Merge remote-tracking branch 'qatar/master'

* qatar/master: (35 commits)
  flvdec: Do not call parse_keyframes_index with a NULL stream
  libspeexdec: include system headers before local headers
  libspeexdec: return meaningful error codes
  libspeexdec: cosmetics: reindent
  libspeexdec: decode one frame at a time.
  swscale: fix signed shift overflows in ff_yuv2rgb_c_init_tables()
  Move timefilter code from lavf to lavd.
  mov: add support for hdvd and pgapmetadata atoms
  mov: rename function _stik, some indentation cosmetics
  mov: rename function _int8 to remove ambiguity, some indentation cosmetics
  mov: parse the gnre atom
  mp3on4: check for allocation failures in decode_init_mp3on4()
  mp3on4: create a separate flush function for MP3onMP4.
  mp3on4: ensure that the frame channel count does not exceed the codec channel count.
  mp3on4: set channel layout
  mp3on4: fix the output channel order
  mp3on4: allocate temp buffer with av_malloc() instead of on the stack.
  mp3on4: copy MPADSPContext from first context to all contexts.
  fmtconvert: port float_to_int16_interleave() 2-channel x86 inline asm to yasm
  fmtconvert: port int32_to_float_fmul_scalar() x86 inline asm to yasm
  ...

Conflicts:
	libavcodec/arm/h264dsp_init_arm.c
	libavcodec/h264.c
	libavcodec/h264.h
	libavcodec/h264_cabac.c
	libavcodec/h264_cavlc.c
	libavcodec/h264_ps.c
	libavcodec/h264dsp_template.c
	libavcodec/h264idct_template.c
	libavcodec/h264pred.c
	libavcodec/h264pred_template.c
	libavcodec/x86/h264dsp_mmx.c
	libavdevice/Makefile
	libavdevice/jack_audio.c
	libavformat/Makefile
	libavformat/flvdec.c
	libavformat/flvenc.c
	libavutil/pixfmt.h
	libswscale/utils.c

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

6 files changed, 360 insertions, 533 deletions

diff --git a/libavcodec/x86/dsputil_yasm.asm b/libavcodec/x86/dsputil_yasm.asm
index 70f1b86c44..6627d21bd8 100644
--- a/libavcodec/x86/dsputil_yasm.asm
+++ b/libavcodec/x86/dsputil_yasm.asm
@@ -1055,14 +1055,6 @@ emu_edge mmx
 ;                           int32_t max, unsigned int len)
 ;-----------------------------------------------------------------------------
 
-%macro SPLATD_MMX 1
-    punpckldq  %1, %1
-%endmacro
-
-%macro SPLATD_SSE2 1
-    pshufd  %1, %1, 0
-%endmacro
-
 %macro VECTOR_CLIP_INT32 4
 cglobal vector_clip_int32_%1, 5,5,%2, dst, src, min, max, len
 %ifidn %1, sse2
diff --git a/libavcodec/x86/fmtconvert.asm b/libavcodec/x86/fmtconvert.asm
index 2deb577ca6..37e7a094ce 100644
--- a/libavcodec/x86/fmtconvert.asm
+++ b/libavcodec/x86/fmtconvert.asm
@@ -24,6 +24,146 @@
 
 SECTION_TEXT
 
+;---------------------------------------------------------------------------------
+; void int32_to_float_fmul_scalar(float *dst, const int *src, float mul, int len);
+;---------------------------------------------------------------------------------
+%macro INT32_TO_FLOAT_FMUL_SCALAR 2
+%ifdef ARCH_X86_64
+cglobal int32_to_float_fmul_scalar_%1, 3,3,%2, dst, src, len
+%else
+cglobal int32_to_float_fmul_scalar_%1, 4,4,%2, dst, src, mul, len
+    movss   m0, mulm
+%endif
+    SPLATD  m0
+    shl     lenq, 2
+    add     srcq, lenq
+    add     dstq, lenq
+    neg     lenq
+.loop:
+%ifidn %1, sse2
+    cvtdq2ps  m1, [srcq+lenq   ]
+    cvtdq2ps  m2, [srcq+lenq+16]
+%else
+    cvtpi2ps  m1, [srcq+lenq   ]
+    cvtpi2ps  m3, [srcq+lenq+ 8]
+    cvtpi2ps  m2, [srcq+lenq+16]
+    cvtpi2ps  m4, [srcq+lenq+24]
+    movlhps   m1, m3
+    movlhps   m2, m4
+%endif
+    mulps     m1, m0
+    mulps     m2, m0
+    mova  [dstq+lenq   ], m1
+    mova  [dstq+lenq+16], m2
+    add     lenq, 32
+    jl .loop
+    REP_RET
+%endmacro
+
+INIT_XMM
+%define SPLATD SPLATD_SSE
+%define movdqa movaps
+INT32_TO_FLOAT_FMUL_SCALAR sse, 5
+%undef movdqa
+%define SPLATD SPLATD_SSE2
+INT32_TO_FLOAT_FMUL_SCALAR sse2, 3
+%undef SPLATD
+
+
+;------------------------------------------------------------------------------
+; void ff_float_to_int16(int16_t *dst, const float *src, long len);
+;------------------------------------------------------------------------------
+%macro FLOAT_TO_INT16 2
+cglobal float_to_int16_%1, 3,3,%2, dst, src, len
+    add       lenq, lenq
+    lea       srcq, [srcq+2*lenq]
+    add       dstq, lenq
+    neg       lenq
+.loop:
+%ifidn %1, sse2
+    cvtps2dq    m0, [srcq+2*lenq   ]
+    cvtps2dq    m1, [srcq+2*lenq+16]
+    packssdw    m0, m1
+    mova  [dstq+lenq], m0
+%else
+    cvtps2pi    m0, [srcq+2*lenq   ]
+    cvtps2pi    m1, [srcq+2*lenq+ 8]
+    cvtps2pi    m2, [srcq+2*lenq+16]
+    cvtps2pi    m3, [srcq+2*lenq+24]
+    packssdw    m0, m1
+    packssdw    m2, m3
+    mova  [dstq+lenq  ], m0
+    mova  [dstq+lenq+8], m2
+%endif
+    add       lenq, 16
+    js .loop
+%ifnidn %1, sse2
+    emms
+%endif
+    REP_RET
+%endmacro
+
+INIT_XMM
+FLOAT_TO_INT16 sse2, 2
+INIT_MMX
+FLOAT_TO_INT16 sse, 0
+%define cvtps2pi pf2id
+FLOAT_TO_INT16 3dnow, 0
+%undef cvtps2pi
+
+
+;-------------------------------------------------------------------------------
+; void ff_float_to_int16_interleave2(int16_t *dst, const float **src, long len);
+;-------------------------------------------------------------------------------
+%macro FLOAT_TO_INT16_INTERLEAVE2 1
+cglobal float_to_int16_interleave2_%1, 3,4,2, dst, src0, src1, len
+    lea      lenq, [4*r2q]
+    mov     src1q, [src0q+gprsize]
+    mov     src0q, [src0q]
+    add      dstq, lenq
+    add     src0q, lenq
+    add     src1q, lenq
+    neg      lenq
+.loop:
+%ifidn %1, sse2
+    cvtps2dq   m0, [src0q+lenq]
+    cvtps2dq   m1, [src1q+lenq]
+    packssdw   m0, m1
+    movhlps    m1, m0
+    punpcklwd  m0, m1
+    mova  [dstq+lenq], m0
+%else
+    cvtps2pi   m0, [src0q+lenq  ]
+    cvtps2pi   m1, [src0q+lenq+8]
+    cvtps2pi   m2, [src1q+lenq  ]
+    cvtps2pi   m3, [src1q+lenq+8]
+    packssdw   m0, m1
+    packssdw   m2, m3
+    mova       m1, m0
+    punpcklwd  m0, m2
+    punpckhwd  m1, m2
+    mova  [dstq+lenq  ], m0
+    mova  [dstq+lenq+8], m1
+%endif
+    add      lenq, 16
+    js .loop
+%ifnidn %1, sse2
+    emms
+%endif
+    REP_RET
+%endmacro
+
+INIT_MMX
+%define cvtps2pi pf2id
+FLOAT_TO_INT16_INTERLEAVE2 3dnow
+%undef cvtps2pi
+%define movdqa movaps
+FLOAT_TO_INT16_INTERLEAVE2 sse
+%undef movdqa
+INIT_XMM
+FLOAT_TO_INT16_INTERLEAVE2 sse2
+
+
 %macro PSWAPD_SSE 2
     pshufw %1, %2, 0x4e
 %endmacro
diff --git a/libavcodec/x86/fmtconvert_mmx.c b/libavcodec/x86/fmtconvert_mmx.c
index ba2c2c9bd5..a3d8f89816 100644
--- a/libavcodec/x86/fmtconvert_mmx.c
+++ b/libavcodec/x86/fmtconvert_mmx.c
@@ -26,133 +26,32 @@
 #include "libavutil/x86_cpu.h"
 #include "libavcodec/fmtconvert.h"
 
-static void int32_to_float_fmul_scalar_sse(float *dst, const int *src, float mul, int len)
-{
-    x86_reg i = -4*len;
-    __asm__ volatile(
-        "movss  %3, %%xmm4 \n"
-        "shufps $0, %%xmm4, %%xmm4 \n"
-        "1: \n"
-        "cvtpi2ps   (%2,%0), %%xmm0 \n"
-        "cvtpi2ps  8(%2,%0), %%xmm1 \n"
-        "cvtpi2ps 16(%2,%0), %%xmm2 \n"
-        "cvtpi2ps 24(%2,%0), %%xmm3 \n"
-        "movlhps  %%xmm1,    %%xmm0 \n"
-        "movlhps  %%xmm3,    %%xmm2 \n"
-        "mulps    %%xmm4,    %%xmm0 \n"
-        "mulps    %%xmm4,    %%xmm2 \n"
-        "movaps   %%xmm0,   (%1,%0) \n"
-        "movaps   %%xmm2, 16(%1,%0) \n"
-        "add $32, %0 \n"
-        "jl 1b \n"
-        :"+r"(i)
-        :"r"(dst+len), "r"(src+len), "m"(mul)
-    );
-}
-
-static void int32_to_float_fmul_scalar_sse2(float *dst, const int *src, float mul, int len)
-{
-    x86_reg i = -4*len;
-    __asm__ volatile(
-        "movss  %3, %%xmm4 \n"
-        "shufps $0, %%xmm4, %%xmm4 \n"
-        "1: \n"
-        "cvtdq2ps   (%2,%0), %%xmm0 \n"
-        "cvtdq2ps 16(%2,%0), %%xmm1 \n"
-        "mulps    %%xmm4,    %%xmm0 \n"
-        "mulps    %%xmm4,    %%xmm1 \n"
-        "movaps   %%xmm0,   (%1,%0) \n"
-        "movaps   %%xmm1, 16(%1,%0) \n"
-        "add $32, %0 \n"
-        "jl 1b \n"
-        :"+r"(i)
-        :"r"(dst+len), "r"(src+len), "m"(mul)
-    );
-}
+#if HAVE_YASM
 
-static void float_to_int16_3dnow(int16_t *dst, const float *src, long len){
-    x86_reg reglen = len;
-    // not bit-exact: pf2id uses different rounding than C and SSE
-    __asm__ volatile(
-        "add        %0          , %0        \n\t"
-        "lea         (%2,%0,2)  , %2        \n\t"
-        "add        %0          , %1        \n\t"
-        "neg        %0                      \n\t"
-        "1:                                 \n\t"
-        "pf2id       (%2,%0,2)  , %%mm0     \n\t"
-        "pf2id      8(%2,%0,2)  , %%mm1     \n\t"
-        "pf2id     16(%2,%0,2)  , %%mm2     \n\t"
-        "pf2id     24(%2,%0,2)  , %%mm3     \n\t"
-        "packssdw   %%mm1       , %%mm0     \n\t"
-        "packssdw   %%mm3       , %%mm2     \n\t"
-        "movq       %%mm0       ,  (%1,%0)  \n\t"
-        "movq       %%mm2       , 8(%1,%0)  \n\t"
-        "add        $16         , %0        \n\t"
-        " js 1b                             \n\t"
-        "femms                              \n\t"
-        :"+r"(reglen), "+r"(dst), "+r"(src)
-    );
-}
+void ff_int32_to_float_fmul_scalar_sse (float *dst, const int *src, float mul, int len);
+void ff_int32_to_float_fmul_scalar_sse2(float *dst, const int *src, float mul, int len);
 
-static void float_to_int16_sse(int16_t *dst, const float *src, long len){
-    x86_reg reglen = len;
-    __asm__ volatile(
-        "add        %0          , %0        \n\t"
-        "lea         (%2,%0,2)  , %2        \n\t"
-        "add        %0          , %1        \n\t"
-        "neg        %0                      \n\t"
-        "1:                                 \n\t"
-        "cvtps2pi    (%2,%0,2)  , %%mm0     \n\t"
-        "cvtps2pi   8(%2,%0,2)  , %%mm1     \n\t"
-        "cvtps2pi  16(%2,%0,2)  , %%mm2     \n\t"
-        "cvtps2pi  24(%2,%0,2)  , %%mm3     \n\t"
-        "packssdw   %%mm1       , %%mm0     \n\t"
-        "packssdw   %%mm3       , %%mm2     \n\t"
-        "movq       %%mm0       ,  (%1,%0)  \n\t"
-        "movq       %%mm2       , 8(%1,%0)  \n\t"
-        "add        $16         , %0        \n\t"
-        " js 1b                             \n\t"
-        "emms                               \n\t"
-        :"+r"(reglen), "+r"(dst), "+r"(src)
-    );
-}
+void ff_float_to_int16_3dnow(int16_t *dst, const float *src, long len);
+void ff_float_to_int16_sse  (int16_t *dst, const float *src, long len);
+void ff_float_to_int16_sse2 (int16_t *dst, const float *src, long len);
 
-static void float_to_int16_sse2(int16_t *dst, const float *src, long len){
-    x86_reg reglen = len;
-    __asm__ volatile(
-        "add        %0          , %0        \n\t"
-        "lea         (%2,%0,2)  , %2        \n\t"
-        "add        %0          , %1        \n\t"
-        "neg        %0                      \n\t"
-        "1:                                 \n\t"
-        "cvtps2dq    (%2,%0,2)  , %%xmm0    \n\t"
-        "cvtps2dq  16(%2,%0,2)  , %%xmm1    \n\t"
-        "packssdw   %%xmm1      , %%xmm0    \n\t"
-        "movdqa     %%xmm0      ,  (%1,%0)  \n\t"
-        "add        $16         , %0        \n\t"
-        " js 1b                             \n\t"
-        :"+r"(reglen), "+r"(dst), "+r"(src)
-    );
-}
+void ff_float_to_int16_interleave2_3dnow(int16_t *dst, const float **src, long len);
+void ff_float_to_int16_interleave2_sse  (int16_t *dst, const float **src, long len);
+void ff_float_to_int16_interleave2_sse2 (int16_t *dst, const float **src, long len);
 
 void ff_float_to_int16_interleave6_sse(int16_t *dst, const float **src, int len);
 void ff_float_to_int16_interleave6_3dnow(int16_t *dst, const float **src, int len);
 void ff_float_to_int16_interleave6_3dn2(int16_t *dst, const float **src, int len);
 
-#if !HAVE_YASM
-#define ff_float_to_int16_interleave6_sse(a,b,c)   float_to_int16_interleave_misc_sse(a,b,c,6)
-#define ff_float_to_int16_interleave6_3dnow(a,b,c) float_to_int16_interleave_misc_3dnow(a,b,c,6)
-#define ff_float_to_int16_interleave6_3dn2(a,b,c)  float_to_int16_interleave_misc_3dnow(a,b,c,6)
-#endif
 #define ff_float_to_int16_interleave6_sse2 ff_float_to_int16_interleave6_sse
 
-#define FLOAT_TO_INT16_INTERLEAVE(cpu, body) \
+#define FLOAT_TO_INT16_INTERLEAVE(cpu) \
 /* gcc pessimizes register allocation if this is in the same function as float_to_int16_interleave_sse2*/\
 static av_noinline void float_to_int16_interleave_misc_##cpu(int16_t *dst, const float **src, long len, int channels){\
     DECLARE_ALIGNED(16, int16_t, tmp)[len];\
     int i,j,c;\
     for(c=0; c<channels; c++){\
-        float_to_int16_##cpu(tmp, src[c], len);\
+        ff_float_to_int16_##cpu(tmp, src[c], len);\
         for(i=0, j=c; i<len; i++, j+=channels)\
             dst[j] = tmp[i];\
     }\
@@ -160,73 +59,18 @@ static av_noinline void float_to_int16_interleave_misc_##cpu(int16_t *dst, const
 \
 static void float_to_int16_interleave_##cpu(int16_t *dst, const float **src, long len, int channels){\
     if(channels==1)\
-        float_to_int16_##cpu(dst, src[0], len);\
+        ff_float_to_int16_##cpu(dst, src[0], len);\
     else if(channels==2){\
-        x86_reg reglen = len; \
-        const float *src0 = src[0];\
-        const float *src1 = src[1];\
-        __asm__ volatile(\
-            "shl $2, %0 \n"\
-            "add %0, %1 \n"\
-            "add %0, %2 \n"\
-            "add %0, %3 \n"\
-            "neg %0 \n"\
-            body\
-            :"+r"(reglen), "+r"(dst), "+r"(src0), "+r"(src1)\
-        );\
+        ff_float_to_int16_interleave2_##cpu(dst, src, len);\
     }else if(channels==6){\
         ff_float_to_int16_interleave6_##cpu(dst, src, len);\
     }else\
         float_to_int16_interleave_misc_##cpu(dst, src, len, channels);\
 }
 
-FLOAT_TO_INT16_INTERLEAVE(3dnow,
-    "1:                         \n"
-    "pf2id     (%2,%0), %%mm0   \n"
-    "pf2id    8(%2,%0), %%mm1   \n"
-    "pf2id     (%3,%0), %%mm2   \n"
-    "pf2id    8(%3,%0), %%mm3   \n"
-    "packssdw    %%mm1, %%mm0   \n"
-    "packssdw    %%mm3, %%mm2   \n"
-    "movq        %%mm0, %%mm1   \n"
-    "punpcklwd   %%mm2, %%mm0   \n"
-    "punpckhwd   %%mm2, %%mm1   \n"
-    "movq        %%mm0,  (%1,%0)\n"
-    "movq        %%mm1, 8(%1,%0)\n"
-    "add $16, %0                \n"
-    "js 1b                      \n"
-    "femms                      \n"
-)
-
-FLOAT_TO_INT16_INTERLEAVE(sse,
-    "1:                         \n"
-    "cvtps2pi  (%2,%0), %%mm0   \n"
-    "cvtps2pi 8(%2,%0), %%mm1   \n"
-    "cvtps2pi  (%3,%0), %%mm2   \n"
-    "cvtps2pi 8(%3,%0), %%mm3   \n"
-    "packssdw    %%mm1, %%mm0   \n"
-    "packssdw    %%mm3, %%mm2   \n"
-    "movq        %%mm0, %%mm1   \n"
-    "punpcklwd   %%mm2, %%mm0   \n"
-    "punpckhwd   %%mm2, %%mm1   \n"
-    "movq        %%mm0,  (%1,%0)\n"
-    "movq        %%mm1, 8(%1,%0)\n"
-    "add $16, %0                \n"
-    "js 1b                      \n"
-    "emms                       \n"
-)
-
-FLOAT_TO_INT16_INTERLEAVE(sse2,
-    "1:                         \n"
-    "cvtps2dq  (%2,%0), %%xmm0  \n"
-    "cvtps2dq  (%3,%0), %%xmm1  \n"
-    "packssdw   %%xmm1, %%xmm0  \n"
-    "movhlps    %%xmm0, %%xmm1  \n"
-    "punpcklwd  %%xmm1, %%xmm0  \n"
-    "movdqa     %%xmm0, (%1,%0) \n"
-    "add $16, %0                \n"
-    "js 1b                      \n"
-)
+FLOAT_TO_INT16_INTERLEAVE(3dnow)
+FLOAT_TO_INT16_INTERLEAVE(sse)
+FLOAT_TO_INT16_INTERLEAVE(sse2)
 
 static void float_to_int16_interleave_3dn2(int16_t *dst, const float **src, long len, int channels){
     if(channels==6)
@@ -235,7 +79,6 @@ static void float_to_int16_interleave_3dn2(int16_t *dst, const float **src, long
         float_to_int16_interleave_3dnow(dst, src, len, channels);
 }
 
-#if HAVE_YASM
 void ff_float_interleave2_mmx(float *dst, const float **src, unsigned int len);
 void ff_float_interleave2_sse(float *dst, const float **src, unsigned int len);
 
@@ -269,34 +112,32 @@ void ff_fmt_convert_init_x86(FmtConvertContext *c, AVCodecContext *avctx)
 {
     int mm_flags = av_get_cpu_flags();
 
-    if (mm_flags & AV_CPU_FLAG_MMX) {
 #if HAVE_YASM
+    if (mm_flags & AV_CPU_FLAG_MMX) {
         c->float_interleave = float_interleave_mmx;
-#endif
 
-        if(mm_flags & AV_CPU_FLAG_3DNOW){
+        if (HAVE_AMD3DNOW && mm_flags & AV_CPU_FLAG_3DNOW) {
             if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
-                c->float_to_int16 = float_to_int16_3dnow;
+                c->float_to_int16 = ff_float_to_int16_3dnow;
                 c->float_to_int16_interleave = float_to_int16_interleave_3dnow;
             }
         }
-        if(mm_flags & AV_CPU_FLAG_3DNOWEXT){
+        if (HAVE_AMD3DNOWEXT && mm_flags & AV_CPU_FLAG_3DNOWEXT) {
             if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
                 c->float_to_int16_interleave = float_to_int16_interleave_3dn2;
             }
         }
-        if(mm_flags & AV_CPU_FLAG_SSE){
-            c->int32_to_float_fmul_scalar = int32_to_float_fmul_scalar_sse;
-            c->float_to_int16 = float_to_int16_sse;
+        if (HAVE_SSE && mm_flags & AV_CPU_FLAG_SSE) {
+            c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_sse;
+            c->float_to_int16 = ff_float_to_int16_sse;
             c->float_to_int16_interleave = float_to_int16_interleave_sse;
-#if HAVE_YASM
             c->float_interleave = float_interleave_sse;
-#endif
         }
-        if(mm_flags & AV_CPU_FLAG_SSE2){
-            c->int32_to_float_fmul_scalar = int32_to_float_fmul_scalar_sse2;
-            c->float_to_int16 = float_to_int16_sse2;
+        if (HAVE_SSE && mm_flags & AV_CPU_FLAG_SSE2) {
+            c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_sse2;
+            c->float_to_int16 = ff_float_to_int16_sse2;
             c->float_to_int16_interleave = float_to_int16_interleave_sse2;
         }
     }
+#endif
 }
diff --git a/libavcodec/x86/h264_weight.asm b/libavcodec/x86/h264_weight.asm
index bb0af86097..cc96cb1f3b 100644
--- a/libavcodec/x86/h264_weight.asm
+++ b/libavcodec/x86/h264_weight.asm
@@ -28,21 +28,20 @@ SECTION .text
 ;-----------------------------------------------------------------------------
 ; biweight pred:
 ;
-; void h264_biweight_16x16_sse2(uint8_t *dst, uint8_t *src, int stride,
-;                               int log2_denom, int weightd, int weights,
-;                               int offset);
+; void h264_biweight_16_sse2(uint8_t *dst, uint8_t *src, int stride,
+;                            int height, int log2_denom, int weightd,
+;                            int weights, int offset);
 ; and
-; void h264_weight_16x16_sse2(uint8_t *dst, int stride,
-;                             int log2_denom, int weight,
-;                             int offset);
+; void h264_weight_16_sse2(uint8_t *dst, int stride, int height,
+;                          int log2_denom, int weight, int offset);
 ;-----------------------------------------------------------------------------
 
 %macro WEIGHT_SETUP 0
-    add        r4, r4
-    inc        r4
-    movd       m3, r3d
-    movd       m5, r4d
-    movd       m6, r2d
+    add        r5, r5
+    inc        r5
+    movd       m3, r4d
+    movd       m5, r5d
+    movd       m6, r3d
     pslld      m5, m6
     psrld      m5, 1
 %if mmsize == 16
@@ -71,60 +70,41 @@ SECTION .text
     packuswb      m0, m1
 %endmacro
 
-%macro WEIGHT_FUNC_DBL_MM 1
-cglobal h264_weight_16x%1_mmx2, 5, 5, 0
+INIT_MMX
+cglobal h264_weight_16_mmx2, 6, 6, 0
     WEIGHT_SETUP
-    mov        r2, %1
-%if %1 == 16
 .nextrow
     WEIGHT_OP 0,  4
     mova     [r0  ], m0
     WEIGHT_OP 8, 12
     mova     [r0+8], m0
     add        r0, r1
-    dec        r2
+    dec        r2d
     jnz .nextrow
     REP_RET
-%else
-    jmp mangle(ff_h264_weight_16x16_mmx2.nextrow)
-%endif
-%endmacro
 
-INIT_MMX
-WEIGHT_FUNC_DBL_MM 16
-WEIGHT_FUNC_DBL_MM  8
-
-%macro WEIGHT_FUNC_MM 4
-cglobal h264_weight_%1x%2_%4, 7, 7, %3
+%macro WEIGHT_FUNC_MM 3
+cglobal h264_weight_%1_%3, 6, 6, %2
     WEIGHT_SETUP
-    mov        r2, %2
-%if %2 == 16
 .nextrow
     WEIGHT_OP 0, mmsize/2
     mova     [r0], m0
     add        r0, r1
-    dec        r2
+    dec        r2d
     jnz .nextrow
     REP_RET
-%else
-    jmp mangle(ff_h264_weight_%1x16_%4.nextrow)
-%endif
 %endmacro
 
 INIT_MMX
-WEIGHT_FUNC_MM  8, 16,  0, mmx2
-WEIGHT_FUNC_MM  8,  8,  0, mmx2
-WEIGHT_FUNC_MM  8,  4,  0, mmx2
+WEIGHT_FUNC_MM  8, 0, mmx2
 INIT_XMM
-WEIGHT_FUNC_MM 16, 16,  8, sse2
-WEIGHT_FUNC_MM 16,  8,  8, sse2
+WEIGHT_FUNC_MM 16, 8, sse2
 
-%macro WEIGHT_FUNC_HALF_MM 5
-cglobal h264_weight_%1x%2_%5, 5, 5, %4
+%macro WEIGHT_FUNC_HALF_MM 3
+cglobal h264_weight_%1_%3, 6, 6, %2
     WEIGHT_SETUP
-    mov        r2, %2/2
+    sar       r2d, 1
     lea        r3, [r1*2]
-%if %2 == mmsize
 .nextrow
     WEIGHT_OP 0, r1
     movh     [r0], m0
@@ -135,31 +115,34 @@ cglobal h264_weight_%1x%2_%5, 5, 5, %4
     movh     [r0+r1], m0
 %endif
     add        r0, r3
-    dec        r2
+    dec        r2d
     jnz .nextrow
     REP_RET
-%else
-    jmp mangle(ff_h264_weight_%1x%3_%5.nextrow)
-%endif
 %endmacro
 
 INIT_MMX
-WEIGHT_FUNC_HALF_MM 4,  8,  8, 0, mmx2
-WEIGHT_FUNC_HALF_MM 4,  4,  8, 0, mmx2
-WEIGHT_FUNC_HALF_MM 4,  2,  8, 0, mmx2
+WEIGHT_FUNC_HALF_MM 4, 0, mmx2
+WEIGHT_FUNC_HALF_MM 4, 0, mmx2
+WEIGHT_FUNC_HALF_MM 4, 0, mmx2
 INIT_XMM
-WEIGHT_FUNC_HALF_MM 8, 16, 16, 8, sse2
-WEIGHT_FUNC_HALF_MM 8,  8, 16, 8, sse2
-WEIGHT_FUNC_HALF_MM 8,  4, 16, 8, sse2
+WEIGHT_FUNC_HALF_MM 8, 8, sse2
+WEIGHT_FUNC_HALF_MM 8, 8, sse2
+WEIGHT_FUNC_HALF_MM 8, 8, sse2
 
 %macro BIWEIGHT_SETUP 0
-    add        r6, 1
-    or         r6, 1
-    add        r3, 1
-    movd       m3, r4d
-    movd       m4, r5d
-    movd       m5, r6d
-    movd       m6, r3d
+%ifdef ARCH_X86_64
+%define off_regd r11d
+%else
+%define off_regd r3d
+%endif
+    mov  off_regd, r7m
+    add  off_regd, 1
+    or   off_regd, 1
+    add        r4, 1
+    movd       m3, r5d
+    movd       m4, r6d
+    movd       m5, off_regd
+    movd       m6, r4d
     pslld      m5, m6
     psrld      m5, 1
 %if mmsize == 16
@@ -195,11 +178,10 @@ WEIGHT_FUNC_HALF_MM 8,  4, 16, 8, sse2
     packuswb   m0, m1
 %endmacro
 
-%macro BIWEIGHT_FUNC_DBL_MM 1
-cglobal h264_biweight_16x%1_mmx2, 7, 7, 0
+INIT_MMX
+cglobal h264_biweight_16_mmx2, 7, 7, 0
     BIWEIGHT_SETUP
-    mov        r3, %1
-%if %1 == 16
+    movifnidn r3d, r3m
 .nextrow
     BIWEIGHT_STEPA 0, 1, 0
     BIWEIGHT_STEPA 1, 2, 4
@@ -211,23 +193,14 @@ cglobal h264_biweight_16x%1_mmx2, 7, 7, 0
     mova     [r0+8], m0
     add        r0, r2
     add        r1, r2
-    dec        r3
+    dec        r3d
     jnz .nextrow
     REP_RET
-%else
-    jmp mangle(ff_h264_biweight_16x16_mmx2.nextrow)
-%endif
-%endmacro
 
-INIT_MMX
-BIWEIGHT_FUNC_DBL_MM 16
-BIWEIGHT_FUNC_DBL_MM  8
-
-%macro BIWEIGHT_FUNC_MM 4
-cglobal h264_biweight_%1x%2_%4, 7, 7, %3
+%macro BIWEIGHT_FUNC_MM 3
+cglobal h264_biweight_%1_%3, 7, 7, %2
     BIWEIGHT_SETUP
-    mov        r3, %2
-%if %2 == 16
+    movifnidn r3d, r3m
 .nextrow
     BIWEIGHT_STEPA 0, 1, 0
     BIWEIGHT_STEPA 1, 2, mmsize/2
@@ -235,28 +208,22 @@ cglobal h264_biweight_%1x%2_%4, 7, 7, %3
     mova       [r0], m0
     add        r0, r2
     add        r1, r2
-    dec        r3
+    dec        r3d
     jnz .nextrow
     REP_RET
-%else
-    jmp mangle(ff_h264_biweight_%1x16_%4.nextrow)
-%endif
 %endmacro
 
 INIT_MMX
-BIWEIGHT_FUNC_MM  8, 16,  0, mmx2
-BIWEIGHT_FUNC_MM  8,  8,  0, mmx2
-BIWEIGHT_FUNC_MM  8,  4,  0, mmx2
+BIWEIGHT_FUNC_MM  8, 0, mmx2
 INIT_XMM
-BIWEIGHT_FUNC_MM 16, 16,  8, sse2
-BIWEIGHT_FUNC_MM 16,  8,  8, sse2
+BIWEIGHT_FUNC_MM 16, 8, sse2
 
-%macro BIWEIGHT_FUNC_HALF_MM 5
-cglobal h264_biweight_%1x%2_%5, 7, 7, %4
+%macro BIWEIGHT_FUNC_HALF_MM 3
+cglobal h264_biweight_%1_%3, 7, 7, %2
     BIWEIGHT_SETUP
-    mov        r3, %2/2
+    movifnidn r3d, r3m
+    sar        r3, 1
     lea        r4, [r2*2]
-%if %2 == mmsize
 .nextrow
     BIWEIGHT_STEPA 0, 1, 0
     BIWEIGHT_STEPA 1, 2, r2
@@ -270,31 +237,30 @@ cglobal h264_biweight_%1x%2_%5, 7, 7, %4
 %endif
     add        r0, r4
     add        r1, r4
-    dec        r3
+    dec        r3d
     jnz .nextrow
     REP_RET
-%else
-    jmp mangle(ff_h264_biweight_%1x%3_%5.nextrow)
-%endif
 %endmacro
 
 INIT_MMX
-BIWEIGHT_FUNC_HALF_MM 4,  8,  8, 0, mmx2
-BIWEIGHT_FUNC_HALF_MM 4,  4,  8, 0, mmx2
-BIWEIGHT_FUNC_HALF_MM 4,  2,  8, 0, mmx2
+BIWEIGHT_FUNC_HALF_MM 4, 0, mmx2
 INIT_XMM
-BIWEIGHT_FUNC_HALF_MM 8, 16, 16, 8, sse2
-BIWEIGHT_FUNC_HALF_MM 8,  8, 16, 8, sse2
-BIWEIGHT_FUNC_HALF_MM 8,  4, 16, 8, sse2
+BIWEIGHT_FUNC_HALF_MM 8, 8, sse2
 
 %macro BIWEIGHT_SSSE3_SETUP 0
-    add        r6, 1
-    or         r6, 1
-    add        r3, 1
-    movd       m4, r4d
-    movd       m0, r5d
-    movd       m5, r6d
-    movd       m6, r3d
+%ifdef ARCH_X86_64
+%define off_regd r11d
+%else
+%define off_regd r3d
+%endif
+    mov  off_regd, r7m
+    add  off_regd, 1
+    or   off_regd, 1
+    add        r4, 1
+    movd       m4, r5d
+    movd       m0, r6d
+    movd       m5, off_regd
+    movd       m6, r4d
     pslld      m5, m6
     psrld      m5, 1
     punpcklbw  m4, m0
@@ -314,12 +280,11 @@ BIWEIGHT_FUNC_HALF_MM 8,  4, 16, 8, sse2
     packuswb   m0, m2
 %endmacro
 
-%macro BIWEIGHT_SSSE3_16 1
-cglobal h264_biweight_16x%1_ssse3, 7, 7, 8
+INIT_XMM
+cglobal h264_biweight_16_ssse3, 7, 7, 8
     BIWEIGHT_SSSE3_SETUP
-    mov        r3, %1
+    movifnidn r3d, r3m
 
-%if %1 == 16
 .nextrow
     movh       m0, [r0]
     movh       m2, [r0+8]
@@ -330,25 +295,17 @@ cglobal h264_biweight_16x%1_ssse3, 7, 7, 8
     mova       [r0], m0
     add        r0, r2
     add        r1, r2
-    dec        r3
+    dec        r3d
     jnz .nextrow
     REP_RET
-%else
-    jmp mangle(ff_h264_biweight_16x16_ssse3.nextrow)
-%endif
-%endmacro
 
 INIT_XMM
-BIWEIGHT_SSSE3_16 16
-BIWEIGHT_SSSE3_16  8
-
-%macro BIWEIGHT_SSSE3_8 1
-cglobal h264_biweight_8x%1_ssse3, 7, 7, 8
+cglobal h264_biweight_8_ssse3, 7, 7, 8
     BIWEIGHT_SSSE3_SETUP
-    mov        r3, %1/2
+    movifnidn r3d, r3m
+    sar        r3, 1
     lea        r4, [r2*2]
 
-%if %1 == 16
 .nextrow
     movh       m0, [r0]
     movh       m1, [r1]
@@ -361,15 +318,6 @@ cglobal h264_biweight_8x%1_ssse3, 7, 7, 8
     movhps     [r0+r2], m0
     add        r0, r4
     add        r1, r4
-    dec        r3
+    dec        r3d
     jnz .nextrow
     REP_RET
-%else
-    jmp mangle(ff_h264_biweight_8x16_ssse3.nextrow)
-%endif
-%endmacro
-
-INIT_XMM
-BIWEIGHT_SSSE3_8 16
-BIWEIGHT_SSSE3_8  8
-BIWEIGHT_SSSE3_8  4
diff --git a/libavcodec/x86/h264_weight_10bit.asm b/libavcodec/x86/h264_weight_10bit.asm
index 1c58d72d94..20df6fbab5 100644
--- a/libavcodec/x86/h264_weight_10bit.asm
+++ b/libavcodec/x86/h264_weight_10bit.asm
@@ -36,33 +36,26 @@ cextern pw_1
 SECTION .text
 
 ;-----------------------------------------------------------------------------
-; void h264_weight(uint8_t *dst, int stride, int log2_denom,
+; void h264_weight(uint8_t *dst, int stride, int height, int log2_denom,
 ;                  int weight, int offset);
 ;-----------------------------------------------------------------------------
-%ifdef ARCH_X86_32
-DECLARE_REG_TMP 2
-%else
-DECLARE_REG_TMP 10
-%endif
-
-%macro WEIGHT_PROLOGUE 1
-    mov t0, %1
+%macro WEIGHT_PROLOGUE 0
 .prologue
-    PROLOGUE 0,5,8
+    PROLOGUE 0,6,8
     movifnidn  r0, r0mp
     movifnidn r1d, r1m
-    movifnidn r3d, r3m
     movifnidn r4d, r4m
+    movifnidn r5d, r5m
 %endmacro
 
 %macro WEIGHT_SETUP 1
     mova       m0, [pw_1]
-    movd       m2, r2m
+    movd       m2, r3m
     pslld      m0, m2       ; 1<<log2_denom
     SPLATW     m0, m0
-    shl        r4, 19       ; *8, move to upper half of dword
-    lea        r4, [r4+r3*2+0x10000]
-    movd       m3, r4d      ; weight<<1 | 1+(offset<<(3))
+    shl        r5, 19       ; *8, move to upper half of dword
+    lea        r5, [r5+r4*2+0x10000]
+    movd       m3, r5d      ; weight<<1 | 1+(offset<<(3))
     pshufd     m3, m3, 0
     mova       m4, [pw_pixel_max]
     paddw      m2, [sq_1]   ; log2_denom+1
@@ -96,8 +89,8 @@ DECLARE_REG_TMP 10
 %endmacro
 
 %macro WEIGHT_FUNC_DBL 1
-cglobal h264_weight_16x16_10_%1
-    WEIGHT_PROLOGUE 16
+cglobal h264_weight_16_10_%1
+    WEIGHT_PROLOGUE
     WEIGHT_SETUP %1
 .nextrow
     WEIGHT_OP %1,  0
@@ -105,13 +98,9 @@ cglobal h264_weight_16x16_10_%1
     WEIGHT_OP %1, 16
     mova [r0+16], m5
     add       r0, r1
-    dec       t0
+    dec       r2d
     jnz .nextrow
     REP_RET
-
-cglobal h264_weight_16x8_10_%1
-    mov t0, 8
-    jmp mangle(ff_h264_weight_16x16_10_%1.prologue)
 %endmacro
 
 INIT_XMM
@@ -120,24 +109,16 @@ WEIGHT_FUNC_DBL sse4
 
 
 %macro WEIGHT_FUNC_MM 1
-cglobal h264_weight_8x16_10_%1
-    WEIGHT_PROLOGUE 16
+cglobal h264_weight_8_10_%1
+    WEIGHT_PROLOGUE
     WEIGHT_SETUP %1
 .nextrow
     WEIGHT_OP  %1, 0
     mova     [r0], m5
     add        r0, r1
-    dec        t0
+    dec        r2d
     jnz .nextrow
     REP_RET
-
-cglobal h264_weight_8x8_10_%1
-    mov t0, 8
-    jmp mangle(ff_h264_weight_8x16_10_%1.prologue)
-
-cglobal h264_weight_8x4_10_%1
-    mov t0, 4
-    jmp mangle(ff_h264_weight_8x16_10_%1.prologue)
 %endmacro
 
 INIT_XMM
@@ -146,8 +127,9 @@ WEIGHT_FUNC_MM sse4
 
 
 %macro WEIGHT_FUNC_HALF_MM 1
-cglobal h264_weight_4x8_10_%1
-    WEIGHT_PROLOGUE 4
+cglobal h264_weight_4_10_%1
+    WEIGHT_PROLOGUE
+    sar         r2d, 1
     WEIGHT_SETUP %1
     lea         r3, [r1*2]
 .nextrow
@@ -155,17 +137,9 @@ cglobal h264_weight_4x8_10_%1
     movh      [r0], m5
     movhps [r0+r1], m5
     add         r0, r3
-    dec         t0
+    dec         r2d
     jnz .nextrow
     REP_RET
-
-cglobal h264_weight_4x4_10_%1
-    mov t0, 2
-    jmp mangle(ff_h264_weight_4x8_10_%1.prologue)
-
-cglobal h264_weight_4x2_10_%1
-    mov t0, 1
-    jmp mangle(ff_h264_weight_4x8_10_%1.prologue)
 %endmacro
 
 INIT_XMM
@@ -174,40 +148,40 @@ WEIGHT_FUNC_HALF_MM sse4
 
 
 ;-----------------------------------------------------------------------------
-; void h264_biweight(uint8_t *dst, uint8_t *src, int stride, int log2_denom,
-;                    int weightd, int weights, int offset);
+; void h264_biweight(uint8_t *dst, uint8_t *src, int stride, int height,
+;                    int log2_denom, int weightd, int weights, int offset);
 ;-----------------------------------------------------------------------------
 %ifdef ARCH_X86_32
-DECLARE_REG_TMP 2,3
+DECLARE_REG_TMP 3
 %else
-DECLARE_REG_TMP 10,2
+DECLARE_REG_TMP 10
 %endif
 
-%macro BIWEIGHT_PROLOGUE 1
-    mov t0, %1
+%macro BIWEIGHT_PROLOGUE 0
 .prologue
     PROLOGUE 0,7,8
     movifnidn  r0, r0mp
     movifnidn  r1, r1mp
-    movifnidn t1d, r2m
-    movifnidn r4d, r4m
+    movifnidn r2d, r2m
     movifnidn r5d, r5m
     movifnidn r6d, r6m
+    movifnidn t0d, r7m
 %endmacro
 
 %macro BIWEIGHT_SETUP 1
-    lea        r6, [r6*4+1] ; (offset<<2)+1
-    or         r6, 1
-    shl        r5, 16
-    or         r4, r5
-    movd       m4, r4d      ; weightd | weights
-    movd       m5, r6d      ; (offset+1)|1
-    movd       m6, r3m      ; log2_denom
+    lea        t0, [t0*4+1] ; (offset<<2)+1
+    or         t0, 1
+    shl        r6, 16
+    or         r5, r6
+    movd       m4, r5d      ; weightd | weights
+    movd       m5, t0d      ; (offset+1)|1
+    movd       m6, r4m      ; log2_denom
     pslld      m5, m6       ; (((offset<<2)+1)|1)<<log2_denom
     paddd      m6, [sq_1]
     pshufd     m4, m4, 0
     pshufd     m5, m5, 0
     mova       m3, [pw_pixel_max]
+    movifnidn r3d, r3m
 %ifnidn %1, sse4
     pxor       m7, m7
 %endif
@@ -243,23 +217,19 @@ DECLARE_REG_TMP 10,2
 %endmacro
 
 %macro BIWEIGHT_FUNC_DBL 1
-cglobal h264_biweight_16x16_10_%1
-    BIWEIGHT_PROLOGUE 16
+cglobal h264_biweight_16_10_%1
+    BIWEIGHT_PROLOGUE
     BIWEIGHT_SETUP %1
 .nextrow
     BIWEIGHT  %1,  0
     mova [r0   ], m0
     BIWEIGHT  %1, 16
     mova [r0+16], m0
-    add       r0, t1
-    add       r1, t1
-    dec       t0
+    add       r0, r2
+    add       r1, r2
+    dec       r3d
     jnz .nextrow
     REP_RET
-
-cglobal h264_biweight_16x8_10_%1
-    mov t0, 8
-    jmp mangle(ff_h264_biweight_16x16_10_%1.prologue)
 %endmacro
 
 INIT_XMM
@@ -267,25 +237,17 @@ BIWEIGHT_FUNC_DBL sse2
 BIWEIGHT_FUNC_DBL sse4
 
 %macro BIWEIGHT_FUNC 1
-cglobal h264_biweight_8x16_10_%1
-    BIWEIGHT_PROLOGUE 16
+cglobal h264_biweight_8_10_%1
+    BIWEIGHT_PROLOGUE
     BIWEIGHT_SETUP %1
 .nextrow
     BIWEIGHT %1, 0
     mova   [r0], m0
-    add      r0, t1
-    add      r1, t1
-    dec      t0
+    add      r0, r2
+    add      r1, r2
+    dec      r3d
     jnz .nextrow
     REP_RET
-
-cglobal h264_biweight_8x8_10_%1
-    mov t0, 8
-    jmp mangle(ff_h264_biweight_8x16_10_%1.prologue)
-
-cglobal h264_biweight_8x4_10_%1
-    mov t0, 4
-    jmp mangle(ff_h264_biweight_8x16_10_%1.prologue)
 %endmacro
 
 INIT_XMM
@@ -293,27 +255,20 @@ BIWEIGHT_FUNC sse2
 BIWEIGHT_FUNC sse4
 
 %macro BIWEIGHT_FUNC_HALF 1
-cglobal h264_biweight_4x8_10_%1
-    BIWEIGHT_PROLOGUE 4
+cglobal h264_biweight_4_10_%1
+    BIWEIGHT_PROLOGUE
     BIWEIGHT_SETUP %1
-    lea        r4, [t1*2]
+    sar        r3d, 1
+    lea        r4, [r2*2]
 .nextrow
-    BIWEIGHT    %1, 0, t1
+    BIWEIGHT    %1, 0, r2
     movh   [r0   ], m0
-    movhps [r0+t1], m0
+    movhps [r0+r2], m0
     add         r0, r4
     add         r1, r4
-    dec         t0
+    dec         r3d
     jnz .nextrow
     REP_RET
-
-cglobal h264_biweight_4x4_10_%1
-    mov t0, 2
-    jmp mangle(ff_h264_biweight_4x8_10_%1.prologue)
-
-cglobal h264_biweight_4x2_10_%1
-    mov t0, 1
-    jmp mangle(ff_h264_biweight_4x8_10_%1.prologue)
 %endmacro
 
 INIT_XMM
diff --git a/libavcodec/x86/h264dsp_mmx.c b/libavcodec/x86/h264dsp_mmx.c
index 71beb262c9..b337462aec 100644
--- a/libavcodec/x86/h264dsp_mmx.c
+++ b/libavcodec/x86/h264dsp_mmx.c
@@ -298,63 +298,53 @@ LF_IFUNC(v,  luma_intra,      10, mmxext)
 /***********************************/
 /* weighted prediction */
 
-#define H264_WEIGHT(W, H, OPT) \
-void ff_h264_weight_ ## W ## x ## H ## _ ## OPT(uint8_t *dst, \
-    int stride, int log2_denom, int weight, int offset);
+#define H264_WEIGHT(W, OPT) \
+void ff_h264_weight_ ## W ## _ ## OPT(uint8_t *dst, \
+    int stride, int height, int log2_denom, int weight, int offset);
 
-#define H264_BIWEIGHT(W, H, OPT) \
-void ff_h264_biweight_ ## W ## x ## H ## _ ## OPT(uint8_t *dst, \
-    uint8_t *src, int stride, int log2_denom, int weightd, \
+#define H264_BIWEIGHT(W, OPT) \
+void ff_h264_biweight_ ## W ## _ ## OPT(uint8_t *dst, \
+    uint8_t *src, int stride, int height, int log2_denom, int weightd, \
     int weights, int offset);
 
-#define H264_BIWEIGHT_MMX(W,H) \
-H264_WEIGHT  (W, H, mmx2) \
-H264_BIWEIGHT(W, H, mmx2)
-
-#define H264_BIWEIGHT_MMX_SSE(W,H) \
-H264_BIWEIGHT_MMX(W, H) \
-H264_WEIGHT      (W, H, sse2) \
-H264_BIWEIGHT    (W, H, sse2) \
-H264_BIWEIGHT    (W, H, ssse3)
-
-H264_BIWEIGHT_MMX_SSE(16, 16)
-H264_BIWEIGHT_MMX_SSE(16,  8)
-H264_BIWEIGHT_MMX_SSE( 8, 16)
-H264_BIWEIGHT_MMX_SSE( 8,  8)
-H264_BIWEIGHT_MMX_SSE( 8,  4)
-H264_BIWEIGHT_MMX    ( 4,  8)
-H264_BIWEIGHT_MMX    ( 4,  4)
-H264_BIWEIGHT_MMX    ( 4,  2)
-
-#define H264_WEIGHT_10(W, H, DEPTH, OPT) \
-void ff_h264_weight_ ## W ## x ## H ## _ ## DEPTH ## _ ## OPT(uint8_t *dst, \
-    int stride, int log2_denom, int weight, int offset);
-
-#define H264_BIWEIGHT_10(W, H, DEPTH, OPT) \
-void ff_h264_biweight_ ## W ## x ## H ## _ ## DEPTH ## _ ## OPT \
-    (uint8_t *dst, uint8_t *src, int stride, int log2_denom, \
+#define H264_BIWEIGHT_MMX(W) \
+H264_WEIGHT  (W, mmx2) \
+H264_BIWEIGHT(W, mmx2)
+
+#define H264_BIWEIGHT_MMX_SSE(W) \
+H264_BIWEIGHT_MMX(W) \
+H264_WEIGHT      (W, sse2) \
+H264_BIWEIGHT    (W, sse2) \
+H264_BIWEIGHT    (W, ssse3)
+
+H264_BIWEIGHT_MMX_SSE(16)
+H264_BIWEIGHT_MMX_SSE( 8)
+H264_BIWEIGHT_MMX    ( 4)
+
+#define H264_WEIGHT_10(W, DEPTH, OPT) \
+void ff_h264_weight_ ## W ## _ ## DEPTH ## _ ## OPT(uint8_t *dst, \
+    int stride, int height, int log2_denom, int weight, int offset);
+
+#define H264_BIWEIGHT_10(W, DEPTH, OPT) \
+void ff_h264_biweight_ ## W ## _ ## DEPTH ## _ ## OPT \
+    (uint8_t *dst, uint8_t *src, int stride, int height, int log2_denom, \
      int weightd, int weights, int offset);
 
-#define H264_BIWEIGHT_10_SSE(W, H, DEPTH) \
-H264_WEIGHT_10  (W, H, DEPTH, sse2) \
-H264_WEIGHT_10  (W, H, DEPTH, sse4) \
-H264_BIWEIGHT_10(W, H, DEPTH, sse2) \
-H264_BIWEIGHT_10(W, H, DEPTH, sse4)
-
-H264_BIWEIGHT_10_SSE(16, 16, 10)
-H264_BIWEIGHT_10_SSE(16,  8, 10)
-H264_BIWEIGHT_10_SSE( 8, 16, 10)
-H264_BIWEIGHT_10_SSE( 8,  8, 10)
-H264_BIWEIGHT_10_SSE( 8,  4, 10)
-H264_BIWEIGHT_10_SSE( 4,  8, 10)
-H264_BIWEIGHT_10_SSE( 4,  4, 10)
-H264_BIWEIGHT_10_SSE( 4,  2, 10)
+#define H264_BIWEIGHT_10_SSE(W, DEPTH) \
+H264_WEIGHT_10  (W, DEPTH, sse2) \
+H264_WEIGHT_10  (W, DEPTH, sse4) \
+H264_BIWEIGHT_10(W, DEPTH, sse2) \
+H264_BIWEIGHT_10(W, DEPTH, sse4)
+
+H264_BIWEIGHT_10_SSE(16, 10)
+H264_BIWEIGHT_10_SSE( 8, 10)
+H264_BIWEIGHT_10_SSE( 4, 10)
 
 void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth, const int chroma_format_idc)
 {
     int mm_flags = av_get_cpu_flags();
 
-    if (mm_flags & AV_CPU_FLAG_MMX2) {
+    if (chroma_format_idc == 1 && mm_flags & AV_CPU_FLAG_MMX2) {
         c->h264_loop_filter_strength= h264_loop_filter_strength_mmx2;
     }
 
@@ -394,23 +384,13 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth, const int chrom
             c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_mmxext;
             c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_8_mmxext;
 #endif
-            c->weight_h264_pixels_tab[0]= ff_h264_weight_16x16_mmx2;
-            c->weight_h264_pixels_tab[1]= ff_h264_weight_16x8_mmx2;
-            c->weight_h264_pixels_tab[2]= ff_h264_weight_8x16_mmx2;
-            c->weight_h264_pixels_tab[3]= ff_h264_weight_8x8_mmx2;
-            c->weight_h264_pixels_tab[4]= ff_h264_weight_8x4_mmx2;
-            c->weight_h264_pixels_tab[5]= ff_h264_weight_4x8_mmx2;
-            c->weight_h264_pixels_tab[6]= ff_h264_weight_4x4_mmx2;
-            c->weight_h264_pixels_tab[7]= ff_h264_weight_4x2_mmx2;
-
-            c->biweight_h264_pixels_tab[0]= ff_h264_biweight_16x16_mmx2;
-            c->biweight_h264_pixels_tab[1]= ff_h264_biweight_16x8_mmx2;
-            c->biweight_h264_pixels_tab[2]= ff_h264_biweight_8x16_mmx2;
-            c->biweight_h264_pixels_tab[3]= ff_h264_biweight_8x8_mmx2;
-            c->biweight_h264_pixels_tab[4]= ff_h264_biweight_8x4_mmx2;
-            c->biweight_h264_pixels_tab[5]= ff_h264_biweight_4x8_mmx2;
-            c->biweight_h264_pixels_tab[6]= ff_h264_biweight_4x4_mmx2;
-            c->biweight_h264_pixels_tab[7]= ff_h264_biweight_4x2_mmx2;
+            c->weight_h264_pixels_tab[0]= ff_h264_weight_16_mmx2;
+            c->weight_h264_pixels_tab[1]= ff_h264_weight_8_mmx2;
+            c->weight_h264_pixels_tab[2]= ff_h264_weight_4_mmx2;
+
+            c->biweight_h264_pixels_tab[0]= ff_h264_biweight_16_mmx2;
+            c->biweight_h264_pixels_tab[1]= ff_h264_biweight_8_mmx2;
+            c->biweight_h264_pixels_tab[2]= ff_h264_biweight_4_mmx2;
 
             if (mm_flags&AV_CPU_FLAG_SSE2) {
                 c->h264_idct8_add           = ff_h264_idct8_add_8_sse2;
@@ -422,17 +402,11 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth, const int chrom
                 c->h264_idct_add16intra     = ff_h264_idct_add16intra_8_sse2;
                 c->h264_luma_dc_dequant_idct= ff_h264_luma_dc_dequant_idct_sse2;
 
-                c->weight_h264_pixels_tab[0]= ff_h264_weight_16x16_sse2;
-                c->weight_h264_pixels_tab[1]= ff_h264_weight_16x8_sse2;
-                c->weight_h264_pixels_tab[2]= ff_h264_weight_8x16_sse2;
-                c->weight_h264_pixels_tab[3]= ff_h264_weight_8x8_sse2;
-                c->weight_h264_pixels_tab[4]= ff_h264_weight_8x4_sse2;
+                c->weight_h264_pixels_tab[0]= ff_h264_weight_16_sse2;
+                c->weight_h264_pixels_tab[1]= ff_h264_weight_8_sse2;
 
-                c->biweight_h264_pixels_tab[0]= ff_h264_biweight_16x16_sse2;
-                c->biweight_h264_pixels_tab[1]= ff_h264_biweight_16x8_sse2;
-                c->biweight_h264_pixels_tab[2]= ff_h264_biweight_8x16_sse2;
-                c->biweight_h264_pixels_tab[3]= ff_h264_biweight_8x8_sse2;
-                c->biweight_h264_pixels_tab[4]= ff_h264_biweight_8x4_sse2;
+                c->biweight_h264_pixels_tab[0]= ff_h264_biweight_16_sse2;
+                c->biweight_h264_pixels_tab[1]= ff_h264_biweight_8_sse2;
 
 #if HAVE_ALIGNED_STACK
                 c->h264_v_loop_filter_luma = ff_deblock_v_luma_8_sse2;
@@ -442,11 +416,8 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth, const int chrom
 #endif
             }
             if (mm_flags&AV_CPU_FLAG_SSSE3) {
-                c->biweight_h264_pixels_tab[0]= ff_h264_biweight_16x16_ssse3;
-                c->biweight_h264_pixels_tab[1]= ff_h264_biweight_16x8_ssse3;
-                c->biweight_h264_pixels_tab[2]= ff_h264_biweight_8x16_ssse3;
-                c->biweight_h264_pixels_tab[3]= ff_h264_biweight_8x8_ssse3;
-                c->biweight_h264_pixels_tab[4]= ff_h264_biweight_8x4_ssse3;
+                c->biweight_h264_pixels_tab[0]= ff_h264_biweight_16_ssse3;
+                c->biweight_h264_pixels_tab[1]= ff_h264_biweight_8_ssse3;
             }
             if (HAVE_AVX && mm_flags&AV_CPU_FLAG_AVX) {
 #if HAVE_ALIGNED_STACK
@@ -485,23 +456,13 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth, const int chrom
                 c->h264_idct8_add4     = ff_h264_idct8_add4_10_sse2;
 #endif
 
-                c->weight_h264_pixels_tab[0] = ff_h264_weight_16x16_10_sse2;
-                c->weight_h264_pixels_tab[1] = ff_h264_weight_16x8_10_sse2;
-                c->weight_h264_pixels_tab[2] = ff_h264_weight_8x16_10_sse2;
-                c->weight_h264_pixels_tab[3] = ff_h264_weight_8x8_10_sse2;
-                c->weight_h264_pixels_tab[4] = ff_h264_weight_8x4_10_sse2;
-                c->weight_h264_pixels_tab[5] = ff_h264_weight_4x8_10_sse2;
-                c->weight_h264_pixels_tab[6] = ff_h264_weight_4x4_10_sse2;
-                c->weight_h264_pixels_tab[7] = ff_h264_weight_4x2_10_sse2;
-
-                c->biweight_h264_pixels_tab[0] = ff_h264_biweight_16x16_10_sse2;
-                c->biweight_h264_pixels_tab[1] = ff_h264_biweight_16x8_10_sse2;
-                c->biweight_h264_pixels_tab[2] = ff_h264_biweight_8x16_10_sse2;
-                c->biweight_h264_pixels_tab[3] = ff_h264_biweight_8x8_10_sse2;
-                c->biweight_h264_pixels_tab[4] = ff_h264_biweight_8x4_10_sse2;
-                c->biweight_h264_pixels_tab[5] = ff_h264_biweight_4x8_10_sse2;
-                c->biweight_h264_pixels_tab[6] = ff_h264_biweight_4x4_10_sse2;
-                c->biweight_h264_pixels_tab[7] = ff_h264_biweight_4x2_10_sse2;
+                c->weight_h264_pixels_tab[0] = ff_h264_weight_16_10_sse2;
+                c->weight_h264_pixels_tab[1] = ff_h264_weight_8_10_sse2;
+                c->weight_h264_pixels_tab[2] = ff_h264_weight_4_10_sse2;
+
+                c->biweight_h264_pixels_tab[0] = ff_h264_biweight_16_10_sse2;
+                c->biweight_h264_pixels_tab[1] = ff_h264_biweight_8_10_sse2;
+                c->biweight_h264_pixels_tab[2] = ff_h264_biweight_4_10_sse2;
 
                 c->h264_v_loop_filter_chroma= ff_deblock_v_chroma_10_sse2;
                 c->h264_v_loop_filter_chroma_intra= ff_deblock_v_chroma_intra_10_sse2;
@@ -513,23 +474,13 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth, const int chrom
 #endif
             }
             if (mm_flags&AV_CPU_FLAG_SSE4) {
-                c->weight_h264_pixels_tab[0] = ff_h264_weight_16x16_10_sse4;
-                c->weight_h264_pixels_tab[1] = ff_h264_weight_16x8_10_sse4;
-                c->weight_h264_pixels_tab[2] = ff_h264_weight_8x16_10_sse4;
-                c->weight_h264_pixels_tab[3] = ff_h264_weight_8x8_10_sse4;
-                c->weight_h264_pixels_tab[4] = ff_h264_weight_8x4_10_sse4;
-                c->weight_h264_pixels_tab[5] = ff_h264_weight_4x8_10_sse4;
-                c->weight_h264_pixels_tab[6] = ff_h264_weight_4x4_10_sse4;
-                c->weight_h264_pixels_tab[7] = ff_h264_weight_4x2_10_sse4;
-
-                c->biweight_h264_pixels_tab[0] = ff_h264_biweight_16x16_10_sse4;
-                c->biweight_h264_pixels_tab[1] = ff_h264_biweight_16x8_10_sse4;
-                c->biweight_h264_pixels_tab[2] = ff_h264_biweight_8x16_10_sse4;
-                c->biweight_h264_pixels_tab[3] = ff_h264_biweight_8x8_10_sse4;
-                c->biweight_h264_pixels_tab[4] = ff_h264_biweight_8x4_10_sse4;
-                c->biweight_h264_pixels_tab[5] = ff_h264_biweight_4x8_10_sse4;
-                c->biweight_h264_pixels_tab[6] = ff_h264_biweight_4x4_10_sse4;
-                c->biweight_h264_pixels_tab[7] = ff_h264_biweight_4x2_10_sse4;
+                c->weight_h264_pixels_tab[0] = ff_h264_weight_16_10_sse4;
+                c->weight_h264_pixels_tab[1] = ff_h264_weight_8_10_sse4;
+                c->weight_h264_pixels_tab[2] = ff_h264_weight_4_10_sse4;
+
+                c->biweight_h264_pixels_tab[0] = ff_h264_biweight_16_10_sse4;
+                c->biweight_h264_pixels_tab[1] = ff_h264_biweight_8_10_sse4;
+                c->biweight_h264_pixels_tab[2] = ff_h264_biweight_4_10_sse4;
             }
 #if HAVE_AVX
             if (mm_flags&AV_CPU_FLAG_AVX) {