Merge remote-tracking branch 'qatar/master'

* qatar/master: (36 commits)
  ARM: allow unaligned buffer in fixed-point NEON FFT4
  fate: test more FFT etc sizes
  dca: set AVCodecContext frame_size for DTS audio
  YASM: Shut up unused variable compiler warning with --disable-yasm.
  x86_32: Fix build on x86_32 with --disable-yasm.
  iirfilter: add fate test
  doxygen: Add qmul docs.
  ogg: propagate return values and return more meaningful error values
  H.264: fix overreads of qscale_table
  Remove unused static tables and static inline functions.
  eval: clear Parser instances before using
  dct-test: remove 'ref' function pointer from tables
  build: Remove deleted 'check' target from .PHONY list.
  oggdec: Abort Ogg header parsing when encountering a data packet.
  Add LGPL license boilerplate to files lacking it.
  mxfenc: small typo fix
  doxygen: Fix documentation for some VP8 functions.
  sha: use AV_RB32() instead of assuming buffer can be cast to uint32_t*
  des: allow unaligned input and output buffers
  aes: allow unaligned input and output buffers
  ...

Conflicts:
	libavcodec/dct-test.c
	libavcodec/libvpxenc.c
	libavcodec/x86/dsputil_mmx.c
	libavcodec/x86/h264_qpel_mmx.c
	libavfilter/x86/gradfun.c
	libavformat/oggdec.c

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

17 files changed, 322 insertions, 273 deletions

diff --git a/libavcodec/arm/fft_fixed_neon.S b/libavcodec/arm/fft_fixed_neon.S
index bd6c853ec8..565a9c0d20 100644
--- a/libavcodec/arm/fft_fixed_neon.S
+++ b/libavcodec/arm/fft_fixed_neon.S
@@ -75,9 +75,9 @@
 .endm
 
 function fft4_neon
-        vld1.16         {d0-d1},  [r0,:128]
+        vld1.16         {d0-d1},  [r0]
         fft4            d0,  d1,  d2,  d3
-        vst1.16         {d0-d1},  [r0,:128]
+        vst1.16         {d0-d1},  [r0]
         bx              lr
 endfunc
 
diff --git a/libavcodec/dca.c b/libavcodec/dca.c
index b56005b7cf..e11439f939 100644
--- a/libavcodec/dca.c
+++ b/libavcodec/dca.c
@@ -1650,6 +1650,7 @@ static int dca_decode_frame(AVCodecContext * avctx,
     //set AVCodec values with parsed data
     avctx->sample_rate = s->sample_rate;
     avctx->bit_rate = s->bit_rate;
+    avctx->frame_size = s->sample_blocks * 32;
 
     s->profile = FF_PROFILE_DTS;
 
diff --git a/libavcodec/dct-test.c b/libavcodec/dct-test.c
index 2abb05bd3b..4bc59d3370 100644
--- a/libavcodec/dct-test.c
+++ b/libavcodec/dct-test.c
@@ -68,12 +68,12 @@ void ff_simple_idct_neon(DCTELEM *data);
 void ff_simple_idct_axp(DCTELEM *data);
 
 struct algo {
-  const char *name;
-  enum { FDCT, IDCT } is_idct;
-  void (* func) (DCTELEM *block);
-  void (* ref)  (DCTELEM *block);
-  enum formattag { NO_PERM,MMX_PERM, MMX_SIMPLE_PERM, SCALE_PERM, SSE2_PERM, PARTTRANS_PERM } format;
-  int  mm_support;
+    const char *name;
+    void (*func)(DCTELEM *block);
+    enum formattag { NO_PERM, MMX_PERM, MMX_SIMPLE_PERM, SCALE_PERM,
+                     SSE2_PERM, PARTTRANS_PERM } format;
+    int mm_support;
+    int nonspec;
 };
 
 #ifndef FAAN_POSTSCALE
@@ -84,61 +84,69 @@ struct algo {
 
 static int cpu_flags;
 
-struct algo algos[] = {
-  {"REF-DBL",         0, ff_ref_fdct,        ff_ref_fdct, NO_PERM},
-  {"FAAN",            0, ff_faandct,         ff_ref_fdct, FAAN_SCALE},
-  {"FAANI",           1, ff_faanidct,        ff_ref_idct, NO_PERM},
-  {"IJG-AAN-INT",     0, fdct_ifast,         ff_ref_fdct, SCALE_PERM},
-  {"IJG-LLM-INT",     0, ff_jpeg_fdct_islow, ff_ref_fdct, NO_PERM},
-  {"REF-DBL",         1, ff_ref_idct,        ff_ref_idct, NO_PERM},
-  {"INT",             1, j_rev_dct,          ff_ref_idct, MMX_PERM},
-  {"SIMPLE-C",        1, ff_simple_idct,     ff_ref_idct, NO_PERM},
+static const struct algo fdct_tab[] = {
+    { "REF-DBL",        ff_ref_fdct,           NO_PERM    },
+    { "FAAN",           ff_faandct,            FAAN_SCALE },
+    { "IJG-AAN-INT",    fdct_ifast,            SCALE_PERM },
+    { "IJG-LLM-INT",    ff_jpeg_fdct_islow,    NO_PERM    },
 
 #if HAVE_MMX
-  {"MMX",             0, ff_fdct_mmx,        ff_ref_fdct, NO_PERM, AV_CPU_FLAG_MMX},
-#if HAVE_MMX2
-  {"MMX2",            0, ff_fdct_mmx2,       ff_ref_fdct, NO_PERM, AV_CPU_FLAG_MMX2},
-  {"SSE2",            0, ff_fdct_sse2,       ff_ref_fdct, NO_PERM, AV_CPU_FLAG_SSE2},
+    { "MMX",            ff_fdct_mmx,           NO_PERM,   AV_CPU_FLAG_MMX     },
+    { "MMX2",           ff_fdct_mmx2,          NO_PERM,   AV_CPU_FLAG_MMX2    },
+    { "SSE2",           ff_fdct_sse2,          NO_PERM,   AV_CPU_FLAG_SSE2    },
 #endif
 
-#if CONFIG_GPL
-  {"LIBMPEG2-MMX",    1, ff_mmx_idct,        ff_ref_idct, MMX_PERM, AV_CPU_FLAG_MMX},
-  {"LIBMPEG2-MMX2",   1, ff_mmxext_idct,     ff_ref_idct, MMX_PERM, AV_CPU_FLAG_MMX2},
+#if HAVE_ALTIVEC
+    { "altivecfdct",    fdct_altivec,          NO_PERM,   AV_CPU_FLAG_ALTIVEC },
 #endif
-  {"SIMPLE-MMX",      1, ff_simple_idct_mmx, ff_ref_idct, MMX_SIMPLE_PERM, AV_CPU_FLAG_MMX},
-  {"XVID-MMX",        1, ff_idct_xvid_mmx,   ff_ref_idct, NO_PERM, AV_CPU_FLAG_MMX},
-  {"XVID-MMX2",       1, ff_idct_xvid_mmx2,  ff_ref_idct, NO_PERM, AV_CPU_FLAG_MMX2},
-  {"XVID-SSE2",       1, ff_idct_xvid_sse2,  ff_ref_idct, SSE2_PERM, AV_CPU_FLAG_SSE2},
+
+#if ARCH_BFIN
+    { "BFINfdct",       ff_bfin_fdct,          NO_PERM  },
 #endif
 
-#if HAVE_ALTIVEC
-  {"altivecfdct",     0, fdct_altivec,       ff_ref_fdct, NO_PERM, AV_CPU_FLAG_ALTIVEC},
+    { 0 }
+};
+
+static const struct algo idct_tab[] = {
+    { "FAANI",          ff_faanidct,           NO_PERM  },
+    { "REF-DBL",        ff_ref_idct,           NO_PERM  },
+    { "INT",            j_rev_dct,             MMX_PERM },
+    { "SIMPLE-C",       ff_simple_idct,        NO_PERM  },
+
+#if HAVE_MMX
+#if CONFIG_GPL
+    { "LIBMPEG2-MMX",   ff_mmx_idct,           MMX_PERM,  AV_CPU_FLAG_MMX,  1 },
+    { "LIBMPEG2-MMX2",  ff_mmxext_idct,        MMX_PERM,  AV_CPU_FLAG_MMX2, 1 },
+#endif
+    { "SIMPLE-MMX",     ff_simple_idct_mmx,  MMX_SIMPLE_PERM, AV_CPU_FLAG_MMX },
+    { "XVID-MMX",       ff_idct_xvid_mmx,      NO_PERM,   AV_CPU_FLAG_MMX,  1 },
+    { "XVID-MMX2",      ff_idct_xvid_mmx2,     NO_PERM,   AV_CPU_FLAG_MMX2, 1 },
+    { "XVID-SSE2",      ff_idct_xvid_sse2,     SSE2_PERM, AV_CPU_FLAG_SSE2, 1 },
 #endif
 
 #if ARCH_BFIN
-  {"BFINfdct",        0, ff_bfin_fdct,       ff_ref_fdct, NO_PERM},
-  {"BFINidct",        1, ff_bfin_idct,       ff_ref_idct, NO_PERM},
+    { "BFINidct",       ff_bfin_idct,          NO_PERM  },
 #endif
 
 #if ARCH_ARM
-  {"SIMPLE-ARM",      1, ff_simple_idct_arm, ff_ref_idct, NO_PERM },
-  {"INT-ARM",         1, ff_j_rev_dct_arm,   ff_ref_idct, MMX_PERM },
+    { "SIMPLE-ARM",     ff_simple_idct_arm,    NO_PERM  },
+    { "INT-ARM",        ff_j_rev_dct_arm,      MMX_PERM },
+#endif
 #if HAVE_ARMV5TE
-  {"SIMPLE-ARMV5TE",  1, ff_simple_idct_armv5te, ff_ref_idct, NO_PERM },
+    { "SIMPLE-ARMV5TE", ff_simple_idct_armv5te,NO_PERM  },
 #endif
 #if HAVE_ARMV6
-  {"SIMPLE-ARMV6",    1, ff_simple_idct_armv6, ff_ref_idct, MMX_PERM },
+    { "SIMPLE-ARMV6",   ff_simple_idct_armv6,  MMX_PERM },
 #endif
 #if HAVE_NEON
-  {"SIMPLE-NEON",     1, ff_simple_idct_neon, ff_ref_idct, PARTTRANS_PERM },
+    { "SIMPLE-NEON",    ff_simple_idct_neon,   PARTTRANS_PERM },
 #endif
-#endif /* ARCH_ARM */
 
 #if ARCH_ALPHA
-  {"SIMPLE-ALPHA",    1, ff_simple_idct_axp,  ff_ref_idct, NO_PERM },
+    { "SIMPLE-ALPHA",   ff_simple_idct_axp,    NO_PERM },
 #endif
 
-  { 0 }
+    { 0 }
 };
 
 #define AANSCALE_BITS 12
@@ -148,7 +156,7 @@ uint8_t cropTbl[256 + 2 * MAX_NEG_CROP];
 static int64_t gettime(void)
 {
     struct timeval tv;
-    gettimeofday(&tv,NULL);
+    gettimeofday(&tv, NULL);
     return (int64_t)tv.tv_sec * 1000000 + tv.tv_usec;
 }
 
@@ -157,18 +165,18 @@ static int64_t gettime(void)
 
 static short idct_mmx_perm[64];
 
-static short idct_simple_mmx_perm[64]={
-        0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D,
-        0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D,
-        0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D,
-        0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F,
-        0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F,
-        0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D,
-        0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F,
-        0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F,
+static short idct_simple_mmx_perm[64] = {
+    0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D,
+    0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D,
+    0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D,
+    0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F,
+    0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F,
+    0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D,
+    0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F,
+    0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F,
 };
 
-static const uint8_t idct_sse2_row_perm[8] = {0, 4, 1, 5, 2, 6, 3, 7};
+static const uint8_t idct_sse2_row_perm[8] = { 0, 4, 1, 5, 2, 6, 3, 7 };
 
 static void idct_mmx_init(void)
 {
@@ -177,13 +185,12 @@ static void idct_mmx_init(void)
     /* the mmx/mmxext idct uses a reordered input, so we patch scan tables */
     for (i = 0; i < 64; i++) {
         idct_mmx_perm[i] = (i & 0x38) | ((i & 6) >> 1) | ((i & 1) << 2);
-//        idct_simple_mmx_perm[i] = simple_block_permute_op(i);
     }
 }
 
 DECLARE_ALIGNED(16, static DCTELEM, block)[64];
-DECLARE_ALIGNED(8, static DCTELEM, block1)[64];
-DECLARE_ALIGNED(8, static DCTELEM, block_org)[64];
+DECLARE_ALIGNED(8,  static DCTELEM, block1)[64];
+DECLARE_ALIGNED(8,  static DCTELEM, block_org)[64];
 
 static inline void mmx_emms(void)
 {
@@ -193,188 +200,177 @@ static inline void mmx_emms(void)
 #endif
 }
 
-static void dct_error(const char *name, int is_idct,
-               void (*fdct_func)(DCTELEM *block),
-               void (*fdct_ref)(DCTELEM *block), int form, int test, const int bits)
+
+static int dct_error(const struct algo *dct, int test, int is_idct, int speed, const int bits)
 {
+    void (*ref)(DCTELEM *block) = is_idct ? ff_ref_idct : ff_ref_fdct;
     int it, i, scale;
     int err_inf, v;
-    int64_t err2, ti, ti1, it1;
-    int64_t sysErr[64], sysErrMax=0;
-    int maxout=0;
-    int blockSumErrMax=0, blockSumErr;
+    int64_t err2, ti, ti1, it1, err_sum = 0;
+    int64_t sysErr[64], sysErrMax = 0;
+    int maxout = 0;
+    int blockSumErrMax = 0, blockSumErr;
     AVLFG prng;
     const int vals=1<<bits;
+    double omse, ome;
+    int spec_err;
 
     av_lfg_init(&prng, 1);
 
     err_inf = 0;
     err2 = 0;
-    for(i=0; i<64; i++) sysErr[i]=0;
-    for(it=0;it<NB_ITS;it++) {
-        for(i=0;i<64;i++)
+    for (i = 0; i < 64; i++)
+        sysErr[i] = 0;
+    for (it = 0; it < NB_ITS; it++) {
+        for (i = 0; i < 64; i++)
             block1[i] = 0;
-        switch(test){
+        switch (test) {
         case 0:
-            for(i=0;i<64;i++)
+            for (i = 0; i < 64; i++)
                 block1[i] = (av_lfg_get(&prng) % (2*vals)) -vals;
-            if (is_idct){
+            if (is_idct) {
                 ff_ref_fdct(block1);
-
-                for(i=0;i<64;i++)
-                    block1[i]>>=3;
+                for (i = 0; i < 64; i++)
+                    block1[i] >>= 3;
             }
-        break;
-        case 1:{
-            int num = av_lfg_get(&prng) % 10 + 1;
-            for(i=0;i<num;i++)
-                block1[av_lfg_get(&prng) % 64] = av_lfg_get(&prng) % (2*vals) -vals;
-        }break;
+            break;
+        case 1: {
+                int num = av_lfg_get(&prng) % 10 + 1;
+                for (i = 0; i < num; i++)
+                    block1[av_lfg_get(&prng) % 64] = av_lfg_get(&prng) % (2*vals) -vals;
+            }
+            break;
         case 2:
             block1[0] = av_lfg_get(&prng) % (16*vals) - (8*vals);
-            block1[63]= (block1[0]&1)^1;
-        break;
+            block1[63] = (block1[0] & 1) ^ 1;
+            break;
         }
 
-#if 0 // simulate mismatch control
-{ int sum=0;
-        for(i=0;i<64;i++)
-           sum+=block1[i];
-
-        if((sum&1)==0) block1[63]^=1;
-}
-#endif
+        for (i = 0; i < 64; i++)
+            block_org[i] = block1[i];
 
-        for(i=0; i<64; i++)
-            block_org[i]= block1[i];
-
-        if (form == MMX_PERM) {
-            for(i=0;i<64;i++)
+        if (dct->format == MMX_PERM) {
+            for (i = 0; i < 64; i++)
                 block[idct_mmx_perm[i]] = block1[i];
-            } else if (form == MMX_SIMPLE_PERM) {
-            for(i=0;i<64;i++)
+        } else if (dct->format == MMX_SIMPLE_PERM) {
+            for (i = 0; i < 64; i++)
                 block[idct_simple_mmx_perm[i]] = block1[i];
-
-        } else if (form == SSE2_PERM) {
-            for(i=0; i<64; i++)
-                block[(i&0x38) | idct_sse2_row_perm[i&7]] = block1[i];
-        } else if (form == PARTTRANS_PERM) {
-            for(i=0; i<64; i++)
-                block[(i&0x24) | ((i&3)<<3) | ((i>>3)&3)] = block1[i];
+        } else if (dct->format == SSE2_PERM) {
+            for (i = 0; i < 64; i++)
+                block[(i & 0x38) | idct_sse2_row_perm[i & 7]] = block1[i];
+        } else if (dct->format == PARTTRANS_PERM) {
+            for (i = 0; i < 64; i++)
+                block[(i & 0x24) | ((i & 3) << 3) | ((i >> 3) & 3)] = block1[i];
         } else {
-            for(i=0; i<64; i++)
-                block[i]= block1[i];
+            for (i = 0; i < 64; i++)
+                block[i] = block1[i];
         }
-#if 0 // simulate mismatch control for tested IDCT but not the ref
-{ int sum=0;
-        for(i=0;i<64;i++)
-           sum+=block[i];
-
-        if((sum&1)==0) block[63]^=1;
-}
-#endif
 
-        fdct_func(block);
+        dct->func(block);
         mmx_emms();
 
-        if (form == SCALE_PERM) {
-            for(i=0; i<64; i++) {
-                scale = 8*(1 << (AANSCALE_BITS + 11)) / ff_aanscales[i];
-                block[i] = (block[i] * scale /*+ (1<<(AANSCALE_BITS-1))*/) >> AANSCALE_BITS;
+        if (dct->format == SCALE_PERM) {
+            for (i = 0; i < 64; i++) {
+                scale = 8 * (1 << (AANSCALE_BITS + 11)) / ff_aanscales[i];
+                block[i] = (block[i] * scale) >> AANSCALE_BITS;
             }
         }
 
-        fdct_ref(block1);
+        ref(block1);
 
-        blockSumErr=0;
-        for(i=0;i<64;i++) {
-            v = abs(block[i] - block1[i]);
+        blockSumErr = 0;
+        for (i = 0; i < 64; i++) {
+            int err = block[i] - block1[i];
+            err_sum += err;
+            v = abs(err);
             if (v > err_inf)
                 err_inf = v;
             err2 += v * v;
             sysErr[i] += block[i] - block1[i];
             blockSumErr += v;
-            if( abs(block[i])>maxout) maxout=abs(block[i]);
+            if (abs(block[i]) > maxout)
+                maxout = abs(block[i]);
         }
-        if(blockSumErrMax < blockSumErr) blockSumErrMax= blockSumErr;
-#if 0 // print different matrix pairs
-        if(blockSumErr){
-            printf("\n");
-            for(i=0; i<64; i++){
-                if((i&7)==0) printf("\n");
-                printf("%4d ", block_org[i]);
-            }
-            for(i=0; i<64; i++){
-                if((i&7)==0) printf("\n");
-                printf("%4d ", block[i] - block1[i]);
-            }
-        }
-#endif
+        if (blockSumErrMax < blockSumErr)
+            blockSumErrMax = blockSumErr;
     }
-    for(i=0; i<64; i++) sysErrMax= FFMAX(sysErrMax, FFABS(sysErr[i]));
+    for (i = 0; i < 64; i++)
+        sysErrMax = FFMAX(sysErrMax, FFABS(sysErr[i]));
 
-    for(i=0; i<64; i++){
-        if(i%8==0) printf("\n");
-        printf("%7d ", (int)sysErr[i]);
+    for (i = 0; i < 64; i++) {
+        if (i % 8 == 0)
+            printf("\n");
+        printf("%7d ", (int) sysErr[i]);
     }
     printf("\n");
 
-    printf("%s %s: err_inf=%d err2=%0.8f syserr=%0.8f maxout=%d blockSumErr=%d\n",
-           is_idct ? "IDCT" : "DCT",
-           name, err_inf, (double)err2 / NB_ITS / 64.0, (double)sysErrMax / NB_ITS, maxout, blockSumErrMax);
+    omse = (double) err2 / NB_ITS / 64;
+    ome  = (double) err_sum / NB_ITS / 64;
+
+    spec_err = is_idct && (err_inf > 1 || omse > 0.02 || fabs(ome) > 0.0015);
+
+    printf("%s %s: ppe=%d omse=%0.8f ome=%0.8f syserr=%0.8f maxout=%d blockSumErr=%d\n",
+           is_idct ? "IDCT" : "DCT", dct->name, err_inf,
+           omse, ome, (double) sysErrMax / NB_ITS,
+           maxout, blockSumErrMax);
+
+    if (spec_err && !dct->nonspec)
+        return 1;
+
+    if (!speed)
+        return 0;
 
     /* speed test */
-    for(i=0;i<64;i++)
+    for (i = 0; i < 64; i++)
         block1[i] = 0;
-    switch(test){
+
+    switch (test) {
     case 0:
-        for(i=0;i<64;i++)
+        for (i = 0; i < 64; i++)
             block1[i] = av_lfg_get(&prng) % (2*vals) -vals;
-        if (is_idct){
+        if (is_idct) {
             ff_ref_fdct(block1);
-
-            for(i=0;i<64;i++)
-                block1[i]>>=3;
+            for (i = 0; i < 64; i++)
+                block1[i] >>= 3;
         }
-    break;
-    case 1:{
+        break;
+    case 1:
     case 2:
         block1[0] = av_lfg_get(&prng) % (2*vals) -vals;
         block1[1] = av_lfg_get(&prng) % (2*vals) -vals;
         block1[2] = av_lfg_get(&prng) % (2*vals) -vals;
         block1[3] = av_lfg_get(&prng) % (2*vals) -vals;
-    }break;
+        break;
     }
 
-    if (form == MMX_PERM) {
-        for(i=0;i<64;i++)
+    if (dct->format == MMX_PERM) {
+        for (i = 0; i < 64; i++)
             block[idct_mmx_perm[i]] = block1[i];
-    } else if(form == MMX_SIMPLE_PERM) {
-        for(i=0;i<64;i++)
+    } else if (dct->format == MMX_SIMPLE_PERM) {
+        for (i = 0; i < 64; i++)
             block[idct_simple_mmx_perm[i]] = block1[i];
     } else {
-        for(i=0; i<64; i++)
-            block[i]= block1[i];
+        for (i = 0; i < 64; i++)
+            block[i] = block1[i];
     }
 
     ti = gettime();
     it1 = 0;
     do {
-        for(it=0;it<NB_ITS_SPEED;it++) {
-            for(i=0; i<64; i++)
-                block[i]= block1[i];
-//            memcpy(block, block1, sizeof(DCTELEM) * 64);
-// do not memcpy especially not fastmemcpy because it does movntq !!!
-            fdct_func(block);
+        for (it = 0; it < NB_ITS_SPEED; it++) {
+            for (i = 0; i < 64; i++)
+                block[i] = block1[i];
+            dct->func(block);
         }
         it1 += NB_ITS_SPEED;
         ti1 = gettime() - ti;
     } while (ti1 < 1000000);
     mmx_emms();
 
-    printf("%s %s: %0.1f kdct/s\n",
-           is_idct ? "IDCT" : "DCT",
-           name, (double)it1 * 1000.0 / (double)ti1);
+    printf("%s %s: %0.1f kdct/s\n", is_idct ? "IDCT" : "DCT", dct->name,
+           (double) it1 * 1000.0 / (double) ti1);
+
+    return 0;
 }
 
 DECLARE_ALIGNED(8, static uint8_t, img_dest)[64];
@@ -392,19 +388,19 @@ static void idct248_ref(uint8_t *dest, int linesize, int16_t *block)
     if (!init) {
         init = 1;
 
-        for(i=0;i<8;i++) {
+        for (i = 0; i < 8; i++) {
             sum = 0;
-            for(j=0;j<8;j++) {
-                s = (i==0) ? sqrt(1.0/8.0) : sqrt(1.0/4.0);
+            for (j = 0; j < 8; j++) {
+                s = (i == 0) ? sqrt(1.0 / 8.0) : sqrt(1.0 / 4.0);
                 c8[i][j] = s * cos(M_PI * i * (j + 0.5) / 8.0);
                 sum += c8[i][j] * c8[i][j];
             }
         }
 
-        for(i=0;i<4;i++) {
+        for (i = 0; i < 4; i++) {
             sum = 0;
-            for(j=0;j<4;j++) {
-                s = (i==0) ? sqrt(1.0/4.0) : sqrt(1.0/2.0);
+            for (j = 0; j < 4; j++) {
+                s = (i == 0) ? sqrt(1.0 / 4.0) : sqrt(1.0 / 2.0);
                 c4[i][j] = s * cos(M_PI * i * (j + 0.5) / 4.0);
                 sum += c4[i][j] * c4[i][j];
             }
@@ -413,58 +409,59 @@ static void idct248_ref(uint8_t *dest, int linesize, int16_t *block)
 
     /* butterfly */
     s = 0.5 * sqrt(2.0);
-    for(i=0;i<4;i++) {
-        for(j=0;j<8;j++) {
-            block1[8*(2*i)+j] = (block[8*(2*i)+j] + block[8*(2*i+1)+j]) * s;
-            block1[8*(2*i+1)+j] = (block[8*(2*i)+j] - block[8*(2*i+1)+j]) * s;
+    for (i = 0; i < 4; i++) {
+        for (j = 0; j < 8; j++) {
+            block1[8 * (2 * i) + j] =
+                (block[8 * (2 * i) + j] + block[8 * (2 * i + 1) + j]) * s;
+            block1[8 * (2 * i + 1) + j] =
+                (block[8 * (2 * i) + j] - block[8 * (2 * i + 1) + j]) * s;
         }
     }
 
     /* idct8 on lines */
-    for(i=0;i<8;i++) {
-        for(j=0;j<8;j++) {
+    for (i = 0; i < 8; i++) {
+        for (j = 0; j < 8; j++) {
             sum = 0;
-            for(k=0;k<8;k++)
-                sum += c8[k][j] * block1[8*i+k];
-            block2[8*i+j] = sum;
+            for (k = 0; k < 8; k++)
+                sum += c8[k][j] * block1[8 * i + k];
+            block2[8 * i + j] = sum;
         }
     }
 
     /* idct4 */
-    for(i=0;i<8;i++) {
-        for(j=0;j<4;j++) {
+    for (i = 0; i < 8; i++) {
+        for (j = 0; j < 4; j++) {
             /* top */
             sum = 0;
-            for(k=0;k<4;k++)
-                sum += c4[k][j] * block2[8*(2*k)+i];
-            block3[8*(2*j)+i] = sum;
+            for (k = 0; k < 4; k++)
+                sum += c4[k][j] * block2[8 * (2 * k) + i];
+            block3[8 * (2 * j) + i] = sum;
 
             /* bottom */
             sum = 0;
-            for(k=0;k<4;k++)
-                sum += c4[k][j] * block2[8*(2*k+1)+i];
-            block3[8*(2*j+1)+i] = sum;
+            for (k = 0; k < 4; k++)
+                sum += c4[k][j] * block2[8 * (2 * k + 1) + i];
+            block3[8 * (2 * j + 1) + i] = sum;
         }
     }
 
     /* clamp and store the result */
-    for(i=0;i<8;i++) {
-        for(j=0;j<8;j++) {
-            v = block3[8*i+j];
-            if (v < 0)
-                v = 0;
-            else if (v > 255)
-                v = 255;
-            dest[i * linesize + j] = (int)rint(v);
+    for (i = 0; i < 8; i++) {
+        for (j = 0; j < 8; j++) {
+            v = block3[8 * i + j];
+            if      (v < 0)   v = 0;
+            else if (v > 255) v = 255;
+            dest[i * linesize + j] = (int) rint(v);
         }
     }
 }
 
 static void idct248_error(const char *name,
-                    void (*idct248_put)(uint8_t *dest, int line_size, int16_t *block))
+                          void (*idct248_put)(uint8_t *dest, int line_size,
+                                              int16_t *block),
+                          int speed)
 {
     int it, i, it1, ti, ti1, err_max, v;
-
     AVLFG prng;
 
     av_lfg_init(&prng, 1);
@@ -472,23 +469,22 @@ static void idct248_error(const char *name,
     /* just one test to see if code is correct (precision is less
        important here) */
     err_max = 0;
-    for(it=0;it<NB_ITS;it++) {
-
+    for (it = 0; it < NB_ITS; it++) {
         /* XXX: use forward transform to generate values */
-        for(i=0;i<64;i++)
+        for (i = 0; i < 64; i++)
             block1[i] = av_lfg_get(&prng) % 256 - 128;
         block1[0] += 1024;
 
-        for(i=0; i<64; i++)
-            block[i]= block1[i];
+        for (i = 0; i < 64; i++)
+            block[i] = block1[i];
         idct248_ref(img_dest1, 8, block);
 
-        for(i=0; i<64; i++)
-            block[i]= block1[i];
+        for (i = 0; i < 64; i++)
+            block[i] = block1[i];
         idct248_put(img_dest, 8, block);
 
-        for(i=0;i<64;i++) {
-            v = abs((int)img_dest[i] - (int)img_dest1[i]);
+        for (i = 0; i < 64; i++) {
+            v = abs((int) img_dest[i] - (int) img_dest1[i]);
             if (v == 255)
                 printf("%d %d\n", img_dest[i], img_dest1[i]);
             if (v > err_max)
@@ -514,18 +510,17 @@ static void idct248_error(const char *name,
         }
 #endif
     }
-    printf("%s %s: err_inf=%d\n",
-           1 ? "IDCT248" : "DCT248",
-           name, err_max);
+    printf("%s %s: err_inf=%d\n", 1 ? "IDCT248" : "DCT248", name, err_max);
+
+    if (!speed)
+        return;
 
     ti = gettime();
     it1 = 0;
     do {
-        for(it=0;it<NB_ITS_SPEED;it++) {
-            for(i=0; i<64; i++)
-                block[i]= block1[i];
-//            memcpy(block, block1, sizeof(DCTELEM) * 64);
-// do not memcpy especially not fastmemcpy because it does movntq !!!
+        for (it = 0; it < NB_ITS_SPEED; it++) {
+            for (i = 0; i < 64; i++)
+                block[i] = block1[i];
             idct248_put(img_dest, 8, block);
         }
         it1 += NB_ITS_SPEED;
@@ -533,9 +528,8 @@ static void idct248_error(const char *name,
     } while (ti1 < 1000000);
     mmx_emms();
 
-    printf("%s %s: %0.1f kdct/s\n",
-           1 ? "IDCT248" : "DCT248",
-           name, (double)it1 * 1000.0 / (double)ti1);
+    printf("%s %s: %0.1f kdct/s\n", 1 ? "IDCT248" : "DCT248", name,
+           (double) it1 * 1000.0 / (double) ti1);
 }
 
 static void help(void)
@@ -545,56 +539,67 @@ static void help(void)
            "            1 -> test with random sparse matrixes\n"
            "            2 -> do 3. test from mpeg4 std\n"
            "-i          test IDCT implementations\n"
-           "-4          test IDCT248 implementations\n");
+           "-4          test IDCT248 implementations\n"
+           "-t          speed test\n");
 }
 
 int main(int argc, char **argv)
 {
     int test_idct = 0, test_248_dct = 0;
-    int c,i;
-    int test=1;
+    int c, i;
+    int test = 1;
+    int speed = 0;
+    int err = 0;
     int bits=8;
+
     cpu_flags = av_get_cpu_flags();
 
     ff_ref_dct_init();
     idct_mmx_init();
 
-    for(i=0;i<256;i++) cropTbl[i + MAX_NEG_CROP] = i;
-    for(i=0;i<MAX_NEG_CROP;i++) {
+    for (i = 0; i < 256; i++)
+        cropTbl[i + MAX_NEG_CROP] = i;
+    for (i = 0; i < MAX_NEG_CROP; i++) {
         cropTbl[i] = 0;
         cropTbl[i + MAX_NEG_CROP + 256] = 255;
     }
 
-    for(;;) {
-        c = getopt(argc, argv, "ih4");
+    for (;;) {
+        c = getopt(argc, argv, "ih4t");
         if (c == -1)
             break;
-        switch(c) {
+        switch (c) {
         case 'i':
             test_idct = 1;
             break;
         case '4':
             test_248_dct = 1;
             break;
-        default :
+        case 't':
+            speed = 1;
+            break;
+        default:
         case 'h':
             help();
             return 0;
         }
     }
 
-    if(optind <argc) test= atoi(argv[optind]);
+    if (optind < argc)
+        test = atoi(argv[optind]);
     if(optind+1 < argc) bits= atoi(argv[optind+1]);
 
     printf("ffmpeg DCT/IDCT test\n");
 
     if (test_248_dct) {
-        idct248_error("SIMPLE-C", ff_simple_idct248_put);
+        idct248_error("SIMPLE-C", ff_simple_idct248_put, speed);
     } else {
-      for (i=0;algos[i].name;i++)
-        if (algos[i].is_idct == test_idct && !(~cpu_flags & algos[i].mm_support)) {
-          dct_error (algos[i].name, algos[i].is_idct, algos[i].func, algos[i].ref, algos[i].format, test, bits);
-        }
+        const struct algo *algos = test_idct ? idct_tab : fdct_tab;
+        for (i = 0; algos[i].name; i++)
+            if (!(~cpu_flags & algos[i].mm_support)) {
+                err |= dct_error(&algos[i], test, test_idct, speed, bits);
+            }
     }
-    return 0;
+
+    return err;
 }
diff --git a/libavcodec/h264idct_template.c b/libavcodec/h264idct_template.c
index 2acdf47a2f..d198c10c5c 100644
--- a/libavcodec/h264idct_template.c
+++ b/libavcodec/h264idct_template.c
@@ -237,6 +237,7 @@ void FUNCC(ff_h264_idct_add8)(uint8_t **dest, const int *block_offset, DCTELEM *
 }
 /**
  * IDCT transforms the 16 dc values and dequantizes them.
+ * @param qmul quantization parameter
  */
 void FUNCC(ff_h264_luma_dc_dequant_idct)(DCTELEM *p_output, DCTELEM *p_input, int qmul){
 #define stride 16
diff --git a/libavcodec/high_bit_depth.h b/libavcodec/high_bit_depth.h
index 511cd00f3a..c0a6eafe89 100644
--- a/libavcodec/high_bit_depth.h
+++ b/libavcodec/high_bit_depth.h
@@ -1,3 +1,21 @@
+/*
+ * 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 "dsputil.h"
 
 #ifndef BIT_DEPTH
diff --git a/libavcodec/iirfilter.c b/libavcodec/iirfilter.c
index 4b9aae2a5d..4785a7c7c2 100644
--- a/libavcodec/iirfilter.c
+++ b/libavcodec/iirfilter.c
@@ -311,6 +311,9 @@ av_cold void ff_iir_filter_free_coeffs(struct FFIIRFilterCoeffs *coeffs)
 }
 
 #ifdef TEST
+#undef printf
+#include <stdio.h>
+
 #define FILT_ORDER 4
 #define SIZE 1024
 int main(void)
@@ -320,7 +323,6 @@ int main(void)
     float cutoff_coeff = 0.4;
     int16_t x[SIZE], y[SIZE];
     int i;
-    FILE* fd;
 
     fcoeffs = ff_iir_filter_init_coeffs(NULL, FF_FILTER_TYPE_BUTTERWORTH,
                                         FF_FILTER_MODE_LOWPASS, FILT_ORDER,
@@ -333,13 +335,8 @@ int main(void)
 
     ff_iir_filter(fcoeffs, fstate, SIZE, x, 1, y, 1);
 
-    fd = fopen("in.bin", "w");
-    fwrite(x, sizeof(x[0]), SIZE, fd);
-    fclose(fd);
-
-    fd = fopen("out.bin", "w");
-    fwrite(y, sizeof(y[0]), SIZE, fd);
-    fclose(fd);
+    for (i = 0; i < SIZE; i++)
+        printf("%6d %6d\n", x[i], y[i]);
 
     ff_iir_filter_free_coeffs(fcoeffs);
     ff_iir_filter_free_state(fstate);
diff --git a/libavcodec/libvorbis.c b/libavcodec/libvorbis.c
index bc219ded9b..4e05268439 100644
--- a/libavcodec/libvorbis.c
+++ b/libavcodec/libvorbis.c
@@ -30,6 +30,7 @@
 #include "avcodec.h"
 #include "bytestream.h"
 #include "vorbis.h"
+#include "libavutil/mathematics.h"
 
 #undef NDEBUG
 #include <assert.h>
diff --git a/libavcodec/libvpxenc.c b/libavcodec/libvpxenc.c
index f5c942e0fc..ac1b79fcc7 100644
--- a/libavcodec/libvpxenc.c
+++ b/libavcodec/libvpxenc.c
@@ -31,6 +31,7 @@
 #include "avcodec.h"
 #include "libavutil/base64.h"
 #include "libavutil/opt.h"
+#include "libavutil/mathematics.h"
 
 /**
  * Portion of struct vpx_codec_cx_pkt from vpx_encoder.h.
diff --git a/libavcodec/libxvidff.c b/libavcodec/libxvidff.c
index effd2db158..c09d9a8ff8 100644
--- a/libavcodec/libxvidff.c
+++ b/libavcodec/libxvidff.c
@@ -30,6 +30,7 @@
 #include "avcodec.h"
 #include "libavutil/cpu.h"
 #include "libavutil/intreadwrite.h"
+#include "libavutil/mathematics.h"
 #include "libxvid_internal.h"
 #if !HAVE_MKSTEMP
 #include <fcntl.h>
diff --git a/libavcodec/mpegvideo.c b/libavcodec/mpegvideo.c
index c3c3e15033..d819cc083f 100644
--- a/libavcodec/mpegvideo.c
+++ b/libavcodec/mpegvideo.c
@@ -285,9 +285,10 @@ int ff_alloc_picture(MpegEncContext *s, Picture *pic, int shared){
         }
 
         FF_ALLOCZ_OR_GOTO(s->avctx, pic->mbskip_table , mb_array_size * sizeof(uint8_t)+2, fail) //the +2 is for the slice end check
-        FF_ALLOCZ_OR_GOTO(s->avctx, pic->qscale_table , mb_array_size * sizeof(uint8_t)  , fail)
+        FF_ALLOCZ_OR_GOTO(s->avctx, pic->qscale_table_base , (big_mb_num + s->mb_stride) * sizeof(uint8_t)  , fail)
         FF_ALLOCZ_OR_GOTO(s->avctx, pic->mb_type_base , (big_mb_num + s->mb_stride) * sizeof(uint32_t), fail)
         pic->mb_type= pic->mb_type_base + 2*s->mb_stride+1;
+        pic->qscale_table = pic->qscale_table_base + 2*s->mb_stride + 1;
         if(s->out_format == FMT_H264){
             for(i=0; i<2; i++){
                 FF_ALLOCZ_OR_GOTO(s->avctx, pic->motion_val_base[i], 2 * (b4_array_size+4)  * sizeof(int16_t), fail)
@@ -339,7 +340,7 @@ static void free_picture(MpegEncContext *s, Picture *pic){
     av_freep(&pic->mc_mb_var);
     av_freep(&pic->mb_mean);
     av_freep(&pic->mbskip_table);
-    av_freep(&pic->qscale_table);
+    av_freep(&pic->qscale_table_base);
     av_freep(&pic->mb_type_base);
     av_freep(&pic->dct_coeff);
     av_freep(&pic->pan_scan);
diff --git a/libavcodec/mpegvideo.h b/libavcodec/mpegvideo.h
index 2a54329a49..ca11543f73 100644
--- a/libavcodec/mpegvideo.h
+++ b/libavcodec/mpegvideo.h
@@ -88,6 +88,7 @@ typedef struct Picture{
      * halfpel luma planes.
      */
     uint8_t *interpolated[3];
+    int8_t *qscale_table_base;
     int16_t (*motion_val_base[2])[2];
     uint32_t *mb_type_base;
 #define MB_TYPE_INTRA MB_TYPE_INTRA4x4 //default mb_type if there is just one type
diff --git a/libavcodec/opt.h b/libavcodec/opt.h
index 70de27d192..2380e74332 100644
--- a/libavcodec/opt.h
+++ b/libavcodec/opt.h
@@ -1,3 +1,21 @@
+/*
+ * 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
+ */
+
 /**
  * @file
  * This header is provided for compatibility only and will be removed
diff --git a/libavcodec/rv34.c b/libavcodec/rv34.c
index 241132e092..72635faccd 100644
--- a/libavcodec/rv34.c
+++ b/libavcodec/rv34.c
@@ -424,17 +424,6 @@ static inline RV34VLC* choose_vlc_set(int quant, int mod, int type)
 }
 
 /**
- * Decode quantizer difference and return modified quantizer.
- */
-static inline int rv34_decode_dquant(GetBitContext *gb, int quant)
-{
-    if(get_bits1(gb))
-        return rv34_dquant_tab[get_bits1(gb)][quant];
-    else
-        return get_bits(gb, 5);
-}
-
-/**
  * Decode macroblock header and return CBP in case of success, -1 otherwise.
  */
 static int rv34_decode_mb_header(RV34DecContext *r, int8_t *intra_types)
@@ -1255,15 +1244,6 @@ static int check_slice_end(RV34DecContext *r, MpegEncContext *s)
     return 0;
 }
 
-static inline int slice_compare(SliceInfo *si1, SliceInfo *si2)
-{
-    return si1->type   != si2->type  ||
-           si1->start  >= si2->start ||
-           si1->width  != si2->width ||
-           si1->height != si2->height||
-           si1->pts    != si2->pts;
-}
-
 static int rv34_decode_slice(RV34DecContext *r, int end, const uint8_t* buf, int buf_size)
 {
     MpegEncContext *s = &r->s;
diff --git a/libavcodec/vc1dec.c b/libavcodec/vc1dec.c
index e82fba2fcb..172eb6f366 100644
--- a/libavcodec/vc1dec.c
+++ b/libavcodec/vc1dec.c
@@ -45,7 +45,6 @@
 #define MB_INTRA_VLC_BITS 9
 #define DC_VLC_BITS 9
 #define AC_VLC_BITS 9
-static const uint16_t table_mb_intra[64][2];
 
 
 static const uint16_t vlc_offs[] = {
diff --git a/libavcodec/vp8.c b/libavcodec/vp8.c
index c7d5a56af4..beb4de3185 100644
--- a/libavcodec/vp8.c
+++ b/libavcodec/vp8.c
@@ -641,8 +641,6 @@ void decode_mb_mode(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, uint8_
  * @param block destination for block coefficients
  * @param probs probabilities to use when reading trees from the bitstream
  * @param i initial coeff index, 0 unless a separate DC block is coded
- * @param zero_nhood the initial prediction context for number of surrounding
- *                   all-zero blocks (only left/top, so 0-2)
  * @param qmul array holding the dc/ac dequant factor at position 0/1
  * @return 0 if no coeffs were decoded
  *         otherwise, the index of the last coeff decoded plus one
@@ -701,6 +699,17 @@ skip_eob:
 }
 #endif
 
+/**
+ * @param c arithmetic bitstream reader context
+ * @param block destination for block coefficients
+ * @param probs probabilities to use when reading trees from the bitstream
+ * @param i initial coeff index, 0 unless a separate DC block is coded
+ * @param zero_nhood the initial prediction context for number of surrounding
+ *                   all-zero blocks (only left/top, so 0-2)
+ * @param qmul array holding the dc/ac dequant factor at position 0/1
+ * @return 0 if no coeffs were decoded
+ *         otherwise, the index of the last coeff decoded plus one
+ */
 static av_always_inline
 int decode_block_coeffs(VP56RangeCoder *c, DCTELEM block[16],
                         uint8_t probs[16][3][NUM_DCT_TOKENS-1],
@@ -1034,10 +1043,9 @@ static const uint8_t subpel_idx[3][8] = {
 };
 
 /**
- * Generic MC function.
+ * luma MC function
  *
  * @param s VP8 decoding context
- * @param luma 1 for luma (Y) planes, 0 for chroma (Cb/Cr) planes
  * @param dst target buffer for block data at block position
  * @param ref reference picture buffer at origin (0, 0)
  * @param mv motion vector (relative to block position) to get pixel data from
@@ -1083,6 +1091,23 @@ void vp8_mc_luma(VP8Context *s, uint8_t *dst, AVFrame *ref, const VP56mv *mv,
     }
 }
 
+/**
+ * chroma MC function
+ *
+ * @param s VP8 decoding context
+ * @param dst1 target buffer for block data at block position (U plane)
+ * @param dst2 target buffer for block data at block position (V plane)
+ * @param ref reference picture buffer at origin (0, 0)
+ * @param mv motion vector (relative to block position) to get pixel data from
+ * @param x_off horizontal position of block from origin (0, 0)
+ * @param y_off vertical position of block from origin (0, 0)
+ * @param block_w width of block (16, 8 or 4)
+ * @param block_h height of block (always same as block_w)
+ * @param width width of src/dst plane data
+ * @param height height of src/dst plane data
+ * @param linesize size of a single line of plane data, including padding
+ * @param mc_func motion compensation function pointers (bilinear or sixtap MC)
+ */
 static av_always_inline
 void vp8_mc_chroma(VP8Context *s, uint8_t *dst1, uint8_t *dst2, AVFrame *ref,
                    const VP56mv *mv, int x_off, int y_off,
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c
index 1fcce36f08..d98c6a0b85 100644
--- a/libavcodec/x86/dsputil_mmx.c
+++ b/libavcodec/x86/dsputil_mmx.c
@@ -2563,8 +2563,8 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
             SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, mmx2, );
             SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, mmx2, );
             }
-#if HAVE_YASM
             else if (bit_depth == 10) {
+#if HAVE_YASM
 #if !ARCH_X86_64
                 SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 10_mmxext, ff_);
                 SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_mmxext, ff_);
@@ -2573,8 +2573,8 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
 #endif
                 SET_QPEL_FUNCS(put_h264_qpel, 2, 4,  10_mmxext, ff_);
                 SET_QPEL_FUNCS(avg_h264_qpel, 2, 4,  10_mmxext, ff_);
-            }
 #endif
+            }
 
             SET_QPEL_FUNCS(put_2tap_qpel, 0, 16, mmx2, );
             SET_QPEL_FUNCS(put_2tap_qpel, 1, 8, mmx2, );
diff --git a/libavcodec/x86/h264_qpel_mmx.c b/libavcodec/x86/h264_qpel_mmx.c
index 66dee45134..807d8548d6 100644
--- a/libavcodec/x86/h264_qpel_mmx.c
+++ b/libavcodec/x86/h264_qpel_mmx.c
@@ -1294,6 +1294,6 @@ QPEL16_OP(mc31, MMX)\
 QPEL16_OP(mc32, MMX)\
 QPEL16_OP(mc33, MMX)
 
-#if HAVE_YASM && ARCH_X86_32 // ARCH_X86_64 implies sse2+
+#if ARCH_X86_32 && HAVE_YASM // ARCH_X86_64 implies sse2+
 QPEL16(mmxext)
 #endif