196; dstOffset+=4) {
+            for (dstOffset = 128; dstOffset < 196; dstOffset += 4) {
                 int srcOffset;
                 memset(dstBuffer, dstByte, SIZE);
 
-                for(srcOffset=128; srcOffset<196; srcOffset+=4) {
-                    uint8_t *src= srcBuffer+srcOffset;
-                    uint8_t *dst= dstBuffer+dstOffset;
-                    const char *name=NULL;
+                for (srcOffset = 128; srcOffset < 196; srcOffset += 4) {
+                    uint8_t *src     = srcBuffer + srcOffset;
+                    uint8_t *dst     = dstBuffer + dstOffset;
+                    const char *name = NULL;
 
-                    if(failed) break; //don't fill the screen with shit ...
+                    // don't fill the screen with shit ...
+                    if (failed)
+                        break;
 
                     srcBpp = func_info[funcNum].src_bpp;
                     dstBpp = func_info[funcNum].dst_bpp;
                     name   = func_info[funcNum].name;
 
-                    func_info[funcNum].func(src, dst, width*srcBpp);
+                    func_info[funcNum].func(src, dst, width * srcBpp);
 
-                    if(!srcBpp) break;
+                    if (!srcBpp)
+                        break;
 
-                    for(i=0; i<SIZE; i++) {
-                        if(srcBuffer[i]!=srcByte) {
-                            av_log(NULL, AV_LOG_INFO, "src damaged at %d w:%d src:%d dst:%d %s\n",
+                    for (i = 0; i < SIZE; i++) {
+                        if (srcBuffer[i] != srcByte) {
+                            av_log(NULL, AV_LOG_INFO,
+                                   "src damaged at %d w:%d src:%d dst:%d %s\n",
                                    i, width, srcOffset, dstOffset, name);
-                            failed=1;
+                            failed = 1;
                             break;
                         }
                     }
-                    for(i=0; i<dstOffset; i++) {
-                        if(dstBuffer[i]!=dstByte) {
-                            av_log(NULL, AV_LOG_INFO, "dst damaged at %d w:%d src:%d dst:%d %s\n",
+                    for (i = 0; i < dstOffset; i++) {
+                        if (dstBuffer[i] != dstByte) {
+                            av_log(NULL, AV_LOG_INFO,
+                                   "dst damaged at %d w:%d src:%d dst:%d %s\n",
                                    i, width, srcOffset, dstOffset, name);
-                            failed=1;
+                            failed = 1;
                             break;
                         }
                     }
-                    for(i=dstOffset + width*dstBpp; i<SIZE; i++) {
-                        if(dstBuffer[i]!=dstByte) {
-                            av_log(NULL, AV_LOG_INFO, "dst damaged at %d w:%d src:%d dst:%d %s\n",
+                    for (i = dstOffset + width * dstBpp; i < SIZE; i++) {
+                        if (dstBuffer[i] != dstByte) {
+                            av_log(NULL, AV_LOG_INFO,
+                                   "dst damaged at %d w:%d src:%d dst:%d %s\n",
                                    i, width, srcOffset, dstOffset, name);
-                            failed=1;
+                            failed = 1;
                             break;
                         }
                     }
                 }
             }
         }
-        if(failed) failedNum++;
-        else if(srcBpp) passedNum++;
+        if (failed)
+            failedNum++;
+        else if (srcBpp)
+            passedNum++;
     }
 
-    av_log(NULL, AV_LOG_INFO, "\n%d converters passed, %d converters randomly overwrote memory\n", passedNum, failedNum);
+    av_log(NULL, AV_LOG_INFO,
+           "\n%d converters passed, %d converters randomly overwrote memory\n",
+           passedNum, failedNum);
     return failedNum;
 }
diff --git a/libswscale/ppc/yuv2rgb_altivec.h b/libswscale/ppc/yuv2rgb_altivec.h
index 7c2a7e547b..8c62c322e7 100644
--- a/libswscale/ppc/yuv2rgb_altivec.h
+++ b/libswscale/ppc/yuv2rgb_altivec.h
@@ -24,13 +24,18 @@
 #ifndef SWSCALE_PPC_YUV2RGB_ALTIVEC_H
 #define SWSCALE_PPC_YUV2RGB_ALTIVEC_H
 
-#define YUV2PACKEDX_HEADER(suffix) \
-void ff_yuv2 ## suffix ## _X_altivec(SwsContext *c, const int16_t *lumFilter, \
-                            const int16_t **lumSrc, int lumFilterSize, \
-                            const int16_t *chrFilter, const int16_t **chrUSrc, \
-                            const int16_t **chrVSrc, int chrFilterSize, \
-                            const int16_t **alpSrc, uint8_t *dest, \
-                            int dstW, int dstY);
+#define YUV2PACKEDX_HEADER(suffix)                                  \
+    void ff_yuv2 ## suffix ## _X_altivec(SwsContext *c,             \
+                                         const int16_t *lumFilter,  \
+                                         const int16_t **lumSrc,    \
+                                         int lumFilterSize,         \
+                                         const int16_t *chrFilter,  \
+                                         const int16_t **chrUSrc,   \
+                                         const int16_t **chrVSrc,   \
+                                         int chrFilterSize,         \
+                                         const int16_t **alpSrc,    \
+                                         uint8_t *dest,             \
+                                         int dstW, int dstY);
 
 YUV2PACKEDX_HEADER(abgr);
 YUV2PACKEDX_HEADER(bgra);
diff --git a/libswscale/rgb2rgb.h b/libswscale/rgb2rgb.h
index e3edac88d4..a7542cb211 100644
--- a/libswscale/rgb2rgb.h
+++ b/libswscale/rgb2rgb.h
@@ -36,32 +36,33 @@ extern void (*rgb24tobgr32)(const uint8_t *src, uint8_t *dst, int src_size);
 extern void (*rgb24tobgr16)(const uint8_t *src, uint8_t *dst, int src_size);
 extern void (*rgb24tobgr15)(const uint8_t *src, uint8_t *dst, int src_size);
 extern void (*rgb32tobgr24)(const uint8_t *src, uint8_t *dst, int src_size);
-extern void (*rgb32to16)   (const uint8_t *src, uint8_t *dst, int src_size);
-extern void (*rgb32to15)   (const uint8_t *src, uint8_t *dst, int src_size);
-extern void (*rgb15to16)   (const uint8_t *src, uint8_t *dst, int src_size);
+extern void    (*rgb32to16)(const uint8_t *src, uint8_t *dst, int src_size);
+extern void    (*rgb32to15)(const uint8_t *src, uint8_t *dst, int src_size);
+extern void    (*rgb15to16)(const uint8_t *src, uint8_t *dst, int src_size);
 extern void (*rgb15tobgr24)(const uint8_t *src, uint8_t *dst, int src_size);
-extern void (*rgb15to32)   (const uint8_t *src, uint8_t *dst, int src_size);
-extern void (*rgb16to15)   (const uint8_t *src, uint8_t *dst, int src_size);
+extern void    (*rgb15to32)(const uint8_t *src, uint8_t *dst, int src_size);
+extern void    (*rgb16to15)(const uint8_t *src, uint8_t *dst, int src_size);
 extern void (*rgb16tobgr24)(const uint8_t *src, uint8_t *dst, int src_size);
-extern void (*rgb16to32)   (const uint8_t *src, uint8_t *dst, int src_size);
+extern void    (*rgb16to32)(const uint8_t *src, uint8_t *dst, int src_size);
 extern void (*rgb24tobgr24)(const uint8_t *src, uint8_t *dst, int src_size);
-extern void (*rgb24to16)   (const uint8_t *src, uint8_t *dst, int src_size);
-extern void (*rgb24to15)   (const uint8_t *src, uint8_t *dst, int src_size);
-extern void (*shuffle_bytes_2103)(const uint8_t *src, uint8_t *dst, int src_size);
+extern void    (*rgb24to16)(const uint8_t *src, uint8_t *dst, int src_size);
+extern void    (*rgb24to15)(const uint8_t *src, uint8_t *dst, int src_size);
 extern void (*rgb32tobgr16)(const uint8_t *src, uint8_t *dst, int src_size);
 extern void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, int src_size);
 
-void rgb24to32   (const uint8_t *src, uint8_t *dst, int src_size);
-void rgb32to24   (const uint8_t *src, uint8_t *dst, int src_size);
+extern void (*shuffle_bytes_2103)(const uint8_t *src, uint8_t *dst, int src_size);
+
+void    rgb24to32(const uint8_t *src, uint8_t *dst, int src_size);
+void    rgb32to24(const uint8_t *src, uint8_t *dst, int src_size);
 void rgb16tobgr32(const uint8_t *src, uint8_t *dst, int src_size);
-void rgb16to24   (const uint8_t *src, uint8_t *dst, int src_size);
+void    rgb16to24(const uint8_t *src, uint8_t *dst, int src_size);
 void rgb16tobgr16(const uint8_t *src, uint8_t *dst, int src_size);
 void rgb16tobgr15(const uint8_t *src, uint8_t *dst, int src_size);
 void rgb15tobgr32(const uint8_t *src, uint8_t *dst, int src_size);
-void rgb15to24   (const uint8_t *src, uint8_t *dst, int src_size);
+void    rgb15to24(const uint8_t *src, uint8_t *dst, int src_size);
 void rgb15tobgr16(const uint8_t *src, uint8_t *dst, int src_size);
 void rgb15tobgr15(const uint8_t *src, uint8_t *dst, int src_size);
-void bgr8torgb8  (const uint8_t *src, uint8_t *dst, int src_size);
+void   bgr8torgb8(const uint8_t *src, uint8_t *dst, int src_size);
 
 void shuffle_bytes_0321(const uint8_t *src, uint8_t *dst, int src_size);
 void shuffle_bytes_1230(const uint8_t *src, uint8_t *dst, int src_size);
@@ -138,7 +139,6 @@ extern void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint
                             int srcStride1, int srcStride2,
                             int srcStride3, int dstStride);
 
-
 extern void (*uyvytoyuv420)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
                             int width, int height,
                             int lumStride, int chromStride, int srcStride);
diff --git a/libswscale/swscale-test.c b/libswscale/swscale-test.c
index 5dd2e34870..f382ce3d6a 100644
--- a/libswscale/swscale-test.c
+++ b/libswscale/swscale-test.c
@@ -35,33 +35,32 @@
 
 /* HACK Duplicated from swscale_internal.h.
  * Should be removed when a cleaner pixel format system exists. */
-#define isGray(x)       (           \
-           (x)==PIX_FMT_GRAY8       \
-        || (x)==PIX_FMT_GRAY16BE    \
-        || (x)==PIX_FMT_GRAY16LE    \
-    )
-#define hasChroma(x)   (!(          \
-            isGray(x)               \
-        || (x)==PIX_FMT_MONOBLACK   \
-        || (x)==PIX_FMT_MONOWHITE   \
-    ))
-#define isALPHA(x)      (           \
-           (x)==PIX_FMT_BGR32       \
-        || (x)==PIX_FMT_BGR32_1     \
-        || (x)==PIX_FMT_RGB32       \
-        || (x)==PIX_FMT_RGB32_1     \
-        || (x)==PIX_FMT_YUVA420P    \
-    )
-
-static uint64_t getSSD(uint8_t *src1, uint8_t *src2, int stride1, int stride2, int w, int h)
+#define isGray(x)                      \
+    ((x) == PIX_FMT_GRAY8       ||     \
+     (x) == PIX_FMT_Y400A       ||     \
+     (x) == PIX_FMT_GRAY16BE    ||     \
+     (x) == PIX_FMT_GRAY16LE)
+#define hasChroma(x)                   \
+    (!(isGray(x)                ||     \
+       (x) == PIX_FMT_MONOBLACK ||     \
+       (x) == PIX_FMT_MONOWHITE))
+#define isALPHA(x)                     \
+    ((x) == PIX_FMT_BGR32   ||         \
+     (x) == PIX_FMT_BGR32_1 ||         \
+     (x) == PIX_FMT_RGB32   ||         \
+     (x) == PIX_FMT_RGB32_1 ||         \
+     (x) == PIX_FMT_YUVA420P)
+
+static uint64_t getSSD(uint8_t *src1, uint8_t *src2, int stride1,
+                       int stride2, int w, int h)
 {
-    int x,y;
-    uint64_t ssd=0;
+    int x, y;
+    uint64_t ssd = 0;
 
-    for (y=0; y<h; y++) {
-        for (x=0; x<w; x++) {
-            int d= src1[x + y*stride1] - src2[x + y*stride2];
-            ssd+= d*d;
+    for (y = 0; y < h; y++) {
+        for (x = 0; x < w; x++) {
+            int d = src1[x + y * stride1] - src2[x + y * stride2];
+            ssd += d * d;
         }
     }
     return ssd;
@@ -86,14 +85,14 @@ static int doTest(uint8_t *ref[4], int refStride[4], int w, int h,
     static int cur_srcW, cur_srcH;
     static uint8_t *src[4];
     static int srcStride[4];
-    uint8_t *dst[4] = {0};
-    uint8_t *out[4] = {0};
+    uint8_t *dst[4] = { 0 };
+    uint8_t *out[4] = { 0 };
     int dstStride[4];
     int i;
-    uint64_t ssdY, ssdU=0, ssdV=0, ssdA=0;
+    uint64_t ssdY, ssdU = 0, ssdV = 0, ssdA = 0;
     struct SwsContext *dstContext = NULL, *outContext = NULL;
     uint32_t crc = 0;
-    int res = 0;
+    int res      = 0;
 
     if (cur_srcFormat != srcFormat || cur_srcW != srcW || cur_srcH != srcH) {
         struct SwsContext *srcContext = NULL;
@@ -106,11 +105,10 @@ static int doTest(uint8_t *ref[4], int refStride[4], int w, int h,
         for (p = 0; p < 4; p++) {
             srcStride[p] = FFALIGN(srcStride[p], 16);
             if (srcStride[p])
-                src[p] = av_mallocz(srcStride[p]*srcH+16);
+                src[p] = av_mallocz(srcStride[p] * srcH + 16);
             if (srcStride[p] && !src[p]) {
                 perror("Malloc");
                 res = -1;
-
                 goto end;
             }
         }
@@ -121,19 +119,18 @@ static int doTest(uint8_t *ref[4], int refStride[4], int w, int h,
                     av_pix_fmt_descriptors[PIX_FMT_YUVA420P].name,
                     av_pix_fmt_descriptors[srcFormat].name);
             res = -1;
-
             goto end;
         }
         sws_scale(srcContext, ref, refStride, 0, h, src, srcStride);
         sws_freeContext(srcContext);
 
         cur_srcFormat = srcFormat;
-        cur_srcW = srcW;
-        cur_srcH = srcH;
+        cur_srcW      = srcW;
+        cur_srcH      = srcH;
     }
 
     av_image_fill_linesizes(dstStride, dstFormat, dstW);
-    for (i=0; i<4; i++) {
+    for (i = 0; i < 4; i++) {
         /* Image buffers passed into libswscale can be allocated any way you
          * prefer, as long as they're aligned enough for the architecture, and
          * they're freed appropriately (such as using av_free for buffers
@@ -142,7 +139,7 @@ static int doTest(uint8_t *ref[4], int refStride[4], int w, int h,
          * out of bounds. */
         dstStride[i] = FFALIGN(dstStride[i], 16);
         if (dstStride[i])
-            dst[i]= av_mallocz(dstStride[i]*dstH+16);
+            dst[i] = av_mallocz(dstStride[i] * dstH + 16);
         if (dstStride[i] && !dst[i]) {
             perror("Malloc");
             res = -1;
@@ -151,13 +148,13 @@ static int doTest(uint8_t *ref[4], int refStride[4], int w, int h,
         }
     }
 
-    dstContext= sws_getContext(srcW, srcH, srcFormat, dstW, dstH, dstFormat, flags, NULL, NULL, NULL);
+    dstContext = sws_getContext(srcW, srcH, srcFormat, dstW, dstH, dstFormat,
+                                flags, NULL, NULL, NULL);
     if (!dstContext) {
         fprintf(stderr, "Failed to get %s ---> %s\n",
                 av_pix_fmt_descriptors[srcFormat].name,
                 av_pix_fmt_descriptors[dstFormat].name);
         res = -1;
-
         goto end;
     }
 
@@ -169,9 +166,9 @@ static int doTest(uint8_t *ref[4], int refStride[4], int w, int h,
 
     sws_scale(dstContext, src, srcStride, 0, srcH, dst, dstStride);
 
-    for (i = 0; i < 4 && dstStride[i]; i++) {
-        crc = av_crc(av_crc_get_table(AV_CRC_32_IEEE), crc, dst[i], dstStride[i] * dstH);
-    }
+    for (i = 0; i < 4 && dstStride[i]; i++)
+        crc = av_crc(av_crc_get_table(AV_CRC_32_IEEE), crc, dst[i],
+                     dstStride[i] * dstH);
 
     if (r && crc == r->crc) {
         ssdY = r->ssdY;
@@ -179,61 +176,60 @@ static int doTest(uint8_t *ref[4], int refStride[4], int w, int h,
         ssdV = r->ssdV;
         ssdA = r->ssdA;
     } else {
-        for (i=0; i<4; i++) {
+        for (i = 0; i < 4; i++) {
             refStride[i] = FFALIGN(refStride[i], 16);
             if (refStride[i])
-                out[i]= av_mallocz(refStride[i]*h);
+                out[i] = av_mallocz(refStride[i] * h);
             if (refStride[i] && !out[i]) {
                 perror("Malloc");
                 res = -1;
-
                 goto end;
             }
         }
-        outContext= sws_getContext(dstW, dstH, dstFormat, w, h, PIX_FMT_YUVA420P, SWS_BILINEAR, NULL, NULL, NULL);
+        outContext = sws_getContext(dstW, dstH, dstFormat, w, h,
+                                    PIX_FMT_YUVA420P, SWS_BILINEAR,
+                                    NULL, NULL, NULL);
         if (!outContext) {
             fprintf(stderr, "Failed to get %s ---> %s\n",
                     av_pix_fmt_descriptors[dstFormat].name,
                     av_pix_fmt_descriptors[PIX_FMT_YUVA420P].name);
             res = -1;
-
             goto end;
         }
         sws_scale(outContext, dst, dstStride, 0, dstH, out, refStride);
 
-        ssdY= getSSD(ref[0], out[0], refStride[0], refStride[0], w, h);
+        ssdY = getSSD(ref[0], out[0], refStride[0], refStride[0], w, h);
         if (hasChroma(srcFormat) && hasChroma(dstFormat)) {
             //FIXME check that output is really gray
-            ssdU= getSSD(ref[1], out[1], refStride[1], refStride[1], (w+1)>>1, (h+1)>>1);
-            ssdV= getSSD(ref[2], out[2], refStride[2], refStride[2], (w+1)>>1, (h+1)>>1);
+            ssdU = getSSD(ref[1], out[1], refStride[1], refStride[1],
+                          (w + 1) >> 1, (h + 1) >> 1);
+            ssdV = getSSD(ref[2], out[2], refStride[2], refStride[2],
+                          (w + 1) >> 1, (h + 1) >> 1);
         }
         if (isALPHA(srcFormat) && isALPHA(dstFormat))
-            ssdA= getSSD(ref[3], out[3], refStride[3], refStride[3], w, h);
+            ssdA = getSSD(ref[3], out[3], refStride[3], refStride[3], w, h);
 
-        ssdY/= w*h;
-        ssdU/= w*h/4;
-        ssdV/= w*h/4;
-        ssdA/= w*h;
+        ssdY /= w * h;
+        ssdU /= w * h / 4;
+        ssdV /= w * h / 4;
+        ssdA /= w * h;
 
         sws_freeContext(outContext);
 
-        for (i=0; i<4; i++) {
+        for (i = 0; i < 4; i++)
             if (refStride[i])
                 av_free(out[i]);
-        }
     }
 
-    printf(" CRC=%08x SSD=%5"PRId64",%5"PRId64",%5"PRId64",%5"PRId64"\n",
+    printf(" CRC=%08x SSD=%5"PRId64 ",%5"PRId64 ",%5"PRId64 ",%5"PRId64 "\n",
            crc, ssdY, ssdU, ssdV, ssdA);
 
 end:
-
     sws_freeContext(dstContext);
 
-    for (i=0; i<4; i++) {
+    for (i = 0; i < 4; i++)
         if (dstStride[i])
             av_free(dst[i]);
-    }
 
     return res;
 }
@@ -242,18 +238,18 @@ static void selfTest(uint8_t *ref[4], int refStride[4], int w, int h,
                      enum PixelFormat srcFormat_in,
                      enum PixelFormat dstFormat_in)
 {
-    const int flags[] = { SWS_FAST_BILINEAR,
-                          SWS_BILINEAR, SWS_BICUBIC,
-                          SWS_X       , SWS_POINT  , SWS_AREA, 0 };
-    const int srcW = w;
-    const int srcH = h;
-    const int dstW[] = { srcW - srcW/3, srcW, srcW + srcW/3, 0 };
-    const int dstH[] = { srcH - srcH/3, srcH, srcH + srcH/3, 0 };
+    const int flags[] = { SWS_FAST_BILINEAR, SWS_BILINEAR, SWS_BICUBIC,
+                          SWS_X, SWS_POINT, SWS_AREA, 0 };
+    const int srcW   = w;
+    const int srcH   = h;
+    const int dstW[] = { srcW - srcW / 3, srcW, srcW + srcW / 3, 0 };
+    const int dstH[] = { srcH - srcH / 3, srcH, srcH + srcH / 3, 0 };
     enum PixelFormat srcFormat, dstFormat;
 
     for (srcFormat = srcFormat_in != PIX_FMT_NONE ? srcFormat_in : 0;
          srcFormat < PIX_FMT_NB; srcFormat++) {
-        if (!sws_isSupportedInput(srcFormat) || !sws_isSupportedOutput(srcFormat))
+        if (!sws_isSupportedInput(srcFormat) ||
+            !sws_isSupportedOutput(srcFormat))
             continue;
 
         for (dstFormat = dstFormat_in != PIX_FMT_NONE ? dstFormat_in : 0;
@@ -261,7 +257,8 @@ static void selfTest(uint8_t *ref[4], int refStride[4], int w, int h,
             int i, j, k;
             int res = 0;
 
-            if (!sws_isSupportedInput(dstFormat) || !sws_isSupportedOutput(dstFormat))
+            if (!sws_isSupportedInput(dstFormat) ||
+                !sws_isSupportedOutput(dstFormat))
                 continue;
 
             printf("%s -> %s\n",
@@ -269,14 +266,13 @@ static void selfTest(uint8_t *ref[4], int refStride[4], int w, int h,
                    av_pix_fmt_descriptors[dstFormat].name);
             fflush(stdout);
 
-            for (k = 0; flags[k] && !res; k++) {
+            for (k = 0; flags[k] && !res; k++)
                 for (i = 0; dstW[i] && !res; i++)
                     for (j = 0; dstH[j] && !res; j++)
                         res = doTest(ref, refStride, w, h,
                                      srcFormat, dstFormat,
                                      srcW, srcH, dstW[i], dstH[j], flags[k],
                                      NULL);
-            }
             if (dstFormat_in != PIX_FMT_NONE)
                 break;
         }
@@ -302,13 +298,14 @@ static int fileTest(uint8_t *ref[4], int refStride[4], int w, int h, FILE *fp,
         int flags;
         int ret;
 
-        ret = sscanf(buf, " %12s %dx%d -> %12s %dx%d flags=%d CRC=%x"
-                          " SSD=%"PRId64", %"PRId64", %"PRId64", %"PRId64"\n",
-                          srcStr, &srcW, &srcH, dstStr, &dstW, &dstH,
-                          &flags, &r.crc, &r.ssdY, &r.ssdU, &r.ssdV, &r.ssdA);
+        ret = sscanf(buf,
+                     " %12s %dx%d -> %12s %dx%d flags=%d CRC=%x"
+                     " SSD=%"PRId64 ", %"PRId64 ", %"PRId64 ", %"PRId64 "\n",
+                     srcStr, &srcW, &srcH, dstStr, &dstW, &dstH,
+                     &flags, &r.crc, &r.ssdY, &r.ssdU, &r.ssdV, &r.ssdA);
         if (ret != 12) {
             srcStr[0] = dstStr[0] = 0;
-            ret = sscanf(buf, "%12s -> %12s\n", srcStr, dstStr);
+            ret       = sscanf(buf, "%12s -> %12s\n", srcStr, dstStr);
         }
 
         srcFormat = av_get_pix_fmt(srcStr);
@@ -342,12 +339,12 @@ int main(int argc, char **argv)
 {
     enum PixelFormat srcFormat = PIX_FMT_NONE;
     enum PixelFormat dstFormat = PIX_FMT_NONE;
-    uint8_t *rgb_data = av_malloc (W*H*4);
-    uint8_t *rgb_src[3]= {rgb_data, NULL, NULL};
-    int rgb_stride[3]={4*W, 0, 0};
-    uint8_t *data = av_malloc (4*W*H);
-    uint8_t *src[4]= {data, data+W*H, data+W*H*2, data+W*H*3};
-    int stride[4]={W, W, W, W};
+    uint8_t *rgb_data   = av_malloc(W * H * 4);
+    uint8_t *rgb_src[3] = { rgb_data, NULL, NULL };
+    int rgb_stride[3]   = { 4 * W, 0, 0 };
+    uint8_t *data       = av_malloc(4 * W * H);
+    uint8_t *src[4]     = { data, data + W * H, data + W * H * 2, data + W * H * 3 };
+    int stride[4]       = { W, W, W, W };
     int x, y;
     struct SwsContext *sws;
     AVLFG rand;
@@ -357,41 +354,40 @@ int main(int argc, char **argv)
     if (!rgb_data || !data)
         return -1;
 
-    sws= sws_getContext(W/12, H/12, PIX_FMT_RGB32, W, H, PIX_FMT_YUVA420P, SWS_BILINEAR, NULL, NULL, NULL);
+    sws = sws_getContext(W / 12, H / 12, PIX_FMT_RGB32, W, H,
+                         PIX_FMT_YUVA420P, SWS_BILINEAR, NULL, NULL, NULL);
 
     av_lfg_init(&rand, 1);
 
-    for (y=0; y<H; y++) {
-        for (x=0; x<W*4; x++) {
-            rgb_data[ x + y*4*W]= av_lfg_get(&rand);
-        }
-    }
+    for (y = 0; y < H; y++)
+        for (x = 0; x < W * 4; x++)
+            rgb_data[ x + y * 4 * W] = av_lfg_get(&rand);
     sws_scale(sws, rgb_src, rgb_stride, 0, H, src, stride);
     sws_freeContext(sws);
     av_free(rgb_data);
 
     for (i = 1; i < argc; i += 2) {
-        if (argv[i][0] != '-' || i+1 == argc)
+        if (argv[i][0] != '-' || i + 1 == argc)
             goto bad_option;
         if (!strcmp(argv[i], "-ref")) {
-            FILE *fp = fopen(argv[i+1], "r");
+            FILE *fp = fopen(argv[i + 1], "r");
             if (!fp) {
-                fprintf(stderr, "could not open '%s'\n", argv[i+1]);
+                fprintf(stderr, "could not open '%s'\n", argv[i + 1]);
                 goto error;
             }
             res = fileTest(src, stride, W, H, fp, srcFormat, dstFormat);
             fclose(fp);
             goto end;
         } else if (!strcmp(argv[i], "-src")) {
-            srcFormat = av_get_pix_fmt(argv[i+1]);
+            srcFormat = av_get_pix_fmt(argv[i + 1]);
             if (srcFormat == PIX_FMT_NONE) {
-                fprintf(stderr, "invalid pixel format %s\n", argv[i+1]);
+                fprintf(stderr, "invalid pixel format %s\n", argv[i + 1]);
                 return -1;
             }
         } else if (!strcmp(argv[i], "-dst")) {
-            dstFormat = av_get_pix_fmt(argv[i+1]);
+            dstFormat = av_get_pix_fmt(argv[i + 1]);
             if (dstFormat == PIX_FMT_NONE) {
-                fprintf(stderr, "invalid pixel format %s\n", argv[i+1]);
+                fprintf(stderr, "invalid pixel format %s\n", argv[i + 1]);
                 return -1;
             }
         } else {
diff --git a/libswscale/swscale.c b/libswscale/swscale.c
index dca6b073d0..3cb9bfdd27 100644
--- a/libswscale/swscale.c
+++ b/libswscale/swscale.c
@@ -1637,12 +1637,16 @@ rgb16_32_wrapper(PIX_FMT_RGB32,    rgb32,   0, 0, 16, 0,   0x00FF, 0xFF00, 0xFF0
 rgb16_32_wrapper(PIX_FMT_RGB32_1,  rgb321,  0, 0, 16, 8,   0x00FF, 0xFF00, 0xFF0000,  8, 0,  8, RGB2YUV_SHIFT+8)
 rgb16_32_wrapper(PIX_FMT_BGR565LE, bgr16le, 0, 0,  0, 0,   0x001F, 0x07E0,   0xF800, 11, 5,  0, RGB2YUV_SHIFT+8)
 rgb16_32_wrapper(PIX_FMT_BGR555LE, bgr15le, 0, 0,  0, 0,   0x001F, 0x03E0,   0x7C00, 10, 5,  0, RGB2YUV_SHIFT+7)
+rgb16_32_wrapper(PIX_FMT_BGR444LE, bgr12le, 0, 0,  0, 0,   0x000F, 0x00F0,   0x0F00,  8, 4,  0, RGB2YUV_SHIFT+4)
 rgb16_32_wrapper(PIX_FMT_RGB565LE, rgb16le, 0, 0,  0, 0,   0xF800, 0x07E0,   0x001F,  0, 5, 11, RGB2YUV_SHIFT+8)
 rgb16_32_wrapper(PIX_FMT_RGB555LE, rgb15le, 0, 0,  0, 0,   0x7C00, 0x03E0,   0x001F,  0, 5, 10, RGB2YUV_SHIFT+7)
+rgb16_32_wrapper(PIX_FMT_RGB444LE, rgb12le, 0, 0,  0, 0,   0x0F00, 0x00F0,   0x000F,  0, 4,  8, RGB2YUV_SHIFT+4)
 rgb16_32_wrapper(PIX_FMT_BGR565BE, bgr16be, 0, 0,  0, 0,   0x001F, 0x07E0,   0xF800, 11, 5,  0, RGB2YUV_SHIFT+8)
 rgb16_32_wrapper(PIX_FMT_BGR555BE, bgr15be, 0, 0,  0, 0,   0x001F, 0x03E0,   0x7C00, 10, 5,  0, RGB2YUV_SHIFT+7)
+rgb16_32_wrapper(PIX_FMT_BGR444BE, bgr12be, 0, 0,  0, 0,   0x000F, 0x00F0,   0x0F00,  8, 4,  0, RGB2YUV_SHIFT+4)
 rgb16_32_wrapper(PIX_FMT_RGB565BE, rgb16be, 0, 0,  0, 0,   0xF800, 0x07E0,   0x001F,  0, 5, 11, RGB2YUV_SHIFT+8)
 rgb16_32_wrapper(PIX_FMT_RGB555BE, rgb15be, 0, 0,  0, 0,   0x7C00, 0x03E0,   0x001F,  0, 5, 10, RGB2YUV_SHIFT+7)
+rgb16_32_wrapper(PIX_FMT_RGB444BE, rgb12be, 0, 0,  0, 0,   0x0F00, 0x00F0,   0x000F,  0, 4,  8, RGB2YUV_SHIFT+4)
 
 static void gbr24pToUV_half_c(uint16_t *dstU, uint16_t *dstV,
                          const uint8_t *gsrc, const uint8_t *bsrc, const uint8_t *rsrc,
@@ -2887,6 +2891,8 @@ static av_cold void sws_init_swScale_c(SwsContext *c)
         case PIX_FMT_BGR565BE: c->chrToYV12 = bgr16beToUV_half_c; break;
         case PIX_FMT_BGR555LE: c->chrToYV12 = bgr15leToUV_half_c; break;
         case PIX_FMT_BGR555BE: c->chrToYV12 = bgr15beToUV_half_c; break;
+        case PIX_FMT_BGR444LE: c->chrToYV12 = bgr12leToUV_half_c; break;
+        case PIX_FMT_BGR444BE: c->chrToYV12 = bgr12beToUV_half_c; break;
         case PIX_FMT_BGR32   : c->chrToYV12 = rgb32ToUV_half_c;   break;
         case PIX_FMT_BGR32_1 : c->chrToYV12 = rgb321ToUV_half_c;  break;
         case PIX_FMT_RGB24   : c->chrToYV12 = rgb24ToUV_half_c;   break;
@@ -2895,6 +2901,8 @@ static av_cold void sws_init_swScale_c(SwsContext *c)
         case PIX_FMT_RGB555LE: c->chrToYV12 = rgb15leToUV_half_c; break;
         case PIX_FMT_RGB555BE: c->chrToYV12 = rgb15beToUV_half_c; break;
         case PIX_FMT_GBR24P  : c->chrToYV12 = gbr24pToUV_half_c;  break;
+        case PIX_FMT_RGB444LE: c->chrToYV12 = rgb12leToUV_half_c; break;
+        case PIX_FMT_RGB444BE: c->chrToYV12 = rgb12beToUV_half_c; break;
         }
     } else {
         switch(srcFormat) {
@@ -2909,6 +2917,8 @@ static av_cold void sws_init_swScale_c(SwsContext *c)
         case PIX_FMT_BGR565BE: c->chrToYV12 = bgr16beToUV_c; break;
         case PIX_FMT_BGR555LE: c->chrToYV12 = bgr15leToUV_c; break;
         case PIX_FMT_BGR555BE: c->chrToYV12 = bgr15beToUV_c; break;
+        case PIX_FMT_BGR444LE: c->chrToYV12 = bgr12leToUV_c; break;
+        case PIX_FMT_BGR444BE: c->chrToYV12 = bgr12beToUV_c; break;
         case PIX_FMT_BGR32   : c->chrToYV12 = rgb32ToUV_c;   break;
         case PIX_FMT_BGR32_1 : c->chrToYV12 = rgb321ToUV_c;  break;
         case PIX_FMT_RGB24   : c->chrToYV12 = rgb24ToUV_c;   break;
@@ -2916,6 +2926,8 @@ static av_cold void sws_init_swScale_c(SwsContext *c)
         case PIX_FMT_RGB565BE: c->chrToYV12 = rgb16beToUV_c; break;
         case PIX_FMT_RGB555LE: c->chrToYV12 = rgb15leToUV_c; break;
         case PIX_FMT_RGB555BE: c->chrToYV12 = rgb15beToUV_c; break;
+        case PIX_FMT_RGB444LE: c->chrToYV12 = rgb12leToUV_c; break;
+        case PIX_FMT_RGB444BE: c->chrToYV12 = rgb12beToUV_c; break;
         }
     }
 
@@ -2960,11 +2972,15 @@ static av_cold void sws_init_swScale_c(SwsContext *c)
     case PIX_FMT_BGR565BE : c->lumToYV12 = bgr16beToY_c; break;
     case PIX_FMT_BGR555LE : c->lumToYV12 = bgr15leToY_c; break;
     case PIX_FMT_BGR555BE : c->lumToYV12 = bgr15beToY_c; break;
+    case PIX_FMT_BGR444LE : c->lumToYV12 = bgr12leToY_c; break;
+    case PIX_FMT_BGR444BE : c->lumToYV12 = bgr12beToY_c; break;
     case PIX_FMT_RGB24    : c->lumToYV12 = rgb24ToY_c;   break;
     case PIX_FMT_RGB565LE : c->lumToYV12 = rgb16leToY_c; break;
     case PIX_FMT_RGB565BE : c->lumToYV12 = rgb16beToY_c; break;
     case PIX_FMT_RGB555LE : c->lumToYV12 = rgb15leToY_c; break;
     case PIX_FMT_RGB555BE : c->lumToYV12 = rgb15beToY_c; break;
+    case PIX_FMT_RGB444LE : c->lumToYV12 = rgb12leToY_c; break;
+    case PIX_FMT_RGB444BE : c->lumToYV12 = rgb12beToY_c; break;
     case PIX_FMT_RGB8     :
     case PIX_FMT_BGR8     :
     case PIX_FMT_PAL8     :
diff --git a/libswscale/swscale.h b/libswscale/swscale.h
index f65d0767c0..fa7100c41a 100644
--- a/libswscale/swscale.h
+++ b/libswscale/swscale.h
@@ -135,7 +135,6 @@ const char *swscale_license(void);
  */
 const int *sws_getCoefficients(int colorspace);
 
-
 // when used for filters they must have an odd number of elements
 // coeffs cannot be shared between vectors
 typedef struct {
@@ -235,9 +234,9 @@ struct SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat
  *                  the destination image
  * @return          the height of the output slice
  */
-int sws_scale(struct SwsContext *c, const uint8_t* const srcSlice[],
+int sws_scale(struct SwsContext *c, const uint8_t *const srcSlice[],
               const int srcStride[], int srcSliceY, int srcSliceH,
-              uint8_t* const dst[], const int dstStride[]);
+              uint8_t *const dst[], const int dstStride[]);
 
 /**
  * @param inv_table the yuv2rgb coefficients, normally ff_yuv2rgb_coeffs[x]
diff --git a/libswscale/swscale_internal.h b/libswscale/swscale_internal.h
index a9830eff20..f05925f842 100644
--- a/libswscale/swscale_internal.h
+++ b/libswscale/swscale_internal.h
@@ -32,9 +32,9 @@
 #include "libavutil/pixfmt.h"
 #include "libavutil/pixdesc.h"
 
-#define STR(s)         AV_TOSTRING(s) //AV_STRINGIFY is too long
+#define STR(s) AV_TOSTRING(s) // AV_STRINGIFY is too long
 
-#define FAST_BGR2YV12 //use 7-bit instead of 15-bit coefficients
+#define FAST_BGR2YV12 // use 7-bit instead of 15-bit coefficients
 
 #define MAX_FILTER_SIZE 256
 
@@ -47,21 +47,20 @@
 #endif
 
 #if ARCH_X86_64
-#   define APCK_PTR2 8
+#   define APCK_PTR2  8
 #   define APCK_COEF 16
 #   define APCK_SIZE 24
 #else
-#   define APCK_PTR2 4
-#   define APCK_COEF 8
+#   define APCK_PTR2  4
+#   define APCK_COEF  8
 #   define APCK_SIZE 16
 #endif
 
 struct SwsContext;
 
-typedef int (*SwsFunc)(struct SwsContext *context, const uint8_t* src[],
+typedef int (*SwsFunc)(struct SwsContext *context, const uint8_t *src[],
                        int srcStride[], int srcSliceY, int srcSliceH,
-                       uint8_t* dst[], int dstStride[]);
-
+                       uint8_t *dst[], int dstStride[]);
 
 /**
  * Write one line of horizontally scaled data to planar output
@@ -75,8 +74,8 @@ typedef int (*SwsFunc)(struct SwsContext *context, const uint8_t* src[],
  * @param dither  ordered dither array of type int16_t and size 8
  * @param offset  Dither offset
  */
-typedef void (*yuv2planar1_fn) (const int16_t *src, uint8_t *dest, int dstW,
-                                const uint8_t *dither, int offset);
+typedef void (*yuv2planar1_fn)(const int16_t *src, uint8_t *dest, int dstW,
+                               const uint8_t *dither, int offset);
 
 /**
  * Write one line of horizontally scaled data to planar output
@@ -91,9 +90,9 @@ typedef void (*yuv2planar1_fn) (const int16_t *src, uint8_t *dest, int dstW,
  * @param dstW          width of destination pixels
  * @param offset        Dither offset
  */
-typedef void (*yuv2planarX_fn) (const int16_t *filter, int filterSize,
-                                const int16_t **src, uint8_t *dest, int dstW,
-                                const uint8_t *dither, int offset);
+typedef void (*yuv2planarX_fn)(const int16_t *filter, int filterSize,
+                               const int16_t **src, uint8_t *dest, int dstW,
+                               const uint8_t *dither, int offset);
 
 /**
  * Write one line of horizontally scaled chroma to interleaved output
@@ -110,9 +109,12 @@ typedef void (*yuv2planarX_fn) (const int16_t *filter, int filterSize,
  *                      output, this is in uint16_t
  * @param dstW          width of chroma planes
  */
-typedef void (*yuv2interleavedX_fn) (struct SwsContext *c, const int16_t *chrFilter, int chrFilterSize,
-                                     const int16_t **chrUSrc, const int16_t **chrVSrc,
-                                     uint8_t *dest, int dstW);
+typedef void (*yuv2interleavedX_fn)(struct SwsContext *c,
+                                    const int16_t *chrFilter,
+                                    int chrFilterSize,
+                                    const int16_t **chrUSrc,
+                                    const int16_t **chrVSrc,
+                                    uint8_t *dest, int dstW);
 
 /**
  * Write one line of horizontally scaled Y/U/V/A to packed-pixel YUV/RGB
@@ -143,10 +145,11 @@ typedef void (*yuv2interleavedX_fn) (struct SwsContext *c, const int16_t *chrFil
  *                but can be used to generate comfort noise using dithering
  *                for some output formats.
  */
-typedef void (*yuv2packed1_fn) (struct SwsContext *c,  const int16_t *lumSrc,
-                                const int16_t *chrUSrc[2], const int16_t *chrVSrc[2],
-                                const int16_t *alpSrc,  uint8_t *dest,
-                                int dstW, int uvalpha, int y);
+typedef void (*yuv2packed1_fn)(struct SwsContext *c, const int16_t *lumSrc,
+                               const int16_t *chrUSrc[2],
+                               const int16_t *chrVSrc[2],
+                               const int16_t *alpSrc, uint8_t *dest,
+                               int dstW, int uvalpha, int y);
 /**
  * Write one line of horizontally scaled Y/U/V/A to packed-pixel YUV/RGB
  * output by doing bilinear scaling between two input lines.
@@ -175,10 +178,12 @@ typedef void (*yuv2packed1_fn) (struct SwsContext *c,  const int16_t *lumSrc,
  *                but can be used to generate comfort noise using dithering
  *                for some output formats.
  */
-typedef void (*yuv2packed2_fn) (struct SwsContext *c,  const int16_t *lumSrc[2],
-                                const int16_t *chrUSrc[2], const int16_t *chrVSrc[2],
-                                const int16_t *alpSrc[2], uint8_t *dest,
-                                int dstW, int yalpha, int uvalpha, int y);
+typedef void (*yuv2packed2_fn)(struct SwsContext *c, const int16_t *lumSrc[2],
+                               const int16_t *chrUSrc[2],
+                               const int16_t *chrVSrc[2],
+                               const int16_t *alpSrc[2],
+                               uint8_t *dest,
+                               int dstW, int yalpha, int uvalpha, int y);
 /**
  * Write one line of horizontally scaled Y/U/V/A to packed-pixel YUV/RGB
  * output by doing multi-point vertical scaling between input pixels.
@@ -205,12 +210,13 @@ typedef void (*yuv2packed2_fn) (struct SwsContext *c,  const int16_t *lumSrc[2],
  *                      but can be used to generate comfort noise using dithering
  *                      or some output formats.
  */
-typedef void (*yuv2packedX_fn) (struct SwsContext *c, const int16_t *lumFilter,
-                                const int16_t **lumSrc, int lumFilterSize,
-                                const int16_t *chrFilter, const int16_t **chrUSrc,
-                                const int16_t **chrVSrc, int chrFilterSize,
-                                const int16_t **alpSrc, uint8_t *dest,
-                                int dstW, int y);
+typedef void (*yuv2packedX_fn)(struct SwsContext *c, const int16_t *lumFilter,
+                               const int16_t **lumSrc, int lumFilterSize,
+                               const int16_t *chrFilter,
+                               const int16_t **chrUSrc,
+                               const int16_t **chrVSrc, int chrFilterSize,
+                               const int16_t **alpSrc, uint8_t *dest,
+                               int dstW, int y);
 
 /* This struct should be aligned on at least a 32-byte boundary. */
 typedef struct SwsContext {
@@ -263,12 +269,12 @@ typedef struct SwsContext {
     int16_t **chrUPixBuf;         ///< Ring buffer for scaled horizontal chroma plane lines to be fed to the vertical scaler.
     int16_t **chrVPixBuf;         ///< Ring buffer for scaled horizontal chroma plane lines to be fed to the vertical scaler.
     int16_t **alpPixBuf;          ///< Ring buffer for scaled horizontal alpha  plane lines to be fed to the vertical scaler.
-    int       vLumBufSize;        ///< Number of vertical luma/alpha lines allocated in the ring buffer.
-    int       vChrBufSize;        ///< Number of vertical chroma     lines allocated in the ring buffer.
-    int       lastInLumBuf;       ///< Last scaled horizontal luma/alpha line from source in the ring buffer.
-    int       lastInChrBuf;       ///< Last scaled horizontal chroma     line from source in the ring buffer.
-    int       lumBufIndex;        ///< Index in ring buffer of the last scaled horizontal luma/alpha line from source.
-    int       chrBufIndex;        ///< Index in ring buffer of the last scaled horizontal chroma     line from source.
+    int vLumBufSize;              ///< Number of vertical luma/alpha lines allocated in the ring buffer.
+    int vChrBufSize;              ///< Number of vertical chroma     lines allocated in the ring buffer.
+    int lastInLumBuf;             ///< Last scaled horizontal luma/alpha line from source in the ring buffer.
+    int lastInChrBuf;             ///< Last scaled horizontal chroma     line from source in the ring buffer.
+    int lumBufIndex;              ///< Index in ring buffer of the last scaled horizontal luma/alpha line from source.
+    int chrBufIndex;              ///< Index in ring buffer of the last scaled horizontal chroma     line from source.
     //@}
 
     uint8_t *formatConvBuffer;
@@ -295,10 +301,10 @@ typedef struct SwsContext {
     int16_t *hChrFilterPos;       ///< Array of horizontal filter starting positions for each dst[i] for chroma     planes.
     int16_t *vLumFilterPos;       ///< Array of vertical   filter starting positions for each dst[i] for luma/alpha planes.
     int16_t *vChrFilterPos;       ///< Array of vertical   filter starting positions for each dst[i] for chroma     planes.
-    int      hLumFilterSize;      ///< Horizontal filter size for luma/alpha pixels.
-    int      hChrFilterSize;      ///< Horizontal filter size for chroma     pixels.
-    int      vLumFilterSize;      ///< Vertical   filter size for luma/alpha pixels.
-    int      vChrFilterSize;      ///< Vertical   filter size for chroma     pixels.
+    int hLumFilterSize;           ///< Horizontal filter size for luma/alpha pixels.
+    int hChrFilterSize;           ///< Horizontal filter size for chroma     pixels.
+    int vLumFilterSize;           ///< Vertical   filter size for luma/alpha pixels.
+    int vChrFilterSize;           ///< Vertical   filter size for chroma     pixels.
     //@}
 
     int lumMmx2FilterCodeSize;    ///< Runtime-generated MMX2 horizontal fast bilinear scaler code size for luma/alpha planes.
@@ -310,11 +316,11 @@ typedef struct SwsContext {
 
     int dstY;                     ///< Last destination vertical line output from last slice.
     int flags;                    ///< Flags passed by the user to select scaler algorithm, optimizations, subsampling, etc...
-    void * yuvTable;            // pointer to the yuv->rgb table start so it can be freed()
-    uint8_t * table_rV[256];
-    uint8_t * table_gU[256];
-    int    table_gV[256];
-    uint8_t * table_bU[256];
+    void *yuvTable;             // pointer to the yuv->rgb table start so it can be freed()
+    uint8_t *table_rV[256];
+    uint8_t *table_gU[256];
+    int table_gV[256];
+    uint8_t *table_bU[256];
 
     //Colorspace stuff
     int contrast, brightness, saturation;    // for sws_getColorspaceDetails
@@ -366,15 +372,15 @@ typedef struct SwsContext {
     DECLARE_ALIGNED(8, uint64_t, yOffset);
     DECLARE_ALIGNED(8, uint64_t, uOffset);
     DECLARE_ALIGNED(8, uint64_t, vOffset);
-    int32_t  lumMmxFilter[4*MAX_FILTER_SIZE];
-    int32_t  chrMmxFilter[4*MAX_FILTER_SIZE];
+    int32_t lumMmxFilter[4 * MAX_FILTER_SIZE];
+    int32_t chrMmxFilter[4 * MAX_FILTER_SIZE];
     int dstW;                     ///< Width  of destination luma/alpha planes.
     DECLARE_ALIGNED(8, uint64_t, esp);
     DECLARE_ALIGNED(8, uint64_t, vRounder);
     DECLARE_ALIGNED(8, uint64_t, u_temp);
     DECLARE_ALIGNED(8, uint64_t, v_temp);
     DECLARE_ALIGNED(8, uint64_t, y_temp);
-    int32_t  alpMmxFilter[4*MAX_FILTER_SIZE];
+    int32_t alpMmxFilter[4 * MAX_FILTER_SIZE];
     // alignment of these values is not necessary, but merely here
     // to maintain the same offset across x8632 and x86-64. Once we
     // use proper offset macros in the asm, they can be removed.
@@ -393,7 +399,7 @@ typedef struct SwsContext {
     vector signed short   CGV;
     vector signed short   OY;
     vector unsigned short CSHIFT;
-    vector signed short   *vYCoeffsBank, *vCCoeffsBank;
+    vector signed short  *vYCoeffsBank, *vCCoeffsBank;
 #endif
 
 #if ARCH_BFIN
@@ -423,21 +429,25 @@ typedef struct SwsContext {
     yuv2packed2_fn yuv2packed2;
     yuv2packedX_fn yuv2packedX;
 
+    /// Unscaled conversion of luma plane to YV12 for horizontal scaler.
     void (*lumToYV12)(uint8_t *dst, const uint8_t *src, const uint8_t *src2, const uint8_t *src3,
-                      int width, uint32_t *pal); ///< Unscaled conversion of luma plane to YV12 for horizontal scaler.
+                      int width, uint32_t *pal);
+    /// Unscaled conversion of alpha plane to YV12 for horizontal scaler.
     void (*alpToYV12)(uint8_t *dst, const uint8_t *src, const uint8_t *src2, const uint8_t *src3,
-                      int width, uint32_t *pal); ///< Unscaled conversion of alpha plane to YV12 for horizontal scaler.
+                      int width, uint32_t *pal);
+    /// Unscaled conversion of chroma planes to YV12 for horizontal scaler.
     void (*chrToYV12)(uint8_t *dstU, uint8_t *dstV,
                       const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
-                      int width, uint32_t *pal); ///< Unscaled conversion of chroma planes to YV12 for horizontal scaler.
+                      int width, uint32_t *pal);
 
     /**
-      * Functions to read planar input, such as planar RGB, and convert
-      * internally to Y/UV.
-      */
+     * Functions to read planar input, such as planar RGB, and convert
+     * internally to Y/UV.
+     */
     /** @{ */
     void (*readLumPlanar)(uint8_t *dst, const uint8_t *src[4], int width);
-    void (*readChrPlanar)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src[4], int width);
+    void (*readChrPlanar)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src[4],
+                          int width);
     /** @} */
 
     /**
@@ -499,19 +509,20 @@ typedef struct SwsContext {
      *                   to simplify creating SIMD code.
      */
     /** @{ */
-    void (*hyScale)(struct SwsContext *c, int16_t *dst, int dstW, const uint8_t *src,
-                    const int16_t *filter, const int16_t *filterPos,
-                    int filterSize);
-    void (*hcScale)(struct SwsContext *c, int16_t *dst, int dstW, const uint8_t *src,
-                    const int16_t *filter, const int16_t *filterPos,
-                    int filterSize);
+    void (*hyScale)(struct SwsContext *c, int16_t *dst, int dstW,
+                    const uint8_t *src, const int16_t *filter,
+                    const int16_t *filterPos, int filterSize);
+    void (*hcScale)(struct SwsContext *c, int16_t *dst, int dstW,
+                    const uint8_t *src, const int16_t *filter,
+                    const int16_t *filterPos, int filterSize);
     /** @} */
 
-    void (*lumConvertRange)(int16_t *dst, int width); ///< Color range conversion function for luma plane if needed.
-    void (*chrConvertRange)(int16_t *dst1, int16_t *dst2, int width); ///< Color range conversion function for chroma planes if needed.
+    /// Color range conversion function for luma plane if needed.
+    void (*lumConvertRange)(int16_t *dst, int width);
+    /// Color range conversion function for chroma planes if needed.
+    void (*chrConvertRange)(int16_t *dst1, int16_t *dst2, int width);
 
     int needs_hcscale; ///< Set if there are chroma planes to be converted.
-
 } SwsContext;
 //FIXME check init (where 0)
 
@@ -567,53 +578,54 @@ const char *sws_format_name(enum PixelFormat format);
     (!(av_pix_fmt_descriptors[x].flags & PIX_FMT_PAL) && \
      av_pix_fmt_descriptors[x].nb_components <= 2)
 #else
-#define isGray(x)       (           \
-           (x)==PIX_FMT_GRAY8       \
-        || (x)==PIX_FMT_GRAY8A      \
-        || (x)==PIX_FMT_GRAY16BE    \
-        || (x)==PIX_FMT_GRAY16LE    \
-    )
+#define isGray(x)                      \
+    ((x) == PIX_FMT_GRAY8       ||     \
+     (x) == PIX_FMT_Y400A       ||     \
+     (x) == PIX_FMT_GRAY16BE    ||     \
+     (x) == PIX_FMT_GRAY16LE)
 #endif
 
-#define isRGBinInt(x)   (           \
-           (x)==PIX_FMT_RGB48BE     \
-        || (x)==PIX_FMT_RGB48LE     \
-        || (x)==PIX_FMT_RGBA64BE    \
-        || (x)==PIX_FMT_RGBA64LE    \
-        || (x)==PIX_FMT_RGB32       \
-        || (x)==PIX_FMT_RGB32_1     \
-        || (x)==PIX_FMT_RGB24       \
-        || (x)==PIX_FMT_RGB565BE    \
-        || (x)==PIX_FMT_RGB565LE    \
-        || (x)==PIX_FMT_RGB555BE    \
-        || (x)==PIX_FMT_RGB555LE    \
-        || (x)==PIX_FMT_RGB444BE    \
-        || (x)==PIX_FMT_RGB444LE    \
-        || (x)==PIX_FMT_RGB8        \
-        || (x)==PIX_FMT_RGB4        \
-        || (x)==PIX_FMT_RGB4_BYTE   \
-        || (x)==PIX_FMT_MONOBLACK   \
-        || (x)==PIX_FMT_MONOWHITE   \
+#define isRGBinInt(x) \
+    (           \
+        (x)==PIX_FMT_RGB48BE   ||   \
+        (x)==PIX_FMT_RGB48LE   ||   \
+        (x)==PIX_FMT_RGBA64BE  ||   \
+        (x)==PIX_FMT_RGBA64LE  ||   \
+        (x)==PIX_FMT_RGB32     ||   \
+        (x)==PIX_FMT_RGB32_1   ||   \
+        (x)==PIX_FMT_RGB24     ||   \
+        (x)==PIX_FMT_RGB565BE  ||   \
+        (x)==PIX_FMT_RGB565LE  ||   \
+        (x)==PIX_FMT_RGB555BE  ||   \
+        (x)==PIX_FMT_RGB555LE  ||   \
+        (x)==PIX_FMT_RGB444BE  ||   \
+        (x)==PIX_FMT_RGB444LE  ||   \
+        (x)==PIX_FMT_RGB8      ||   \
+        (x)==PIX_FMT_RGB4      ||   \
+        (x)==PIX_FMT_RGB4_BYTE ||   \
+        (x)==PIX_FMT_MONOBLACK ||   \
+        (x)==PIX_FMT_MONOWHITE   \
     )
-#define isBGRinInt(x)   (           \
-           (x)==PIX_FMT_BGR48BE     \
-        || (x)==PIX_FMT_BGR48LE     \
-        || (x)==PIX_FMT_BGRA64BE    \
-        || (x)==PIX_FMT_BGRA64LE    \
-        || (x)==PIX_FMT_BGR32       \
-        || (x)==PIX_FMT_BGR32_1     \
-        || (x)==PIX_FMT_BGR24       \
-        || (x)==PIX_FMT_BGR565BE    \
-        || (x)==PIX_FMT_BGR565LE    \
-        || (x)==PIX_FMT_BGR555BE    \
-        || (x)==PIX_FMT_BGR555LE    \
-        || (x)==PIX_FMT_BGR444BE    \
-        || (x)==PIX_FMT_BGR444LE    \
-        || (x)==PIX_FMT_BGR8        \
-        || (x)==PIX_FMT_BGR4        \
-        || (x)==PIX_FMT_BGR4_BYTE   \
-        || (x)==PIX_FMT_MONOBLACK   \
-        || (x)==PIX_FMT_MONOWHITE   \
+#define isBGRinInt(x) \
+    (           \
+         (x)==PIX_FMT_BGR48BE  ||   \
+         (x)==PIX_FMT_BGR48LE  ||   \
+         (x)==PIX_FMT_BGRA64BE ||   \
+         (x)==PIX_FMT_BGRA64LE ||   \
+         (x)==PIX_FMT_BGR32    ||   \
+         (x)==PIX_FMT_BGR32_1  ||   \
+         (x)==PIX_FMT_BGR24    ||   \
+         (x)==PIX_FMT_BGR565BE ||   \
+         (x)==PIX_FMT_BGR565LE ||   \
+         (x)==PIX_FMT_BGR555BE ||   \
+         (x)==PIX_FMT_BGR555LE ||   \
+         (x)==PIX_FMT_BGR444BE ||   \
+         (x)==PIX_FMT_BGR444LE ||   \
+         (x)==PIX_FMT_BGR8     ||   \
+         (x)==PIX_FMT_BGR4     ||   \
+         (x)==PIX_FMT_BGR4_BYTE||   \
+         (x)==PIX_FMT_MONOBLACK||   \
+         (x)==PIX_FMT_MONOWHITE   \
     )
 
 #define isRGBinBytes(x) (           \
@@ -635,10 +647,11 @@ const char *sws_format_name(enum PixelFormat format);
         || (x)==PIX_FMT_BGR24       \
     )
 
-#define isAnyRGB(x)     (           \
-            isRGBinInt(x)           \
-        ||  isBGRinInt(x)           \
-        ||  (x)==PIX_FMT_GBR24P     \
+#define isAnyRGB(x) \
+    (           \
+          isRGBinInt(x)       ||    \
+          isBGRinInt(x)       ||    \
+          (x)==PIX_FMT_GBR24P     \
     )
 
 #define isALPHA(x) \
@@ -655,15 +668,14 @@ const char *sws_format_name(enum PixelFormat format);
         ||  isBGRinInt(x)           \
     )
 #else
-#define isPacked(x) (\
-    (av_pix_fmt_descriptors[x].nb_components >= 2 && \
-     !(av_pix_fmt_descriptors[x].flags & PIX_FMT_PLANAR)) || \
-    (x) == PIX_FMT_PAL8\
-    )
+#define isPacked(x)                                            \
+    ((av_pix_fmt_descriptors[x].nb_components >= 2         &&  \
+      !(av_pix_fmt_descriptors[x].flags & PIX_FMT_PLANAR)) ||  \
+     (x) == PIX_FMT_PAL8)
 
 #endif
 #define isPlanar(x) \
-    (av_pix_fmt_descriptors[x].nb_components >= 2 && \
+    (av_pix_fmt_descriptors[x].nb_components >= 2          &&  \
      (av_pix_fmt_descriptors[x].flags & PIX_FMT_PLANAR))
 
 #define usePal(x) ((av_pix_fmt_descriptors[x].flags & PIX_FMT_PAL) || (x) == PIX_FMT_Y400A)
diff --git a/libswscale/utils.c b/libswscale/utils.c
index 7d87a13617..1bbe58e58b 100644
--- a/libswscale/utils.c
+++ b/libswscale/utils.c
@@ -129,10 +129,10 @@ const static FormatEntry format_entries[PIX_FMT_NB] = {
     [PIX_FMT_YUV422P16BE] = { 1 , 1 },
     [PIX_FMT_YUV444P16LE] = { 1 , 1 },
     [PIX_FMT_YUV444P16BE] = { 1 , 1 },
-    [PIX_FMT_RGB444LE]    = { 0 , 1 },
-    [PIX_FMT_RGB444BE]    = { 0 , 1 },
-    [PIX_FMT_BGR444LE]    = { 0 , 1 },
-    [PIX_FMT_BGR444BE]    = { 0 , 1 },
+    [PIX_FMT_RGB444LE]    = { 1 , 1 },
+    [PIX_FMT_RGB444BE]    = { 1 , 1 },
+    [PIX_FMT_BGR444LE]    = { 1 , 1 },
+    [PIX_FMT_BGR444BE]    = { 1 , 1 },
     [PIX_FMT_Y400A]       = { 1 , 0 },
     [PIX_FMT_BGR48BE]     = { 1 , 1 },
     [PIX_FMT_BGR48LE]     = { 1 , 1 },
diff --git a/libswscale/x86/input.asm b/libswscale/x86/input.asm
new file mode 100644
index 0000000000..af53dab7d6
--- /dev/null
+++ b/libswscale/x86/input.asm
@@ -0,0 +1,242 @@
+;******************************************************************************
+;* x86-optimized input routines; does shuffling of packed
+;* YUV formats into individual planes, and converts RGB
+;* into YUV planes also.
+;* Copyright (c) 2012 Ronald S. Bultje <rsbultje@gmail.com>
+;*
+;* This file is part of Libav.
+;*
+;* Libav is free software; you can redistribute it and/or
+;* modify it under the terms of the GNU Lesser General Public
+;* License as published by the Free Software Foundation; either
+;* version 2.1 of the License, or (at your option) any later version.
+;*
+;* Libav 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
+;* Lesser General Public License for more details.
+;*
+;* You should have received a copy of the GNU Lesser General Public
+;* License along with Libav; if not, write to the Free Software
+;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+;******************************************************************************
+
+%include "x86inc.asm"
+%include "x86util.asm"
+
+SECTION_RODATA
+
+SECTION .text
+
+;-----------------------------------------------------------------------------
+; YUYV/UYVY/NV12/NV21 packed pixel shuffling.
+;
+; void <fmt>ToY_<opt>(uint8_t *dst, const uint8_t *src, int w);
+; and
+; void <fmt>toUV_<opt>(uint8_t *dstU, uint8_t *dstV, const uint8_t *src,
+;                      const uint8_t *unused, int w);
+;-----------------------------------------------------------------------------
+
+; %1 = a (aligned) or u (unaligned)
+; %2 = yuyv or uyvy
+%macro LOOP_YUYV_TO_Y 2
+.loop_%1:
+    mov%1          m0, [srcq+wq*2]        ; (byte) { Y0, U0, Y1, V0, ... }
+    mov%1          m1, [srcq+wq*2+mmsize] ; (byte) { Y8, U4, Y9, V4, ... }
+%ifidn %2, yuyv
+    pand           m0, m2                 ; (word) { Y0, Y1, ..., Y7 }
+    pand           m1, m2                 ; (word) { Y8, Y9, ..., Y15 }
+%else ; uyvy
+    psrlw          m0, 8                  ; (word) { Y0, Y1, ..., Y7 }
+    psrlw          m1, 8                  ; (word) { Y8, Y9, ..., Y15 }
+%endif ; yuyv/uyvy
+    packuswb       m0, m1                 ; (byte) { Y0, ..., Y15 }
+    mova    [dstq+wq], m0
+    add            wq, mmsize
+    jl .loop_%1
+    REP_RET
+%endmacro
+
+; %1 = nr. of XMM registers
+; %2 = yuyv or uyvy
+; %3 = if specified, it means that unaligned and aligned code in loop
+;      will be the same (i.e. YUYV+AVX), and thus we don't need to
+;      split the loop in an aligned and unaligned case
+%macro YUYV_TO_Y_FN 2-3
+cglobal %2ToY, 5, 5, %1, dst, unused0, unused1, src, w
+%ifdef ARCH_X86_64
+    movsxd         wq, wd
+%endif
+    add          dstq, wq
+%if mmsize == 16
+    test         srcq, 15
+%endif
+    lea          srcq, [srcq+wq*2]
+%ifidn %2, yuyv
+    pcmpeqb        m2, m2                 ; (byte) { 0xff } x 16
+    psrlw          m2, 8                  ; (word) { 0x00ff } x 8
+%endif ; yuyv
+%if mmsize == 16
+    jnz .loop_u_start
+    neg            wq
+    LOOP_YUYV_TO_Y  a, %2
+.loop_u_start:
+    neg            wq
+    LOOP_YUYV_TO_Y  u, %2
+%else ; mmsize == 8
+    neg            wq
+    LOOP_YUYV_TO_Y  a, %2
+%endif ; mmsize == 8/16
+%endmacro
+
+; %1 = a (aligned) or u (unaligned)
+; %2 = yuyv or uyvy
+%macro LOOP_YUYV_TO_UV 2
+.loop_%1:
+%ifidn %2, yuyv
+    mov%1          m0, [srcq+wq*4]        ; (byte) { Y0, U0, Y1, V0, ... }
+    mov%1          m1, [srcq+wq*4+mmsize] ; (byte) { Y8, U4, Y9, V4, ... }
+    psrlw          m0, 8                  ; (word) { U0, V0, ..., U3, V3 }
+    psrlw          m1, 8                  ; (word) { U4, V4, ..., U7, V7 }
+%else ; uyvy
+%if cpuflag(avx)
+    vpand          m0, m2, [srcq+wq*4]        ; (word) { U0, V0, ..., U3, V3 }
+    vpand          m1, m2, [srcq+wq*4+mmsize] ; (word) { U4, V4, ..., U7, V7 }
+%else
+    mov%1          m0, [srcq+wq*4]        ; (byte) { Y0, U0, Y1, V0, ... }
+    mov%1          m1, [srcq+wq*4+mmsize] ; (byte) { Y8, U4, Y9, V4, ... }
+    pand           m0, m2                 ; (word) { U0, V0, ..., U3, V3 }
+    pand           m1, m2                 ; (word) { U4, V4, ..., U7, V7 }
+%endif
+%endif ; yuyv/uyvy
+    packuswb       m0, m1                 ; (byte) { U0, V0, ..., U7, V7 }
+    pand           m1, m0, m2             ; (word) { U0, U1, ..., U7 }
+    psrlw          m0, 8                  ; (word) { V0, V1, ..., V7 }
+%if mmsize == 16
+    packuswb       m1, m0                 ; (byte) { U0, ... U7, V1, ... V7 }
+    movh   [dstUq+wq], m1
+    movhps [dstVq+wq], m1
+%else ; mmsize == 8
+    packuswb       m1, m1                 ; (byte) { U0, ... U3 }
+    packuswb       m0, m0                 ; (byte) { V0, ... V3 }
+    movh   [dstUq+wq], m1
+    movh   [dstVq+wq], m0
+%endif ; mmsize == 8/16
+    add            wq, mmsize / 2
+    jl .loop_%1
+    REP_RET
+%endmacro
+
+; %1 = nr. of XMM registers
+; %2 = yuyv or uyvy
+; %3 = if specified, it means that unaligned and aligned code in loop
+;      will be the same (i.e. UYVY+AVX), and thus we don't need to
+;      split the loop in an aligned and unaligned case
+%macro YUYV_TO_UV_FN 2-3
+cglobal %2ToUV, 4, 5, %1, dstU, dstV, unused, src, w
+%ifdef ARCH_X86_64
+    movsxd         wq, r5m
+%else ; x86-32
+    mov            wq, r5m
+%endif
+    add         dstUq, wq
+    add         dstVq, wq
+%if mmsize == 16 && %0 == 2
+    test         srcq, 15
+%endif
+    lea          srcq, [srcq+wq*4]
+    pcmpeqb        m2, m2                 ; (byte) { 0xff } x 16
+    psrlw          m2, 8                  ; (word) { 0x00ff } x 8
+    ; NOTE: if uyvy+avx, u/a are identical
+%if mmsize == 16 && %0 == 2
+    jnz .loop_u_start
+    neg            wq
+    LOOP_YUYV_TO_UV a, %2
+.loop_u_start:
+    neg            wq
+    LOOP_YUYV_TO_UV u, %2
+%else ; mmsize == 8
+    neg            wq
+    LOOP_YUYV_TO_UV a, %2
+%endif ; mmsize == 8/16
+%endmacro
+
+; %1 = a (aligned) or u (unaligned)
+; %2 = nv12 or nv21
+%macro LOOP_NVXX_TO_UV 2
+.loop_%1:
+    mov%1          m0, [srcq+wq*2]        ; (byte) { U0, V0, U1, V1, ... }
+    mov%1          m1, [srcq+wq*2+mmsize] ; (byte) { U8, V8, U9, V9, ... }
+    pand           m2, m0, m5             ; (word) { U0, U1, ..., U7 }
+    pand           m3, m1, m5             ; (word) { U8, U9, ..., U15 }
+    psrlw          m0, 8                  ; (word) { V0, V1, ..., V7 }
+    psrlw          m1, 8                  ; (word) { V8, V9, ..., V15 }
+    packuswb       m2, m3                 ; (byte) { U0, ..., U15 }
+    packuswb       m0, m1                 ; (byte) { V0, ..., V15 }
+%ifidn %2, nv12
+    mova   [dstUq+wq], m2
+    mova   [dstVq+wq], m0
+%else ; nv21
+    mova   [dstVq+wq], m2
+    mova   [dstUq+wq], m0
+%endif ; nv12/21
+    add            wq, mmsize
+    jl .loop_%1
+    REP_RET
+%endmacro
+
+; %1 = nr. of XMM registers
+; %2 = nv12 or nv21
+%macro NVXX_TO_UV_FN 2
+cglobal %2ToUV, 4, 5, %1, dstU, dstV, unused, src, w
+%ifdef ARCH_X86_64
+    movsxd         wq, r5m
+%else ; x86-32
+    mov            wq, r5m
+%endif
+    add         dstUq, wq
+    add         dstVq, wq
+%if mmsize == 16
+    test         srcq, 15
+%endif
+    lea          srcq, [srcq+wq*2]
+    pcmpeqb        m5, m5                 ; (byte) { 0xff } x 16
+    psrlw          m5, 8                  ; (word) { 0x00ff } x 8
+%if mmsize == 16
+    jnz .loop_u_start
+    neg            wq
+    LOOP_NVXX_TO_UV a, %2
+.loop_u_start:
+    neg            wq
+    LOOP_NVXX_TO_UV u, %2
+%else ; mmsize == 8
+    neg            wq
+    LOOP_NVXX_TO_UV a, %2
+%endif ; mmsize == 8/16
+%endmacro
+
+%ifdef ARCH_X86_32
+INIT_MMX mmx
+YUYV_TO_Y_FN  0, yuyv
+YUYV_TO_Y_FN  0, uyvy
+YUYV_TO_UV_FN 0, yuyv
+YUYV_TO_UV_FN 0, uyvy
+NVXX_TO_UV_FN 0, nv12
+NVXX_TO_UV_FN 0, nv21
+%endif
+
+INIT_XMM sse2
+YUYV_TO_Y_FN  3, yuyv
+YUYV_TO_Y_FN  2, uyvy
+YUYV_TO_UV_FN 3, yuyv
+YUYV_TO_UV_FN 3, uyvy
+NVXX_TO_UV_FN 5, nv12
+NVXX_TO_UV_FN 5, nv21
+
+INIT_XMM avx
+; in theory, we could write a yuy2-to-y using vpand (i.e. AVX), but
+; that's not faster in practice
+YUYV_TO_UV_FN 3, yuyv
+YUYV_TO_UV_FN 3, uyvy, 1
+NVXX_TO_UV_FN 5, nv12
+NVXX_TO_UV_FN 5, nv21
diff --git a/libswscale/x86/swscale_mmx.c b/libswscale/x86/swscale_mmx.c
index 7cac5d80fd..66c4f69394 100644
--- a/libswscale/x86/swscale_mmx.c
+++ b/libswscale/x86/swscale_mmx.c
@@ -307,6 +307,26 @@ VSCALE_FUNCS(sse2, sse2);
 VSCALE_FUNC(16, sse4);
 VSCALE_FUNCS(avx, avx);
 
+#define INPUT_UV_FUNC(fmt, opt) \
+extern void ff_ ## fmt ## ToUV_ ## opt(uint8_t *dstU, uint8_t *dstV, \
+                                       const uint8_t *src, const uint8_t *unused1, \
+                                       int w, uint32_t *unused2)
+#define INPUT_FUNC(fmt, opt) \
+extern void ff_ ## fmt ## ToY_  ## opt(uint8_t *dst, const uint8_t *src, \
+                                       int w, uint32_t *unused); \
+    INPUT_UV_FUNC(fmt, opt)
+#define INPUT_FUNCS(opt) \
+    INPUT_FUNC(uyvy, opt); \
+    INPUT_FUNC(yuyv, opt); \
+    INPUT_UV_FUNC(nv12, opt); \
+    INPUT_UV_FUNC(nv21, opt)
+
+#if ARCH_X86_32
+INPUT_FUNCS(mmx);
+#endif
+INPUT_FUNCS(sse2);
+INPUT_FUNCS(avx);
+
 void ff_sws_init_swScale_mmx(SwsContext *c)
 {
     int cpu_flags = av_get_cpu_flags();
@@ -366,6 +386,30 @@ switch(c->dstBpc){ \
         ASSIGN_MMX_SCALE_FUNC(c->hyScale, c->hLumFilterSize, mmx, mmx);
         ASSIGN_MMX_SCALE_FUNC(c->hcScale, c->hChrFilterSize, mmx, mmx);
         ASSIGN_VSCALE_FUNC(c->yuv2plane1, mmx, mmx2, cpu_flags & AV_CPU_FLAG_MMX2);
+
+        switch (c->srcFormat) {
+        case PIX_FMT_Y400A:
+            c->lumToYV12 = ff_yuyvToY_mmx;
+            if (c->alpPixBuf)
+                c->alpToYV12 = ff_uyvyToY_mmx;
+            break;
+        case PIX_FMT_YUYV422:
+            c->lumToYV12 = ff_yuyvToY_mmx;
+            c->chrToYV12 = ff_yuyvToUV_mmx;
+            break;
+        case PIX_FMT_UYVY422:
+            c->lumToYV12 = ff_uyvyToY_mmx;
+            c->chrToYV12 = ff_uyvyToUV_mmx;
+            break;
+        case PIX_FMT_NV12:
+            c->chrToYV12 = ff_nv12ToUV_mmx;
+            break;
+        case PIX_FMT_NV21:
+            c->chrToYV12 = ff_nv21ToUV_mmx;
+            break;
+        default:
+            break;
+        }
     }
     if (cpu_flags & AV_CPU_FLAG_MMX2) {
         ASSIGN_VSCALEX_FUNC(c->yuv2planeX, mmx2,);
@@ -384,6 +428,28 @@ switch(c->dstBpc){ \
         ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, sse2, sse2);
         ASSIGN_VSCALEX_FUNC(c->yuv2planeX, sse2,);
         ASSIGN_VSCALE_FUNC(c->yuv2plane1, sse2, sse2, 1);
+
+        switch (c->srcFormat) {
+        case PIX_FMT_Y400A:
+            c->lumToYV12 = ff_yuyvToY_sse2;
+            if (c->alpPixBuf)
+                c->alpToYV12 = ff_uyvyToY_sse2;
+            break;
+        case PIX_FMT_YUYV422:
+            c->lumToYV12 = ff_yuyvToY_sse2;
+            c->chrToYV12 = ff_yuyvToUV_sse2;
+            break;
+        case PIX_FMT_UYVY422:
+            c->lumToYV12 = ff_uyvyToY_sse2;
+            c->chrToYV12 = ff_uyvyToUV_sse2;
+            break;
+        case PIX_FMT_NV12:
+            c->chrToYV12 = ff_nv12ToUV_sse2;
+            break;
+        case PIX_FMT_NV21:
+            c->chrToYV12 = ff_nv21ToUV_sse2;
+            break;
+        }
     }
     if (cpu_flags & AV_CPU_FLAG_SSSE3) {
         ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, ssse3, ssse3);
@@ -402,6 +468,23 @@ switch(c->dstBpc){ \
     if (cpu_flags & AV_CPU_FLAG_AVX) {
         ASSIGN_VSCALEX_FUNC(c->yuv2planeX, avx,);
         ASSIGN_VSCALE_FUNC(c->yuv2plane1, avx, avx, 1);
+
+        switch (c->srcFormat) {
+        case PIX_FMT_YUYV422:
+            c->chrToYV12 = ff_yuyvToUV_avx;
+            break;
+        case PIX_FMT_UYVY422:
+            c->chrToYV12 = ff_uyvyToUV_avx;
+            break;
+        case PIX_FMT_NV12:
+            c->chrToYV12 = ff_nv12ToUV_avx;
+            break;
+        case PIX_FMT_NV21:
+            c->chrToYV12 = ff_nv21ToUV_avx;
+            break;
+        default:
+            break;
+        }
     }
 #endif
 }
diff --git a/libswscale/x86/swscale_template.c b/libswscale/x86/swscale_template.c
index bb351c2394..79c63b7d47 100644
--- a/libswscale/x86/swscale_template.c
+++ b/libswscale/x86/swscale_template.c
@@ -1435,147 +1435,6 @@ static void RENAME(yuv2yuyv422_1)(SwsContext *c, const int16_t *buf0,
     }
 }
 
-#if !COMPILE_TEMPLATE_MMX2
-//FIXME yuy2* can read up to 7 samples too much
-
-static void RENAME(yuy2ToY)(uint8_t *dst, const uint8_t *src, const uint8_t *unused1, const uint8_t *unused2,
-                            int width, uint32_t *unused)
-{
-    __asm__ volatile(
-        "movq "MANGLE(bm01010101)", %%mm2           \n\t"
-        "mov                    %0, %%"REG_a"       \n\t"
-        "1:                                         \n\t"
-        "movq    (%1, %%"REG_a",2), %%mm0           \n\t"
-        "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"
-        "pand                %%mm2, %%mm0           \n\t"
-        "pand                %%mm2, %%mm1           \n\t"
-        "packuswb            %%mm1, %%mm0           \n\t"
-        "movq                %%mm0, (%2, %%"REG_a") \n\t"
-        "add                    $8, %%"REG_a"       \n\t"
-        " js                    1b                  \n\t"
-        : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst+width)
-        : "%"REG_a
-    );
-}
-
-static void RENAME(yuy2ToUV)(uint8_t *dstU, uint8_t *dstV,
-                             const uint8_t *unused1, const uint8_t *src1, const uint8_t *src2,
-                             int width, uint32_t *unused)
-{
-    __asm__ volatile(
-        "movq "MANGLE(bm01010101)", %%mm4           \n\t"
-        "mov                    %0, %%"REG_a"       \n\t"
-        "1:                                         \n\t"
-        "movq    (%1, %%"REG_a",4), %%mm0           \n\t"
-        "movq   8(%1, %%"REG_a",4), %%mm1           \n\t"
-        "psrlw                  $8, %%mm0           \n\t"
-        "psrlw                  $8, %%mm1           \n\t"
-        "packuswb            %%mm1, %%mm0           \n\t"
-        "movq                %%mm0, %%mm1           \n\t"
-        "psrlw                  $8, %%mm0           \n\t"
-        "pand                %%mm4, %%mm1           \n\t"
-        "packuswb            %%mm0, %%mm0           \n\t"
-        "packuswb            %%mm1, %%mm1           \n\t"
-        "movd                %%mm0, (%3, %%"REG_a") \n\t"
-        "movd                %%mm1, (%2, %%"REG_a") \n\t"
-        "add                    $4, %%"REG_a"       \n\t"
-        " js                    1b                  \n\t"
-        : : "g" ((x86_reg)-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width)
-        : "%"REG_a
-    );
-    assert(src1 == src2);
-}
-
-/* This is almost identical to the previous, end exists only because
- * yuy2ToY/UV)(dst, src+1, ...) would have 100% unaligned accesses. */
-static void RENAME(uyvyToY)(uint8_t *dst, const uint8_t *src, const uint8_t *unused1, const uint8_t *unused2,
-                            int width, uint32_t *unused)
-{
-    __asm__ volatile(
-        "mov                  %0, %%"REG_a"         \n\t"
-        "1:                                         \n\t"
-        "movq  (%1, %%"REG_a",2), %%mm0             \n\t"
-        "movq 8(%1, %%"REG_a",2), %%mm1             \n\t"
-        "psrlw                $8, %%mm0             \n\t"
-        "psrlw                $8, %%mm1             \n\t"
-        "packuswb          %%mm1, %%mm0             \n\t"
-        "movq              %%mm0, (%2, %%"REG_a")   \n\t"
-        "add                  $8, %%"REG_a"         \n\t"
-        " js                  1b                    \n\t"
-        : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst+width)
-        : "%"REG_a
-    );
-}
-
-static void RENAME(uyvyToUV)(uint8_t *dstU, uint8_t *dstV,
-                             const uint8_t *unused1, const uint8_t *src1, const uint8_t *src2,
-                             int width, uint32_t *unused)
-{
-    __asm__ volatile(
-        "movq "MANGLE(bm01010101)", %%mm4           \n\t"
-        "mov                    %0, %%"REG_a"       \n\t"
-        "1:                                         \n\t"
-        "movq    (%1, %%"REG_a",4), %%mm0           \n\t"
-        "movq   8(%1, %%"REG_a",4), %%mm1           \n\t"
-        "pand                %%mm4, %%mm0           \n\t"
-        "pand                %%mm4, %%mm1           \n\t"
-        "packuswb            %%mm1, %%mm0           \n\t"
-        "movq                %%mm0, %%mm1           \n\t"
-        "psrlw                  $8, %%mm0           \n\t"
-        "pand                %%mm4, %%mm1           \n\t"
-        "packuswb            %%mm0, %%mm0           \n\t"
-        "packuswb            %%mm1, %%mm1           \n\t"
-        "movd                %%mm0, (%3, %%"REG_a") \n\t"
-        "movd                %%mm1, (%2, %%"REG_a") \n\t"
-        "add                    $4, %%"REG_a"       \n\t"
-        " js                    1b                  \n\t"
-        : : "g" ((x86_reg)-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width)
-        : "%"REG_a
-    );
-    assert(src1 == src2);
-}
-
-static av_always_inline void RENAME(nvXXtoUV)(uint8_t *dst1, uint8_t *dst2,
-                                              const uint8_t *src, int width)
-{
-    __asm__ volatile(
-        "movq "MANGLE(bm01010101)", %%mm4           \n\t"
-        "mov                    %0, %%"REG_a"       \n\t"
-        "1:                                         \n\t"
-        "movq    (%1, %%"REG_a",2), %%mm0           \n\t"
-        "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"
-        "movq                %%mm0, %%mm2           \n\t"
-        "movq                %%mm1, %%mm3           \n\t"
-        "pand                %%mm4, %%mm0           \n\t"
-        "pand                %%mm4, %%mm1           \n\t"
-        "psrlw                  $8, %%mm2           \n\t"
-        "psrlw                  $8, %%mm3           \n\t"
-        "packuswb            %%mm1, %%mm0           \n\t"
-        "packuswb            %%mm3, %%mm2           \n\t"
-        "movq                %%mm0, (%2, %%"REG_a") \n\t"
-        "movq                %%mm2, (%3, %%"REG_a") \n\t"
-        "add                    $8, %%"REG_a"       \n\t"
-        " js                    1b                  \n\t"
-        : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst1+width), "r" (dst2+width)
-        : "%"REG_a
-    );
-}
-
-static void RENAME(nv12ToUV)(uint8_t *dstU, uint8_t *dstV,
-                             const uint8_t *unused1, const uint8_t *src1, const uint8_t *src2,
-                             int width, uint32_t *unused)
-{
-    RENAME(nvXXtoUV)(dstU, dstV, src1, width);
-}
-
-static void RENAME(nv21ToUV)(uint8_t *dstU, uint8_t *dstV,
-                             const uint8_t *unused1, const uint8_t *src1, const uint8_t *src2,
-                             int width, uint32_t *unused)
-{
-    RENAME(nvXXtoUV)(dstV, dstU, src1, width);
-}
-#endif /* !COMPILE_TEMPLATE_MMX2 */
-
 static av_always_inline void RENAME(bgr24ToY_mmx)(int16_t *dst, const uint8_t *src,
                                                   int width, enum PixelFormat srcFormat)
 {
@@ -1927,15 +1786,6 @@ static av_cold void RENAME(sws_init_swScale)(SwsContext *c)
 #endif /* COMPILE_TEMPLATE_MMX2 */
     }
 
-#if !COMPILE_TEMPLATE_MMX2
-    switch(srcFormat) {
-        case PIX_FMT_YUYV422  : c->chrToYV12 = RENAME(yuy2ToUV); break;
-        case PIX_FMT_UYVY422  : c->chrToYV12 = RENAME(uyvyToUV); break;
-        case PIX_FMT_NV12     : c->chrToYV12 = RENAME(nv12ToUV); break;
-        case PIX_FMT_NV21     : c->chrToYV12 = RENAME(nv21ToUV); break;
-        default: break;
-    }
-#endif /* !COMPILE_TEMPLATE_MMX2 */
     if (!c->chrSrcHSubSample) {
         switch(srcFormat) {
         case PIX_FMT_BGR24  : c->chrToYV12 = RENAME(bgr24ToUV); break;
@@ -1945,21 +1795,8 @@ static av_cold void RENAME(sws_init_swScale)(SwsContext *c)
     }
 
     switch (srcFormat) {
-#if !COMPILE_TEMPLATE_MMX2
-    case PIX_FMT_YUYV422  :
-    case PIX_FMT_Y400A    : c->lumToYV12 = RENAME(yuy2ToY); break;
-    case PIX_FMT_UYVY422  : c->lumToYV12 = RENAME(uyvyToY); break;
-#endif /* !COMPILE_TEMPLATE_MMX2 */
     case PIX_FMT_BGR24    : c->lumToYV12 = RENAME(bgr24ToY); break;
     case PIX_FMT_RGB24    : c->lumToYV12 = RENAME(rgb24ToY); break;
     default: break;
     }
-#if !COMPILE_TEMPLATE_MMX2
-    if (c->alpPixBuf) {
-        switch (srcFormat) {
-        case PIX_FMT_Y400A  : c->alpToYV12 = RENAME(yuy2ToY); break;
-        default: break;
-        }
-    }
-#endif /* !COMPILE_TEMPLATE_MMX2 */
 }
diff --git a/tests/fate/aac.mak b/tests/fate/aac.mak
index f17c914d1a..ea87436698 100644
--- a/tests/fate/aac.mak
+++ b/tests/fate/aac.mak
@@ -14,18 +14,34 @@ FATE_AAC += fate-aac-al07_96
 fate-aac-al07_96: CMD = pcm -i $(SAMPLES)/aac/al07_96.mp4
 fate-aac-al07_96: REF = $(SAMPLES)/aac/al07_96.s16
 
+FATE_AAC += fate-aac-al15_44
+fate-aac-al15_44: CMD = pcm -i $(SAMPLES)/aac/al15_44.mp4
+fate-aac-al15_44: REF = $(SAMPLES)/aac/al15_44.s16
+
 FATE_AAC += fate-aac-al17_44
 fate-aac-al17_44: CMD = pcm -i $(SAMPLES)/aac/al17_44.mp4
 fate-aac-al17_44: REF = $(SAMPLES)/aac/al17_44.s16
 
+FATE_AAC += fate-aac-al18_44
+fate-aac-al18_44: CMD = pcm -i $(SAMPLES)/aac/al18_44.mp4
+fate-aac-al18_44: REF = $(SAMPLES)/aac/al18_44.s16
+
 FATE_AAC += fate-aac-am00_88
 fate-aac-am00_88: CMD = pcm -i $(SAMPLES)/aac/am00_88.mp4
 fate-aac-am00_88: REF = $(SAMPLES)/aac/am00_88.s16
 
+FATE_AAC += fate-aac-am05_44
+fate-aac-am05_44: CMD = pcm -i $(SAMPLES)/aac/am05_44.mp4
+fate-aac-am05_44: REF = $(SAMPLES)/aac/am05_44.s16
+
 FATE_AAC += fate-aac-al_sbr_hq_cm_48_2
 fate-aac-al_sbr_hq_cm_48_2: CMD = pcm -i $(SAMPLES)/aac/al_sbr_cm_48_2.mp4
 fate-aac-al_sbr_hq_cm_48_2: REF = $(SAMPLES)/aac/al_sbr_hq_cm_48_2.s16
 
+FATE_AAC += fate-aac-al_sbr_hq_cm_48_5.1
+fate-aac-al_sbr_hq_cm_48_5.1: CMD = pcm -i $(SAMPLES)/aac/al_sbr_cm_48_5.1.mp4
+fate-aac-al_sbr_hq_cm_48_5.1: REF = $(SAMPLES)/aac/al_sbr_hq_cm_48_5.1.s16
+
 FATE_AAC += fate-aac-al_sbr_ps_06_ur
 fate-aac-al_sbr_ps_06_ur: CMD = pcm -i $(SAMPLES)/aac/al_sbr_ps_06_new.mp4
 fate-aac-al_sbr_ps_06_ur: REF = $(SAMPLES)/aac/al_sbr_ps_06_ur.s16
diff --git a/tests/fate/demux.mak b/tests/fate/demux.mak
index 269c155557..415af59e20 100644
--- a/tests/fate/demux.mak
+++ b/tests/fate/demux.mak
@@ -79,11 +79,20 @@ fate-sierra-vmd: CMD = framecrc -i $(SAMPLES)/vmd/12.vmd -vsync 0 -pix_fmt rgb24
 FATE_DEMUX += fate-siff
 fate-siff: CMD = framecrc -i $(SAMPLES)/SIFF/INTRO_B.VB -t 3 -pix_fmt rgb24
 
+FATE_DEMUX += fate-smjpeg
+fate-smjpeg: CMD = framecrc -i $(SAMPLES)/smjpeg/scenwin.mjpg -vcodec copy
+
 FATE_DEMUX += fate-westwood-aud
 fate-westwood-aud: CMD = md5 -i $(SAMPLES)/westwood-aud/excellent.aud -f s16le
 
+FATE_DEMUX += fate-wtv-demux
+fate-wtv-demux: CMD = framecrc -i $(SAMPLES)/wtv/law-and-order-partial.wtv -vcodec copy -acodec copy
+
 FATE_DEMUX += fate-xmv-demux
 fate-xmv-demux: CMD = framecrc -i $(SAMPLES)/xmv/logos1p.fmv -vcodec copy -acodec copy
 
+FATE_DEMUX += fate-xwma-demux
+fate-xwma-demux: CMD = crc -i $(SAMPLES)/xwma/ergon.xwma -acodec copy
+
 FATE_TESTS += $(FATE_DEMUX)
 fate-demux: $(FATE_DEMUX)
diff --git a/tests/fate/microsoft.mak b/tests/fate/microsoft.mak
index 2499ec2bad..df1013b8ad 100644
--- a/tests/fate/microsoft.mak
+++ b/tests/fate/microsoft.mak
@@ -29,5 +29,8 @@ fate-vc1_sa10091: CMD = framecrc -i $(SAMPLES)/vc1/SA10091.vc1
 FATE_MICROSOFT += fate-vc1_sa20021
 fate-vc1_sa20021: CMD = framecrc -i $(SAMPLES)/vc1/SA20021.vc1
 
+FATE_MICROSOFT += fate-vc1-ism
+fate-vc1-ism: CMD = framecrc -i $(SAMPLES)/isom/vc1-wmapro.ism -an
+
 FATE_TESTS += $(FATE_MICROSOFT)
 fate-microsoft: $(FATE_MICROSOFT)
diff --git a/tests/fate/video.mak b/tests/fate/video.mak
index 59317381c7..bce941d6bd 100644
--- a/tests/fate/video.mak
+++ b/tests/fate/video.mak
@@ -49,6 +49,9 @@ fate-cvid: CMD = framecrc -i $(SAMPLES)/cvid/laracroft-cinepak-partial.avi -an
 FATE_VIDEO += fate-cvid-palette
 fate-cvid-palette: CMD = framecrc -i $(SAMPLES)/cvid/catfight-cvid-pal8-partial.mov -pix_fmt rgb24 -an
 
+FATE_VIDEO += fate-cvid-grayscale
+fate-cvid-grayscale: CMD = framecrc -i $(SAMPLES)/cvid/pcitva15.avi -an
+
 FATE_VIDEO += fate-cyberia-c93
 fate-cyberia-c93: CMD = framecrc -i $(SAMPLES)/cyberia-c93/intro1.c93 -t 3 -pix_fmt rgb24
 
diff --git a/tests/fate/wma.mak b/tests/fate/wma.mak
index ca4b8243ae..c57ffe54c6 100644
--- a/tests/fate/wma.mak
+++ b/tests/fate/wma.mak
@@ -8,6 +8,11 @@ fate-wmapro-5.1: CMD = pcm -i $(SAMPLES)/wmapro/latin_192_mulitchannel_cut.wma
 fate-wmapro-5.1: CMP = oneoff
 fate-wmapro-5.1: REF = $(SAMPLES)/wmapro/latin_192_mulitchannel_cut.pcm
 
+FATE_WMA += fate-wmapro-ism
+fate-wmapro-ism: CMD = pcm -i $(SAMPLES)/isom/vc1-wmapro.ism -vn
+fate-wmapro-ism: CMP = oneoff
+fate-wmapro-ism: REF = $(SAMPLES)/isom/vc1-wmapro.pcm
+
 FATE_WMA += fate-wmavoice-7k
 fate-wmavoice-7k: CMD = pcm -i $(SAMPLES)/wmavoice/streaming_CBR-7K.wma
 fate-wmavoice-7k: CMP = stddev
diff --git a/tests/ref/fate/cvid-grayscale b/tests/ref/fate/cvid-grayscale
new file mode 100644
index 0000000000..4e3957156f
--- /dev/null
+++ b/tests/ref/fate/cvid-grayscale
@@ -0,0 +1,152 @@
+0, 0, 11300, 0x46c78923
+0, 17921, 11300, 0x3f2a1175
+0, 35842, 11300, 0x722de221
+0, 53763, 11300, 0x01746b88
+0, 71684, 11300, 0x549587a7
+0, 89605, 11300, 0x843ab943
+0, 107526, 11300, 0x62fdee48
+0, 125447, 11300, 0x74a62867
+0, 143368, 11300, 0x35a20e2f
+0, 161289, 11300, 0x4e9ef54d
+0, 179210, 11300, 0xec7201f5
+0, 197131, 11300, 0x363bfe27
+0, 215052, 11300, 0x2aaab418
+0, 232973, 11300, 0x6a48ab3f
+0, 250894, 11300, 0x3fecea34
+0, 268815, 11300, 0xa371f55e
+0, 286736, 11300, 0xa86b147c
+0, 304657, 11300, 0x49e9206e
+0, 322578, 11300, 0x6c9a2155
+0, 340499, 11300, 0x2c8a4798
+0, 358420, 11300, 0x3485676c
+0, 376341, 11300, 0xb0b293f2
+0, 394262, 11300, 0xe4a9b068
+0, 412183, 11300, 0xd68d0556
+0, 430104, 11300, 0xc28e5193
+0, 448025, 11300, 0xf6948483
+0, 465945, 11300, 0xf21fbf57
+0, 483866, 11300, 0x8345eb44
+0, 501787, 11300, 0x8124f045
+0, 519708, 11300, 0x18e31f10
+0, 537629, 11300, 0xdb1943fc
+0, 555550, 11300, 0x8701699f
+0, 573471, 11300, 0xd7b18550
+0, 591392, 11300, 0xa56faccc
+0, 609313, 11300, 0xf8bcc17c
+0, 627234, 11300, 0x446acab9
+0, 645155, 11300, 0x755fd295
+0, 663076, 11300, 0x92e3d100
+0, 680997, 11300, 0x54895bb3
+0, 698918, 11300, 0xd18bffda
+0, 716839, 11300, 0x480dbe4f
+0, 734760, 11300, 0x49ea9dbe
+0, 752681, 11300, 0x00d3a003
+0, 770602, 11300, 0xda7bbfb2
+0, 788523, 11300, 0x9700d9c2
+0, 806444, 11300, 0xa0a9e490
+0, 824365, 11300, 0x00eb0979
+0, 842286, 11300, 0x32b04630
+0, 860207, 11300, 0xdfb73e51
+0, 878128, 11300, 0x3d8e4f96
+0, 896049, 11300, 0x2ca83271
+0, 913970, 11300, 0xb5b123c0
+0, 931891, 11300, 0x8a570e58
+0, 949812, 11300, 0xc6c805bc
+0, 967733, 11300, 0x27caf7a5
+0, 985654, 11300, 0x5319ecb0
+0, 1003575, 11300, 0x5471e3fd
+0, 1021496, 11300, 0x6d68a6f4
+0, 1039417, 11300, 0x872b7194
+0, 1057338, 11300, 0x007c36bd
+0, 1075259, 11300, 0x2714f1b5
+0, 1093180, 11300, 0x6c8eb50f
+0, 1111101, 11300, 0xf5d57be8
+0, 1129022, 11300, 0x981f412b
+0, 1146943, 11300, 0x1a9804a1
+0, 1164864, 11300, 0xf0c1d24a
+0, 1182785, 11300, 0xa70a9d9b
+0, 1200706, 11300, 0x8c466876
+0, 1218627, 11300, 0xcf2e32df
+0, 1236548, 11300, 0xcb8cfebf
+0, 1254469, 11300, 0xb961ca99
+0, 1272390, 11300, 0x666d9619
+0, 1290311, 11300, 0x84bf5b55
+0, 1308232, 11300, 0xbfa22ccc
+0, 1326153, 11300, 0xcde41849
+0, 1344074, 11300, 0x71372dcd
+0, 1361994, 11300, 0x13402cfd
+0, 1379915, 11300, 0xdebdd321
+0, 1397836, 11300, 0xdda66de1
+0, 1415757, 11300, 0x7f4bb682
+0, 1433678, 11300, 0xf67fd528
+0, 1451599, 11300, 0xe739ff8c
+0, 1469520, 11300, 0x2e131774
+0, 1487441, 11300, 0xfa942811
+0, 1505362, 11300, 0x0cd93ac2
+0, 1523283, 11300, 0xd0445e0e
+0, 1541204, 11300, 0x3f3497c7
+0, 1559125, 11300, 0x11b5bd2c
+0, 1577046, 11300, 0xccd5e62a
+0, 1594967, 11300, 0xa9d4fcb5
+0, 1612888, 11300, 0x34aa1a03
+0, 1630809, 11300, 0x1ce6299e
+0, 1648730, 11300, 0x661c2745
+0, 1666651, 11300, 0x27d8a8b3
+0, 1684572, 11300, 0x9eb07467
+0, 1702493, 11300, 0x128374d2
+0, 1720414, 11300, 0x05c36ff5
+0, 1738335, 11300, 0x8a136bde
+0, 1756256, 11300, 0x15c47c99
+0, 1774177, 11300, 0xcc4a93f4
+0, 1792098, 11300, 0x19529b2b
+0, 1810019, 11300, 0x9943c076
+0, 1827940, 11300, 0xf898e583
+0, 1845861, 11300, 0x40f71f94
+0, 1863782, 11300, 0x5b604afb
+0, 1881703, 11300, 0x8c176af4
+0, 1899624, 11300, 0x0f1a6216
+0, 1917545, 11300, 0x38bbd13d
+0, 1935466, 11300, 0x90c8d1fc
+0, 1953387, 11300, 0x253000d7
+0, 1971308, 11300, 0xb94b03b1
+0, 1989229, 11300, 0xbc872268
+0, 2007150, 11300, 0xe77adb8c
+0, 2025071, 11300, 0xa38936b7
+0, 2042992, 11300, 0xd6153632
+0, 2060913, 11300, 0x1ae633cc
+0, 2078834, 11300, 0xb90c286e
+0, 2096755, 11300, 0xbc7e333d
+0, 2114676, 11300, 0x1b5421f8
+0, 2132597, 11300, 0xdde6506d
+0, 2150518, 11300, 0xd3eb757e
+0, 2168439, 11300, 0x5ad1929c
+0, 2186360, 11300, 0x4f6aa47d
+0, 2204281, 11300, 0xab3caf55
+0, 2222202, 11300, 0x5ff9b39a
+0, 2240123, 11300, 0x1454e12e
+0, 2258043, 11300, 0xf18216e8
+0, 2275964, 11300, 0x62144880
+0, 2293885, 11300, 0x54284241
+0, 2311806, 11300, 0x8e8c7228
+0, 2329727, 11300, 0xb498d06e
+0, 2347648, 11300, 0x7b1e6be1
+0, 2365569, 11300, 0x5e5ea1f4
+0, 2383490, 11300, 0x41eda28e
+0, 2401411, 11300, 0x7ba6aa92
+0, 2419332, 11300, 0xa8a8b1c7
+0, 2437253, 11300, 0x0d30bd08
+0, 2455174, 11300, 0xc610bf16
+0, 2473095, 11300, 0xed57c075
+0, 2491016, 11300, 0xb86dbfea
+0, 2508937, 11300, 0x0970c03d
+0, 2526858, 11300, 0x743ac2ac
+0, 2544779, 11300, 0x0a44c816
+0, 2562700, 11300, 0xe32acd6b
+0, 2580621, 11300, 0x209bcdab
+0, 2598542, 11300, 0x3cd0d105
+0, 2616463, 11300, 0xc0bcd330
+0, 2634384, 11300, 0x4785d6dc
+0, 2652305, 11300, 0xe85f9c90
+0, 2670226, 11300, 0xd4a72850
+0, 2688147, 11300, 0x04766e41
+0, 2706068, 11300, 0x04766e41
diff --git a/tests/ref/fate/smjpeg b/tests/ref/fate/smjpeg
new file mode 100644
index 0000000000..ec91959e7c
--- /dev/null
+++ b/tests/ref/fate/smjpeg
@@ -0,0 +1,423 @@
+0, 0, 734, 0x5a042c2c
+1, 0, 1024, 0x00000000
+1, 2090, 1024, 0x00000000
+1, 4180, 1024, 0xd89a448e
+1, 6269, 1024, 0x695b369c
+1, 8359, 1024, 0xc8ba5707
+0, 9990, 763, 0xb5893f2f
+1, 10449, 1024, 0xdf241fc6
+1, 12539, 1024, 0x61cf4166
+1, 14629, 1024, 0x97cbc386
+1, 16718, 1024, 0x44899d04
+1, 18808, 1024, 0xa7cbaa62
+0, 19980, 3023, 0x0f3907d3
+1, 20898, 1024, 0xa7aea60c
+1, 22988, 1024, 0xd7b18a89
+1, 25078, 1024, 0x268e81f6
+1, 27167, 1024, 0x9cf83a2f
+1, 29257, 1024, 0x5559b508
+0, 29970, 4800, 0x22e6e18a
+1, 31347, 1024, 0xe1b9e71c
+1, 33437, 1024, 0xdcee733e
+1, 35527, 1024, 0xe5918f60
+1, 37616, 1024, 0x29dbd209
+1, 39706, 1024, 0x9bcbcf16
+0, 39960, 6417, 0x427adde5
+1, 41796, 1024, 0x86f5f458
+1, 43886, 1024, 0xabcbda86
+1, 45976, 1024, 0xc51f77b9
+1, 48065, 1024, 0xf6b3a504
+0, 49950, 6776, 0x7a74c6ad
+1, 50155, 1024, 0x1af3e40e
+1, 52245, 1024, 0x3866b03b
+1, 54335, 1024, 0xbc005403
+1, 56424, 1024, 0xe9dfcc51
+1, 58514, 1024, 0x83c837cb
+0, 59940, 6808, 0x1f6eb7c3
+1, 60604, 1024, 0xfa649580
+1, 62694, 1024, 0x519452ea
+1, 64784, 1024, 0xd4978774
+1, 66873, 1024, 0xe2a3b1cd
+1, 68963, 1024, 0x9a9472ad
+0, 69930, 6726, 0x452087e6
+1, 71053, 1024, 0xa12d4060
+1, 73143, 1024, 0x31fb0646
+1, 75233, 1024, 0xfc44343f
+1, 77322, 1024, 0x0847751a
+1, 79412, 1024, 0x227968a2
+0, 79920, 6829, 0xee82b109
+1, 81502, 1024, 0x7cce9f1c
+1, 83592, 1024, 0xb8356713
+1, 85682, 1024, 0xb29f6e6f
+1, 87771, 1024, 0x9e1430ab
+1, 89861, 1024, 0x26d85423
+0, 89910, 7055, 0xf41f1108
+1, 91951, 1024, 0x6496547d
+1, 94041, 1024, 0x316b1a86
+1, 96131, 1024, 0x3cd83afc
+1, 98220, 1024, 0x993ff633
+0, 99990, 6977, 0xf8fe1ede
+1, 100310, 1024, 0x0708d1a2
+1, 102400, 1024, 0xd7230db9
+1, 104490, 1024, 0xbb0779ca
+1, 106580, 1024, 0xc6094e1b
+1, 108669, 1024, 0x15a8b039
+0, 109980, 6942, 0x9ad105c6
+1, 110759, 1024, 0xd6dbe88c
+1, 112849, 1024, 0x7e8d1140
+1, 114939, 1024, 0xef88e525
+1, 117029, 1024, 0x44e21149
+1, 119118, 1024, 0x65b0f5f4
+0, 119970, 6926, 0xe239dad6
+1, 121208, 1024, 0xb955f687
+1, 123298, 1024, 0xc85fba9c
+1, 125388, 1024, 0xf59655ad
+1, 127478, 1024, 0x6de80bf1
+1, 129567, 1024, 0x2dcf6e41
+0, 129960, 6966, 0x81dcfab1
+1, 131657, 1024, 0xd0ddcf8a
+1, 133747, 1024, 0x00135c2d
+1, 135837, 1024, 0x697f8efd
+1, 137927, 1024, 0x7a9bada5
+0, 139950, 6896, 0x31e6cc02
+1, 140016, 1024, 0x0d22783c
+1, 142106, 1024, 0x7726d07d
+1, 144196, 1024, 0xa2f14f67
+1, 146286, 1024, 0x7f51060d
+1, 148376, 1024, 0xc4ec6aea
+0, 149940, 6889, 0x1cc1006e
+1, 150465, 1024, 0x9bb37ca4
+1, 152555, 1024, 0x9b085577
+1, 154645, 1024, 0x8812f8af
+1, 156735, 1024, 0x788f5221
+1, 158824, 1024, 0x3a2ce642
+0, 159930, 6933, 0xc303f87f
+1, 160914, 1024, 0x72415692
+1, 163004, 1024, 0xe3dcc105
+1, 165094, 1024, 0xb26c0599
+1, 167184, 1024, 0x5c9e55eb
+1, 169273, 1024, 0x8fe88707
+0, 169920, 7034, 0xb4970a20
+1, 171363, 1024, 0xc5d7beb6
+1, 173453, 1024, 0xe1d3a3b4
+1, 175543, 1024, 0x012da0c6
+1, 177633, 1024, 0x8d010922
+1, 179722, 1024, 0x3366eb0d
+0, 179910, 6961, 0xf064095d
+1, 181812, 1024, 0xc9381a27
+1, 183902, 1024, 0x0774f685
+1, 185992, 1024, 0xc5cae0a5
+1, 188082, 1024, 0xa6f4737c
+0, 189990, 7089, 0x5ba350f9
+1, 190171, 1024, 0x8fb6d0d1
+1, 192261, 1024, 0x05f579c2
+1, 194351, 1024, 0x56905d99
+1, 196441, 1024, 0x002ee18d
+1, 198531, 1024, 0xeb37ef51
+0, 199980, 7078, 0xa83f3e88
+1, 200620, 1024, 0x38025635
+1, 202710, 1024, 0x4fe643c8
+1, 204800, 1024, 0x11d66ab1
+1, 206890, 1024, 0xcc3051e9
+1, 208980, 1024, 0xcd93e854
+0, 209970, 7147, 0xcda66cfc
+1, 211069, 1024, 0x38f1196d
+1, 213159, 1024, 0x657a15fc
+1, 215249, 1024, 0x669ce2a9
+1, 217339, 1024, 0x95862dda
+1, 219429, 1024, 0x1726a7b2
+0, 219960, 7173, 0xb7455859
+1, 221518, 1024, 0xd6ece2a1
+1, 223608, 1024, 0x33ab9553
+1, 225698, 1024, 0xd50c73a6
+1, 227788, 1024, 0xfe25b63a
+1, 229878, 1024, 0x7e2959e3
+0, 229950, 7213, 0x97b89994
+1, 231967, 1024, 0xa4c07b34
+1, 234057, 1024, 0xd6d8f15c
+1, 236147, 1024, 0x1eccddd7
+1, 238237, 1024, 0x2b69f9cb
+0, 239940, 7170, 0xca8b2948
+1, 240327, 1024, 0x667b775f
+1, 242416, 1024, 0xad3b84e9
+1, 244506, 1024, 0x4f29fc67
+1, 246596, 1024, 0x8d611ab7
+1, 248686, 1024, 0x278966ea
+0, 249930, 7174, 0xc7cc6bbb
+1, 250776, 1024, 0xaf33812b
+1, 252865, 1024, 0xa55f4265
+1, 254955, 1024, 0x023cb51c
+1, 257045, 1024, 0x1d1f1005
+1, 259135, 1024, 0x874cccf7
+0, 259920, 7235, 0xc2e68d2b
+1, 261224, 1024, 0xda705428
+1, 263314, 1024, 0x48d9b440
+1, 265404, 1024, 0xa14e0712
+1, 267494, 1024, 0x7efbad1f
+1, 269584, 1024, 0xdb82c17f
+0, 270000, 7261, 0x8204a423
+1, 271673, 1024, 0xcbe87613
+1, 273763, 1024, 0x3a63df1d
+1, 275853, 1024, 0xd5636bba
+1, 277943, 1024, 0x9397af23
+0, 279990, 7353, 0xacc7e7c0
+1, 280033, 1024, 0x32a07c98
+1, 282122, 1024, 0x202ca667
+1, 284212, 1024, 0xdf969011
+1, 286302, 1024, 0xc434d238
+1, 288392, 1024, 0xe9ad7562
+0, 289980, 7065, 0x45035c5c
+1, 290482, 1024, 0xb51b6b50
+1, 292571, 1024, 0xe70aecd3
+1, 294661, 1024, 0x03c816b2
+1, 296751, 1024, 0x869fdf25
+1, 298841, 1024, 0xd40a0a62
+0, 299970, 7269, 0x72edbb76
+1, 300931, 1024, 0x5af7dd35
+1, 303020, 1024, 0x891ffc72
+1, 305110, 1024, 0x1ff68a08
+1, 307200, 1024, 0x5a7517a9
+1, 309290, 1024, 0x0f959f74
+0, 309960, 7220, 0xb926772f
+1, 311380, 1024, 0xe92a12a2
+1, 313469, 1024, 0x38000e55
+1, 315559, 1024, 0x39fbdd70
+1, 317649, 1024, 0xca3d9184
+1, 319739, 1024, 0x66c8995b
+0, 319950, 7326, 0x0a66c632
+1, 321829, 1024, 0xac25acea
+1, 323918, 1024, 0x3cd1046c
+1, 326008, 1024, 0x6a1df31c
+1, 328098, 1024, 0x21ca10a1
+0, 329940, 7225, 0xe39076ab
+1, 330188, 1024, 0x1aeccedc
+1, 332278, 1024, 0xddea1335
+1, 334367, 1024, 0x19f5ca9f
+1, 336457, 1024, 0x88e95e43
+1, 338547, 1024, 0x726284fe
+0, 339930, 7265, 0xe0209036
+1, 340637, 1024, 0x6b85b40e
+1, 342727, 1024, 0x111fee2a
+1, 344816, 1024, 0x3656b588
+1, 346906, 1024, 0xa5a2b552
+1, 348996, 1024, 0x38fb2467
+0, 349920, 7337, 0x7a5dc093
+1, 351086, 1024, 0xaa919ccc
+1, 353176, 1024, 0x15993dbc
+1, 355265, 1024, 0xbe01a7b9
+1, 357355, 1024, 0xefe93c09
+1, 359445, 1024, 0x1bb566e5
+0, 360000, 7246, 0x519a7a3c
+1, 361535, 1024, 0x15ce6237
+1, 363624, 1024, 0xa8552e66
+1, 365714, 1024, 0x9d80187e
+1, 367804, 1024, 0x5df3fc30
+1, 369894, 1024, 0x1a312aa5
+0, 369990, 7266, 0x352c8078
+1, 371984, 1024, 0x6bb8e302
+1, 374073, 1024, 0xbd9684bb
+1, 376163, 1024, 0x78b0b166
+1, 378253, 1024, 0xd9af5eae
+0, 379980, 7323, 0xcaf69d7c
+1, 380343, 1024, 0xdb90fe82
+1, 382433, 1024, 0x327614e9
+1, 384522, 1024, 0x1f19b7fe
+1, 386612, 1024, 0x46c53f96
+1, 388702, 1024, 0x921b2189
+0, 389970, 7309, 0x98c1e6f7
+1, 390792, 1024, 0xa8fbc85a
+1, 392882, 1024, 0xabfdaaae
+1, 394971, 1024, 0x6acc7387
+1, 397061, 1024, 0x0d9c27b5
+1, 399151, 1024, 0xba4dd809
+0, 399960, 7121, 0x913d5bd6
+1, 401241, 1024, 0x2a2ad521
+1, 403331, 1024, 0x892de38a
+1, 405420, 1024, 0xdc97a2eb
+1, 407510, 1024, 0x4f614ca4
+1, 409600, 1024, 0x9c8a77ea
+0, 409950, 7088, 0x56302362
+1, 411690, 1024, 0x2d30e646
+1, 413780, 1024, 0x74e800a7
+1, 415869, 1024, 0x1e01fb02
+1, 417959, 1024, 0x4ed2c1d8
+0, 419940, 7104, 0xc0d14f78
+1, 420049, 1024, 0xf2fdbe63
+1, 422139, 1024, 0x8d6f63a1
+1, 424229, 1024, 0xded468d9
+1, 426318, 1024, 0xccad839e
+1, 428408, 1024, 0xdde7c082
+0, 429930, 7169, 0xd03c825b
+1, 430498, 1024, 0x548613c5
+1, 432588, 1024, 0x383909bd
+1, 434678, 1024, 0xfd37627b
+1, 436767, 1024, 0x6d95a481
+1, 438857, 1024, 0x56aa87fa
+0, 439920, 7038, 0x1ecc201d
+1, 440947, 1024, 0x7b67258c
+1, 443037, 1024, 0x7dd99a92
+1, 445127, 1024, 0x4a66d102
+1, 447216, 1024, 0x7b3fce51
+1, 449306, 1024, 0xbbd968aa
+0, 450000, 7015, 0x83c94454
+1, 451396, 1024, 0x8283ec36
+1, 453486, 1024, 0x3c96493d
+1, 455576, 1024, 0xfa4f8cf8
+1, 457665, 1024, 0xe2cf872d
+1, 459755, 1024, 0x0a9e7aa6
+0, 459990, 6983, 0x9e51f54d
+1, 461845, 1024, 0x6e7a0550
+1, 463935, 1024, 0x3acfea2f
+1, 466024, 1024, 0x7111d0fa
+1, 468114, 1024, 0xe9a1eca9
+0, 469980, 7088, 0x70d33de1
+1, 470204, 1024, 0x24da6c46
+1, 472294, 1024, 0x117cff37
+1, 474384, 1024, 0x0f27cab6
+1, 476473, 1024, 0x69b6b4e6
+1, 478563, 1024, 0x1e6cc841
+0, 479970, 7096, 0x4d0f81b5
+1, 480653, 1024, 0xb01e2365
+1, 482743, 1024, 0x14e200d3
+1, 484833, 1024, 0xd1184c98
+1, 486922, 1024, 0xef9140e9
+1, 489012, 1024, 0x4cbb645e
+0, 489960, 7106, 0xd1a83ddc
+1, 491102, 1024, 0xe7fe2f06
+1, 493192, 1024, 0xf8c45028
+1, 495282, 1024, 0x561358f4
+1, 497371, 1024, 0xd0129b77
+1, 499461, 1024, 0xcc636e88
+0, 499950, 7219, 0x20f47fe4
+1, 501551, 1024, 0xe9406321
+1, 503641, 1024, 0x9f16a041
+1, 505731, 1024, 0x468bf409
+1, 507820, 1024, 0x3df70f7b
+1, 509910, 1024, 0xa880b11b
+0, 509940, 7184, 0x45dc6a0e
+1, 512000, 1024, 0x3286c489
+1, 514090, 1024, 0x39fe9ebc
+1, 516180, 1024, 0xc533d83b
+1, 518269, 1024, 0x153b195d
+0, 519930, 7222, 0x488c6499
+1, 520359, 1024, 0xd84786a1
+1, 522449, 1024, 0xdc295aaa
+1, 524539, 1024, 0xfb764d8c
+1, 526629, 1024, 0xeebc9db9
+1, 528718, 1024, 0x7ba9403e
+0, 529920, 7254, 0xbd097ba7
+1, 530808, 1024, 0x4e5571ec
+1, 532898, 1024, 0xd965fad4
+1, 534988, 1024, 0x87e259f2
+1, 537078, 1024, 0xae7e533b
+1, 539167, 1024, 0x313cf4d6
+0, 540000, 7189, 0x46e06d43
+1, 541257, 1024, 0xe1844c90
+1, 543347, 1024, 0xbb057b44
+1, 545437, 1024, 0xa5099687
+1, 547527, 1024, 0xbff10707
+1, 549616, 1024, 0x37c4ffc0
+0, 549990, 7283, 0x19dd7319
+1, 551706, 1024, 0xf9fb6caa
+1, 553796, 1024, 0x3b6a3a1f
+1, 555886, 1024, 0x83431edb
+1, 557976, 1024, 0x1eb713cf
+0, 559980, 7161, 0x23171d02
+1, 560065, 1024, 0xd7b07a6d
+1, 562155, 1024, 0x81ae3391
+1, 564245, 1024, 0xf150130a
+1, 566335, 1024, 0x09678eaa
+1, 568424, 1024, 0xb94e06f1
+0, 569970, 6976, 0xcc610c26
+1, 570514, 1024, 0x67b1dbc9
+1, 572604, 1024, 0xd6edc235
+1, 574694, 1024, 0x34e4c499
+1, 576784, 1024, 0xeefd89c0
+1, 578873, 1024, 0x38afdaf1
+0, 579960, 7056, 0x6cd917b0
+1, 580963, 1024, 0x29a60d76
+1, 583053, 1024, 0xe28a4372
+1, 585143, 1024, 0x7089454d
+1, 587233, 1024, 0x0c01bb7b
+1, 589322, 1024, 0xbd776a72
+0, 589950, 6736, 0x02b78951
+1, 591412, 1024, 0x86776fd0
+1, 593502, 1024, 0xb37c88f7
+1, 595592, 1024, 0x5f90aaf8
+1, 597682, 1024, 0x203d4222
+1, 599771, 1024, 0x382692a6
+0, 599940, 6540, 0x767e0854
+1, 601861, 1024, 0xf37c95fd
+1, 603951, 1024, 0x6c0b8877
+1, 606041, 1024, 0x2e54a8b6
+1, 608131, 1024, 0x7f266488
+0, 609930, 6170, 0xc84962fb
+1, 610220, 1024, 0xfbf20f9a
+1, 612310, 1024, 0xf2985cc0
+1, 614400, 1024, 0xc7075340
+1, 616490, 1024, 0xe4585695
+1, 618580, 1024, 0xbdffa380
+0, 619920, 6169, 0x27e06c03
+1, 620669, 1024, 0x2422a8a9
+1, 622759, 1024, 0x59cbd75f
+1, 624849, 1024, 0x04ad1a8c
+1, 626939, 1024, 0x33c09191
+1, 629029, 1024, 0x55efa6fd
+0, 630000, 5864, 0xd14db83f
+1, 631118, 1024, 0xf73d0e5d
+1, 633208, 1024, 0x6141ebae
+1, 635298, 1024, 0x7db17a68
+1, 637388, 1024, 0xa6c690b6
+1, 639478, 1024, 0xa6fd6725
+0, 639990, 5375, 0x4a21055d
+1, 641567, 1024, 0x50a90b9b
+1, 643657, 1024, 0xef990dc8
+1, 645747, 1024, 0x75adf6b5
+1, 647837, 1024, 0x61eac43e
+1, 649927, 1024, 0x67797a19
+0, 649980, 5206, 0x95ead3cb
+1, 652016, 1024, 0xf325277a
+1, 654106, 1024, 0x18bf254a
+1, 656196, 1024, 0x2ce6bee3
+1, 658286, 1024, 0x8d320860
+0, 659970, 5220, 0xcfdcc37e
+1, 660376, 1024, 0xc979b6e8
+1, 662465, 1024, 0xdb644b41
+1, 664555, 1024, 0xe1b368ba
+1, 666645, 1024, 0xacc53d15
+1, 668735, 1024, 0x42ea8c18
+0, 669960, 4946, 0x2d864a77
+1, 670824, 1024, 0xe52c99a4
+1, 672914, 1024, 0xd7db54a6
+1, 675004, 1024, 0x7f27a7e3
+1, 677094, 1024, 0xf7ffeaa9
+1, 679184, 1024, 0x792b6088
+0, 679950, 4390, 0x2ab9f462
+1, 681273, 1024, 0x61d99724
+1, 683363, 1024, 0x5213720e
+1, 685453, 1024, 0xac09dd30
+1, 687543, 1024, 0x960bf6bb
+1, 689633, 1024, 0xc90168e1
+0, 689940, 4051, 0x1d09592e
+1, 691722, 1024, 0x43b45768
+1, 693812, 1024, 0x935d60a1
+1, 695902, 1024, 0x9a342ef2
+1, 697992, 1024, 0xc894709f
+0, 699930, 3680, 0x39bd6a12
+1, 700082, 1024, 0x59b43b07
+1, 702171, 1024, 0x36a1a98d
+1, 704261, 1024, 0x9e1a121c
+1, 706351, 1024, 0x02208b78
+1, 708441, 1024, 0xd1d7b274
+0, 709920, 2910, 0x6337ece9
+1, 710531, 1024, 0xdacd5096
+1, 712620, 1024, 0x51b71ead
+1, 714710, 1024, 0xd009a7ca
+1, 716800, 1024, 0xb6d5a938
+1, 718890, 1024, 0xf3d45e47
+0, 720000, 2153, 0xf4e3bc17
+1, 720980, 1024, 0xea8e04fc
+1, 723069, 1024, 0x0b928bd8
+1, 725159, 1024, 0x0f02caec
+1, 727249, 1024, 0xe2b137a8
+1, 729339, 1024, 0xd5f94892
diff --git a/tests/ref/fate/vc1-ism b/tests/ref/fate/vc1-ism
new file mode 100644
index 0000000000..886e583c5d
--- /dev/null
+++ b/tests/ref/fate/vc1-ism
@@ -0,0 +1,120 @@
+0, 0, 37440, 0xd1bc5235
+0, 3750, 37440, 0x158e6167
+0, 7500, 37440, 0x0faa4481
+0, 11250, 37440, 0x427158c5
+0, 15000, 37440, 0x4eb53ac6
+0, 18750, 37440, 0x99304eea
+0, 22500, 37440, 0xcc554a6f
+0, 26250, 37440, 0xabeb6c35
+0, 30000, 37440, 0xddfc7e18
+0, 33750, 37440, 0xaa79b504
+0, 37500, 37440, 0x5cb1c839
+0, 41250, 37440, 0x7e36ecca
+0, 45000, 37440, 0xf486f425
+0, 48750, 37440, 0xf1b4138f
+0, 52500, 37440, 0x966f1a49
+0, 56250, 37440, 0x5eff21da
+0, 60000, 37440, 0x333f39b1
+0, 63750, 37440, 0x62e5963e
+0, 67500, 37440, 0x26930671
+0, 71250, 37440, 0x27b4bb6c
+0, 75000, 37440, 0xdbd07766
+0, 78750, 37440, 0x04260104
+0, 82500, 37440, 0x9b1e078b
+0, 86250, 37440, 0xdf4e2474
+0, 90000, 37440, 0x57d44986
+0, 93750, 37440, 0x8780e34c
+0, 97500, 37440, 0xf80c8bc0
+0, 101250, 37440, 0x630a7583
+0, 105000, 37440, 0x235ae089
+0, 108750, 37440, 0x984b8f0e
+0, 112500, 37440, 0x865cf592
+0, 116250, 37440, 0x70f376f2
+0, 120000, 37440, 0x8b30c035
+0, 123750, 37440, 0xde772d79
+0, 127500, 37440, 0x8e076be5
+0, 131250, 37440, 0x3dc2bd9f
+0, 135000, 37440, 0xb782eb67
+0, 138750, 37440, 0x02025d73
+0, 142500, 37440, 0x86bbbce8
+0, 146250, 37440, 0xd6554f62
+0, 150000, 37440, 0xb831b917
+0, 153750, 37440, 0x80643560
+0, 157500, 37440, 0x4ecf9afd
+0, 161250, 37440, 0x9ce51e0b
+0, 165000, 37440, 0x179466cd
+0, 168750, 37440, 0x145fc900
+0, 172500, 37440, 0xb1b50402
+0, 176250, 37440, 0x0a87552a
+0, 180000, 37440, 0x8f53821d
+0, 183750, 37440, 0x1c07c825
+0, 187500, 37440, 0x49dde82f
+0, 191250, 37440, 0xb1a32605
+0, 195000, 37440, 0x410f3cd5
+0, 198750, 37440, 0xff5e6696
+0, 202500, 37440, 0x96f678c9
+0, 206250, 37440, 0x6c9e9e68
+0, 210000, 37440, 0x79a2a655
+0, 213750, 37440, 0xf237bd6c
+0, 217500, 37440, 0x4051b611
+0, 221250, 37440, 0xc7ccc918
+0, 225000, 37440, 0xbd02c122
+0, 228750, 37440, 0xacb3c881
+0, 232500, 37440, 0x2abdb940
+0, 236250, 37440, 0x19d5be85
+0, 240000, 37440, 0xfa5fb1ba
+0, 243750, 37440, 0xdae7a7aa
+0, 247500, 37440, 0x6b0f9f69
+0, 251250, 37440, 0x353e8201
+0, 255000, 37440, 0xa21443aa
+0, 258750, 37440, 0x66c8d7e0
+0, 262500, 37440, 0xc332068e
+0, 266250, 37440, 0x71431b9b
+0, 270000, 37440, 0x392f15cb
+0, 273750, 37440, 0x95a146bb
+0, 277500, 37440, 0x7c51740a
+0, 281250, 37440, 0xa3bdd43c
+0, 285000, 37440, 0xa079f965
+0, 288750, 37440, 0xa95423ea
+0, 292500, 37440, 0xd1bd2c67
+0, 296250, 37440, 0x6cf82844
+0, 300000, 37440, 0xd401e128
+0, 303750, 37440, 0x1f7db118
+0, 307500, 37440, 0x2e0a65a9
+0, 311250, 37440, 0x321c1c40
+0, 315000, 37440, 0x95b2a127
+0, 318750, 37440, 0xa1471f4b
+0, 322500, 37440, 0x29d148c0
+0, 326250, 37440, 0x24c07107
+0, 330000, 37440, 0x0ead678d
+0, 333750, 37440, 0xd0ca6495
+0, 337500, 37440, 0x08f935ef
+0, 341250, 37440, 0xb5ec3c38
+0, 345000, 37440, 0xce371628
+0, 348750, 37440, 0x68170812
+0, 352500, 37440, 0xe222699e
+0, 356250, 37440, 0xd688706c
+0, 360000, 37440, 0x81a033f9
+0, 363750, 37440, 0x28bd0fbf
+0, 367500, 37440, 0xe36db7b2
+0, 371250, 37440, 0x30559121
+0, 375000, 37440, 0xbf2b5fc8
+0, 378750, 37440, 0x4b427672
+0, 382500, 37440, 0x0544b0b4
+0, 386250, 37440, 0x38a70b06
+0, 390000, 37440, 0x4ed62607
+0, 393750, 37440, 0x6efe8ea6
+0, 397500, 37440, 0x81197e11
+0, 401250, 37440, 0xf4060050
+0, 405000, 37440, 0xaf205f13
+0, 408750, 37440, 0x5fa21382
+0, 412500, 37440, 0x8627ad05
+0, 416250, 37440, 0xf7130133
+0, 420000, 37440, 0x76dea7ba
+0, 423750, 37440, 0x1dbae1be
+0, 427500, 37440, 0x74a933f7
+0, 431250, 37440, 0xbdcd41a3
+0, 435000, 37440, 0xf0fe8c1c
+0, 438750, 37440, 0xc0036222
+0, 442500, 37440, 0x3058385c
+0, 446250, 37440, 0x68141016
diff --git a/tests/ref/fate/wtv-demux b/tests/ref/fate/wtv-demux
new file mode 100644
index 0000000000..b3744ee2d1
--- /dev/null
+++ b/tests/ref/fate/wtv-demux
@@ -0,0 +1,139 @@
+1, 0, 576, 0x9b6e1638
+1, 1620, 576, 0x0ca91183
+1, 3780, 576, 0xec6a180f
+1, 5940, 576, 0x478a2b9b
+1, 8100, 576, 0x00fa15b3
+1, 10260, 576, 0xfb551816
+1, 12960, 576, 0x422e12bd
+1, 15120, 576, 0xa7581b29
+1, 17280, 576, 0xd4b31a74
+1, 19440, 576, 0x11521b10
+1, 21600, 576, 0x3dcc1474
+1, 23760, 576, 0x66c31aab
+1, 25920, 576, 0x97f318a8
+1, 28080, 576, 0xd3fb1a30
+1, 30240, 576, 0xd2bd16af
+1, 32400, 576, 0x6c10146a
+1, 34560, 576, 0x10d81468
+1, 36720, 576, 0x3813162d
+1, 38880, 576, 0x89e71d95
+1, 41040, 576, 0xd1c717f9
+1, 43200, 576, 0x1a311e5f
+1, 45360, 576, 0x0ea80e05
+1, 47520, 576, 0x2f1718f2
+1, 49680, 576, 0xffe01e13
+1, 51840, 576, 0xa7b02296
+1, 54000, 576, 0x199f1597
+1, 56160, 576, 0xdea217ba
+1, 58320, 576, 0x8a790f01
+1, 60480, 576, 0x23e80038
+1, 62640, 576, 0x75dc048a
+1, 64800, 576, 0xeb4b0d93
+1, 66960, 576, 0xde1322f5
+1, 69120, 576, 0xc3131f35
+1, 71280, 576, 0x708f1381
+1, 73440, 576, 0x1f00137e
+0, 74578, 41980, 0xd4920915
+1, 75600, 576, 0x05131eb0
+1, 77760, 576, 0x78151c22
+0, 78178, 7228, 0x1b141fa3
+1, 79920, 576, 0x31771239
+0, 81777, 7492, 0x1a47f3e4
+1, 82080, 576, 0x3ce4097c
+1, 84240, 576, 0x180e15f4
+0, 85378, 25068, 0xcb70a744
+1, 86400, 576, 0x30db0604
+1, 88560, 576, 0x9b290284
+0, 88978, 7212, 0x0ab9f558
+1, 90720, 576, 0xcf340753
+0, 92578, 7612, 0xa93054f0
+1, 92880, 576, 0xdaa41457
+1, 95040, 576, 0x34d310a2
+0, 96177, 22868, 0xa77db64a
+1, 97200, 576, 0x58b31010
+1, 99360, 576, 0x19610f54
+0, 99778, 6260, 0x6cf76411
+1, 101520, 576, 0x17762352
+0, 103377, 6156, 0xe168394b
+1, 103680, 576, 0x1fea1448
+1, 105840, 576, 0x55840a01
+0, 106977, 23364, 0x53164f1e
+1, 108000, 576, 0x6c9c24ce
+1, 110160, 576, 0x955f1e97
+0, 110578, 6708, 0x89877269
+1, 112320, 576, 0x2827134f
+0, 114178, 6908, 0x8d62a249
+1, 114480, 576, 0x34a01c29
+1, 116640, 576, 0x7d351e52
+0, 117778, 38156, 0xec41f682
+1, 118800, 576, 0x00c91d9e
+1, 120960, 576, 0x57ea1a97
+0, 121377, 5764, 0xcc04534b
+1, 123120, 576, 0xef3a1c74
+0, 124977, 5388, 0xb8a1c3c5
+1, 125280, 576, 0x11fc217d
+1, 127440, 576, 0x59ce20e5
+0, 128578, 16764, 0x59460d96
+1, 129600, 576, 0xaafc1dbf
+1, 131760, 576, 0xdd941609
+0, 132177, 5548, 0x5c91e93d
+1, 133920, 576, 0x900420b0
+0, 135777, 5652, 0x5e321aed
+1, 136080, 576, 0x5f4f1aa1
+1, 138240, 576, 0x7d7e18de
+0, 139377, 15564, 0xefdf5080
+1, 140400, 576, 0x986c0d9d
+1, 142560, 576, 0xcb4c21c0
+0, 142977, 6492, 0xd1d5c5f8
+1, 144720, 576, 0xbcfb1e8b
+0, 146577, 5604, 0xf9472b44
+1, 146880, 576, 0xcb541b4c
+1, 149040, 576, 0x980426e9
+0, 150177, 17924, 0x45815b7b
+1, 151200, 576, 0x09d00aa0
+1, 153360, 576, 0xad591374
+0, 153778, 5020, 0x3cc5e554
+1, 155520, 576, 0x97bf1461
+0, 157378, 5276, 0xa0554c12
+1, 157680, 576, 0xdc871cc4
+1, 159840, 576, 0x56781896
+0, 160977, 31460, 0x5765eb5f
+1, 162000, 576, 0xc77714e3
+1, 164160, 576, 0x280e18d4
+0, 164577, 4972, 0x91adbab7
+1, 166320, 576, 0xbc0d2302
+0, 168178, 5580, 0xfea707cb
+1, 168480, 576, 0x79191384
+1, 170640, 576, 0x65481c97
+0, 171778, 17412, 0x0afe4d27
+1, 172800, 576, 0xc94d227d
+1, 174960, 576, 0xa68a1f14
+0, 175378, 5236, 0x03f55309
+1, 177120, 576, 0x6af11a5c
+0, 178977, 4924, 0x558e753c
+1, 179280, 576, 0x4d1019ef
+1, 181440, 576, 0x3b1b17b5
+0, 182577, 15396, 0xf145d121
+1, 183600, 576, 0xcdd8159f
+1, 185760, 576, 0x97cd1d06
+0, 186177, 4708, 0x43066a92
+1, 187920, 576, 0x5d1b1123
+0, 189778, 4332, 0x9e22bcba
+1, 190080, 576, 0x888d0cb0
+1, 192240, 576, 0x556e1dad
+0, 193377, 12876, 0x46ff9ef4
+1, 194400, 576, 0xf7af0bce
+1, 196560, 576, 0xb5da160a
+0, 196978, 5940, 0x27cba62e
+1, 198720, 576, 0x4a8d0e98
+0, 200578, 6124, 0x6bab0a6d
+1, 200880, 576, 0x183b1c7e
+1, 203040, 576, 0xc47120e6
+0, 204178, 36428, 0x942f9648
+1, 205200, 576, 0xb1f31346
+0, 207777, 6660, 0x545a0db7
+0, 211377, 6780, 0x2d1d4189
+0, 214978, 16460, 0x7c3b3ca4
+0, 218578, 6724, 0x8538cc6f
+0, 222178, 7068, 0x69574fd0
+0, 225777, 19552, 0xf230e854
diff --git a/tests/ref/fate/xwma-demux b/tests/ref/fate/xwma-demux
new file mode 100644
index 0000000000..83a3b086cf
--- /dev/null
+++ b/tests/ref/fate/xwma-demux
@@ -0,0 +1 @@
+CRC=0x2ac2159e
diff --git a/tests/ref/lavfi/pixdesc b/tests/ref/lavfi/pixdesc
index 07d1d36417..f5e398e560 100644
--- a/tests/ref/lavfi/pixdesc
+++ b/tests/ref/lavfi/pixdesc
@@ -4,6 +4,8 @@ abgr                037bf9df6a765520ad6d490066bf4b89
 argb                c442a8261c2265a07212ef0f72e35f5a
 bgr0                0ef8ffe42bfe53be74973a0c118b775e
 bgr24               0d0cb38ab3fa0b2ec0865c14f78b217b
+bgr444be            d9ea9307d21b162225b8b2c524cf9477
+bgr444le            88035350e9da3a8f67387890b956f0bc
 bgr48be             00624e6c7ec7ab19897ba2f0a3257fe8
 bgr48le             d02c235ebba7167881ca2d576497ff84
 bgr4_byte           50d23cc82d9dcef2fd12adb81fb9b806
@@ -22,6 +24,8 @@ nv12                e0af357888584d36eec5aa0f673793ef
 nv21                9a3297f3b34baa038b1f37cb202b512f
 rgb0                7c03f81f5e5346bf8ea42c4187f20605
 rgb24               b41eba9651e1b5fe386289b506188105
+rgb444be            9e89db334568c6b2e3d5d0540f4ba960
+rgb444le            0a68cb6de8bf530aa30c5c1205c25155
 rgb48be             cc139ec1dd9451f0e049c0cb3a0c8aa2
 rgb48le             86c5608904f75360d492dbc5c9589969
 rgb4_byte           c93ba89b74c504e7f5ae9d9ab1546c73
diff --git a/tests/ref/lavfi/pixfmts_copy b/tests/ref/lavfi/pixfmts_copy
index 5ced4785c8..11ed0238ed 100644
--- a/tests/ref/lavfi/pixfmts_copy
+++ b/tests/ref/lavfi/pixfmts_copy
@@ -4,6 +4,8 @@ abgr                037bf9df6a765520ad6d490066bf4b89
 argb                c442a8261c2265a07212ef0f72e35f5a
 bgr0                328a76e72c55508cdf04dc93a5b056fc
 bgr24               0d0cb38ab3fa0b2ec0865c14f78b217b
+bgr444be            d9ea9307d21b162225b8b2c524cf9477
+bgr444le            88035350e9da3a8f67387890b956f0bc
 bgr48be             00624e6c7ec7ab19897ba2f0a3257fe8
 bgr48le             d02c235ebba7167881ca2d576497ff84
 bgr4_byte           50d23cc82d9dcef2fd12adb81fb9b806
@@ -23,6 +25,8 @@ nv21                9a3297f3b34baa038b1f37cb202b512f
 pal8                dec8ed2258ec89b8a796f21cad4df867
 rgb0                ff1a9f355d43f9d25f07a191b5aa906c
 rgb24               b41eba9651e1b5fe386289b506188105
+rgb444be            9e89db334568c6b2e3d5d0540f4ba960
+rgb444le            0a68cb6de8bf530aa30c5c1205c25155
 rgb48be             cc139ec1dd9451f0e049c0cb3a0c8aa2
 rgb48le             86c5608904f75360d492dbc5c9589969
 rgb4_byte           c93ba89b74c504e7f5ae9d9ab1546c73
diff --git a/tests/ref/lavfi/pixfmts_null b/tests/ref/lavfi/pixfmts_null
index 5ced4785c8..11ed0238ed 100644
--- a/tests/ref/lavfi/pixfmts_null
+++ b/tests/ref/lavfi/pixfmts_null
@@ -4,6 +4,8 @@ abgr                037bf9df6a765520ad6d490066bf4b89
 argb                c442a8261c2265a07212ef0f72e35f5a
 bgr0                328a76e72c55508cdf04dc93a5b056fc
 bgr24               0d0cb38ab3fa0b2ec0865c14f78b217b
+bgr444be            d9ea9307d21b162225b8b2c524cf9477
+bgr444le            88035350e9da3a8f67387890b956f0bc
 bgr48be             00624e6c7ec7ab19897ba2f0a3257fe8
 bgr48le             d02c235ebba7167881ca2d576497ff84
 bgr4_byte           50d23cc82d9dcef2fd12adb81fb9b806
@@ -23,6 +25,8 @@ nv21                9a3297f3b34baa038b1f37cb202b512f
 pal8                dec8ed2258ec89b8a796f21cad4df867
 rgb0                ff1a9f355d43f9d25f07a191b5aa906c
 rgb24               b41eba9651e1b5fe386289b506188105
+rgb444be            9e89db334568c6b2e3d5d0540f4ba960
+rgb444le            0a68cb6de8bf530aa30c5c1205c25155
 rgb48be             cc139ec1dd9451f0e049c0cb3a0c8aa2
 rgb48le             86c5608904f75360d492dbc5c9589969
 rgb4_byte           c93ba89b74c504e7f5ae9d9ab1546c73
diff --git a/tests/ref/lavfi/pixfmts_scale b/tests/ref/lavfi/pixfmts_scale
index 597904d02c..f45295128d 100644
--- a/tests/ref/lavfi/pixfmts_scale
+++ b/tests/ref/lavfi/pixfmts_scale
@@ -4,6 +4,8 @@ abgr                cff82561a074874027ac1cc896fd2730
 argb                756dd1eaa5baca2238ce23dbdc452684
 bgr0                ff6e1dfa26d4c2ada3a59e8b0b600d1f
 bgr24               e44192347a45586c6c157e3059610cd1
+bgr444be            c23768338d76693f0da76e8a9b6fd8df
+bgr444le            846c431a47bfb745437941bde768469c
 bgr48be             390d3058a12a99c2b153ed7922508bea
 bgr48le             39fe06feb4ec1d9730dccc04a0cfac4c
 bgr4_byte           ee1d35a7baf8e9016891929a2f565c0b
@@ -23,6 +25,8 @@ nv21                69c699510ff1fb777b118ebee1002f14
 pal8                6324fa058e1bc157ed7132bfe4022317
 rgb0                1bd6f54ad067503ac9783a70062c8f87
 rgb24               13ff53ebeab74dc05492836f1cfbd2c1
+rgb444be            46e466b2709f62b2fffc63708063eaaf
+rgb444le            f0c57a48be671428e2e53c9b54a6c4e2
 rgb48be             8fac63787a711886030f8e056872b488
 rgb48le             ab92f2763a2eb264c3870cc758f97149
 rgb4_byte           d81ffd3add95842a618eec81024f0b5c
diff --git a/tests/ref/lavfi/pixfmts_vflip b/tests/ref/lavfi/pixfmts_vflip
index 333bd46484..1e9a636291 100644
--- a/tests/ref/lavfi/pixfmts_vflip
+++ b/tests/ref/lavfi/pixfmts_vflip
@@ -4,6 +4,8 @@ abgr                25e72e9dbd01ab00727c976d577f7be5
 argb                19869bf1a5ac0b6af4d8bbe2c104533c
 bgr0                e03d3ee0b977f6d86e5116b20494bef5
 bgr24               89108a4ba00201f79b75b9305c42352d
+bgr444be            9ef12c42fb791948ca4423c452dc6b9a
+bgr444le            3650ecfc163abd1596c0cd29d130c4b0
 bgr48be             2f23931844f57641f3737348182d118c
 bgr48le             4242a026012b6c135a6aa138a6d67031
 bgr4_byte           407fcf564ed764c38e1d748f700ab921
@@ -23,6 +25,8 @@ nv21                01ea369dd2d0d3ed7451dc5c8d61497f
 pal8                47ed19a7e128b0e33c25d2a463b0611a
 rgb0                330bd6168e46c0d5eb4acbdbb50afa2e
 rgb24               eaefabc168d0b14576bab45bc1e56e1e
+rgb444be            06722e03f8404e7d2226665ed2444a32
+rgb444le            185c9a5d9c2877484310d4196ef4cd6f
 rgb48be             62dd185862ed142283bd300eb6dbd216
 rgb48le             dcb76353268bc5862194d131762220da
 rgb4_byte           8c6ff02df0b06dd2d574836c3741b2a2