sync to FFmpeg 51.11.0

CVS patchset: 8146
CVS date: 2006/08/02 07:02:37

10 files changed, 844 insertions, 289 deletions

diff --git a/src/libffmpeg/libavcodec/i386/cputest.c b/src/libffmpeg/libavcodec/i386/cputest.c
index 64656c65a..a66bdbe98 100644
--- a/src/libffmpeg/libavcodec/i386/cputest.c
+++ b/src/libffmpeg/libavcodec/i386/cputest.c
@@ -64,6 +64,8 @@ int mm_support(void)
             rval |= MM_MMXEXT | MM_SSE;
         if (std_caps & (1<<26))
             rval |= MM_SSE2;
+        if (ecx & 1)
+            rval |= MM_SSE3;
     }
 
     cpuid(0x80000000, max_ext_level, ebx, ecx, edx);
@@ -76,38 +78,10 @@ int mm_support(void)
             rval |= MM_3DNOWEXT;
         if (ext_caps & (1<<23))
             rval |= MM_MMX;
-    }
-
-    cpuid(0, eax, ebx, ecx, edx);
-    if (       ebx == 0x68747541 &&
-               edx == 0x69746e65 &&
-               ecx == 0x444d4163) {
-        /* AMD */
-        if(ext_caps & (1<<22))
-            rval |= MM_MMXEXT;
-    } else if (ebx == 0x746e6543 &&
-               edx == 0x48727561 &&
-               ecx == 0x736c7561) {  /*  "CentaurHauls" */
-        /* VIA C3 */
-        if(ext_caps & (1<<24))
-          rval |= MM_MMXEXT;
-    } else if (ebx == 0x69727943 &&
-               edx == 0x736e4978 &&
-               ecx == 0x64616574) {
-        /* Cyrix Section */
-        /* See if extended CPUID level 80000001 is supported */
-        /* The value of CPUID/80000001 for the 6x86MX is undefined
-           according to the Cyrix CPU Detection Guide (Preliminary
-           Rev. 1.01 table 1), so we'll check the value of eax for
-           CPUID/0 to see if standard CPUID level 2 is supported.
-           According to the table, the only CPU which supports level
-           2 is also the only one which supports extended CPUID levels.
-        */
-        if (eax < 2)
-            return rval;
-        if (ext_caps & (1<<24))
+        if (ext_caps & (1<<22))
             rval |= MM_MMXEXT;
     }
+
 #if 0
     av_log(NULL, AV_LOG_DEBUG, "%s%s%s%s%s%s\n",
         (rval&MM_MMX) ? "MMX ":"",
diff --git a/src/libffmpeg/libavcodec/i386/dsputil_h264_template_mmx.c b/src/libffmpeg/libavcodec/i386/dsputil_h264_template_mmx.c
index d52938ccf..b49c880a7 100644
--- a/src/libffmpeg/libavcodec/i386/dsputil_h264_template_mmx.c
+++ b/src/libffmpeg/libavcodec/i386/dsputil_h264_template_mmx.c
@@ -37,112 +37,56 @@ static void H264_CHROMA_MC8_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*
 
     assert(x<8 && y<8 && x>=0 && y>=0);
 
-    if(y==0)
+    if(y==0 || x==0)
     {
-        /* horizontal filter only */
-        asm volatile("movd %0, %%mm5\n\t"
-                     "punpcklwd %%mm5, %%mm5\n\t"
-                     "punpckldq %%mm5, %%mm5\n\t" /* mm5 = B = x */
-                     "movq %1, %%mm4\n\t"
-                     "pxor %%mm7, %%mm7\n\t"
-                     "psubw %%mm5, %%mm4\n\t"     /* mm4 = A = 8-x */
-                     : : "rm" (x), "m" (ff_pw_8));
+        /* 1 dimensional filter only */
+        const int dxy = x ? 1 : stride;
+
+        asm volatile(
+            "movd %0, %%mm5\n\t"
+            "movq %1, %%mm4\n\t"
+            "punpcklwd %%mm5, %%mm5\n\t"
+            "punpckldq %%mm5, %%mm5\n\t" /* mm5 = B = x */
+            "movq %%mm4, %%mm6\n\t"
+            "pxor %%mm7, %%mm7\n\t"
+            "psubw %%mm5, %%mm4\n\t"     /* mm4 = A = 8-x */
+            "psrlw $1, %%mm6\n\t"        /* mm6 = 4 */
+            :: "rm"(x+y), "m"(ff_pw_8));
 
         for(i=0; i<h; i++) {
             asm volatile(
                 /* mm0 = src[0..7], mm1 = src[1..8] */
                 "movq %0, %%mm0\n\t"
-                "movq %1, %%mm1\n\t"
-                : : "m" (src[0]), "m" (src[1]));
+                "movq %1, %%mm2\n\t"
+                :: "m"(src[0]), "m"(src[dxy]));
 
             asm volatile(
-                /* [mm2,mm3] = A * src[0..7] */
-                "movq %%mm0, %%mm2\n\t"
-                "punpcklbw %%mm7, %%mm2\n\t"
-                "pmullw %%mm4, %%mm2\n\t"
-                "movq %%mm0, %%mm3\n\t"
-                "punpckhbw %%mm7, %%mm3\n\t"
-                "pmullw %%mm4, %%mm3\n\t"
-
-                /* [mm2,mm3] += B * src[1..8] */
-                "movq %%mm1, %%mm0\n\t"
+                /* [mm0,mm1] = A * src[0..7] */
+                /* [mm2,mm3] = B * src[1..8] */
+                "movq %%mm0, %%mm1\n\t"
+                "movq %%mm2, %%mm3\n\t"
                 "punpcklbw %%mm7, %%mm0\n\t"
-                "pmullw %%mm5, %%mm0\n\t"
                 "punpckhbw %%mm7, %%mm1\n\t"
-                "pmullw %%mm5, %%mm1\n\t"
-                "paddw %%mm0, %%mm2\n\t"
-                "paddw %%mm1, %%mm3\n\t"
-
-                /* dst[0..7] = pack(([mm2,mm3] + 32) >> 6) */
-                "paddw %1, %%mm2\n\t"
-                "paddw %1, %%mm3\n\t"
-                "psrlw $3, %%mm2\n\t"
-                "psrlw $3, %%mm3\n\t"
-                "packuswb %%mm3, %%mm2\n\t"
-                H264_CHROMA_OP(%0, %%mm2)
-                "movq %%mm2, %0\n\t"
-                : "=m" (dst[0]) : "m" (ff_pw_4));
-
-            src += stride;
-            dst += stride;
-        }
-        return;
-    }
-
-    if(x==0)
-    {
-        /* vertical filter only */
-        asm volatile("movd %0, %%mm6\n\t"
-                     "punpcklwd %%mm6, %%mm6\n\t"
-                     "punpckldq %%mm6, %%mm6\n\t" /* mm6 = C = y */
-                     "movq %1, %%mm4\n\t"
-                     "pxor %%mm7, %%mm7\n\t"
-                     "psubw %%mm6, %%mm4\n\t"     /* mm4 = A = 8-y */
-                     : : "rm" (y), "m" (ff_pw_8));
-
-        asm volatile(
-            /* mm0 = src[0..7] */
-            "movq %0, %%mm0\n\t"
-            : : "m" (src[0]));
-
-        for(i=0; i<h; i++) {
-            asm volatile(
-                /* [mm2,mm3] = A * src[0..7] */
-                "movq %mm0, %mm2\n\t"
-                "punpcklbw %mm7, %mm2\n\t"
-                "pmullw %mm4, %mm2\n\t"
-                "movq %mm0, %mm3\n\t"
-                "punpckhbw %mm7, %mm3\n\t"
-                "pmullw %mm4, %mm3\n\t");
+                "punpcklbw %%mm7, %%mm2\n\t"
+                "punpckhbw %%mm7, %%mm3\n\t"
+                "pmullw %%mm4, %%mm0\n\t"
+                "pmullw %%mm4, %%mm1\n\t"
+                "pmullw %%mm5, %%mm2\n\t"
+                "pmullw %%mm5, %%mm3\n\t"
+
+                /* dst[0..7] = (A * src[0..7] + B * src[1..8] + 4) >> 3 */
+                "paddw %%mm6, %%mm0\n\t"
+                "paddw %%mm6, %%mm1\n\t"
+                "paddw %%mm2, %%mm0\n\t"
+                "paddw %%mm3, %%mm1\n\t"
+                "psrlw $3, %%mm0\n\t"
+                "psrlw $3, %%mm1\n\t"
+                "packuswb %%mm1, %%mm0\n\t"
+                H264_CHROMA_OP(%0, %%mm0)
+                "movq %%mm0, %0\n\t"
+                : "=m" (dst[0]));
 
             src += stride;
-            asm volatile(
-                /* mm0 = src[0..7] */
-                "movq %0, %%mm0\n\t"
-                : : "m" (src[0]));
-
-            asm volatile(
-                /* [mm2,mm3] += C * src[0..7] */
-                "movq %mm0, %mm1\n\t"
-                "punpcklbw %mm7, %mm1\n\t"
-                "pmullw %mm6, %mm1\n\t"
-                "paddw %mm1, %mm2\n\t"
-                "movq %mm0, %mm5\n\t"
-                "punpckhbw %mm7, %mm5\n\t"
-                "pmullw %mm6, %mm5\n\t"
-                "paddw %mm5, %mm3\n\t");
-
-            asm volatile(
-                /* dst[0..7] = pack(([mm2,mm3] + 32) >> 6) */
-                "paddw %1, %%mm2\n\t"
-                "paddw %1, %%mm3\n\t"
-                "psrlw $3, %%mm2\n\t"
-                "psrlw $3, %%mm3\n\t"
-                "packuswb %%mm3, %%mm2\n\t"
-                H264_CHROMA_OP(%0, %%mm2)
-                "movq %%mm2, %0\n\t"
-                : "=m" (dst[0]) : "m" (ff_pw_4));
-
             dst += stride;
         }
         return;
@@ -177,57 +121,53 @@ static void H264_CHROMA_MC8_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*
         : : "m" (src[0]), "m" (src[1]));
 
     for(i=0; i<h; i++) {
+        src += stride;
+
         asm volatile(
-            /* [mm2,mm3] = A * src[0..7] */
+            /* mm2 = A * src[0..3] + B * src[1..4] */
+            /* mm3 = A * src[4..7] + B * src[5..8] */
             "movq %%mm0, %%mm2\n\t"
+            "movq %%mm1, %%mm3\n\t"
+            "punpckhbw %%mm7, %%mm0\n\t"
+            "punpcklbw %%mm7, %%mm1\n\t"
             "punpcklbw %%mm7, %%mm2\n\t"
-            "pmullw %0, %%mm2\n\t"
-            "movq %%mm0, %%mm3\n\t"
             "punpckhbw %%mm7, %%mm3\n\t"
-            "pmullw %0, %%mm3\n\t"
+            "pmullw %0, %%mm0\n\t"
+            "pmullw %0, %%mm2\n\t"
+            "pmullw %%mm5, %%mm1\n\t"
+            "pmullw %%mm5, %%mm3\n\t"
+            "paddw %%mm1, %%mm2\n\t"
+            "paddw %%mm0, %%mm3\n\t"
+            : : "m" (AA));
 
-            /* [mm2,mm3] += B * src[1..8] */
-            "movq %%mm1, %%mm0\n\t"
+        asm volatile(
+            /* [mm2,mm3] += C * src[0..7] */
+            "movq %0, %%mm0\n\t"
+            "movq %%mm0, %%mm1\n\t"
             "punpcklbw %%mm7, %%mm0\n\t"
-            "pmullw %%mm5, %%mm0\n\t"
             "punpckhbw %%mm7, %%mm1\n\t"
-            "pmullw %%mm5, %%mm1\n\t"
+            "pmullw %%mm6, %%mm0\n\t"
+            "pmullw %%mm6, %%mm1\n\t"
             "paddw %%mm0, %%mm2\n\t"
             "paddw %%mm1, %%mm3\n\t"
-            : : "m" (AA));
+            : : "m" (src[0]));
 
-        src += stride;
         asm volatile(
-            /* mm0 = src[0..7], mm1 = src[1..8] */
-            "movq %0, %%mm0\n\t"
+            /* [mm2,mm3] += D * src[1..8] */
             "movq %1, %%mm1\n\t"
-            : : "m" (src[0]), "m" (src[1]));
-
-        asm volatile(
-            /* [mm2,mm3] += C *  src[0..7] */
-            "movq %mm0, %mm4\n\t"
-            "punpcklbw %mm7, %mm4\n\t"
-            "pmullw %mm6, %mm4\n\t"
-            "paddw %mm4, %mm2\n\t"
-            "movq %mm0, %mm4\n\t"
-            "punpckhbw %mm7, %mm4\n\t"
-            "pmullw %mm6, %mm4\n\t"
-            "paddw %mm4, %mm3\n\t");
-
-        asm volatile(
-            /* [mm2,mm3] += D *  src[1..8] */
-            "movq %%mm1, %%mm4\n\t"
-            "punpcklbw %%mm7, %%mm4\n\t"
-            "pmullw %0, %%mm4\n\t"
-            "paddw %%mm4, %%mm2\n\t"
+            "movq %%mm1, %%mm0\n\t"
             "movq %%mm1, %%mm4\n\t"
+            "punpcklbw %%mm7, %%mm0\n\t"
             "punpckhbw %%mm7, %%mm4\n\t"
-            "pmullw %0, %%mm4\n\t"
+            "pmullw %2, %%mm0\n\t"
+            "pmullw %2, %%mm4\n\t"
+            "paddw %%mm0, %%mm2\n\t"
             "paddw %%mm4, %%mm3\n\t"
-            : : "m" (DD));
+            "movq %0, %%mm0\n\t"
+            : : "m" (src[0]), "m" (src[1]), "m" (DD));
 
         asm volatile(
-            /* dst[0..7] = pack(([mm2,mm3] + 32) >> 6) */
+            /* dst[0..7] = ([mm2,mm3] + 32) >> 6 */
             "paddw %1, %%mm2\n\t"
             "paddw %1, %%mm3\n\t"
             "psrlw $6, %%mm2\n\t"
@@ -240,7 +180,7 @@ static void H264_CHROMA_MC8_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*
     }
 }
 
-static void H264_CHROMA_MC4_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
+static void H264_CHROMA_MC4_TMPL(uint8_t *dst/*align 4*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
 {
     DECLARE_ALIGNED_8(uint64_t, AA);
     DECLARE_ALIGNED_8(uint64_t, DD);
@@ -319,3 +259,66 @@ static void H264_CHROMA_MC4_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*
         dst += stride;
     }
 }
+
+#ifdef H264_CHROMA_MC2_TMPL
+static void H264_CHROMA_MC2_TMPL(uint8_t *dst/*align 2*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
+{
+    int CD=((1<<16)-1)*x*y + 8*y;
+    int AB=((8<<16)-8)*x + 64 - CD;
+    int i;
+
+    asm volatile(
+        /* mm5 = {A,B,A,B} */
+        /* mm6 = {C,D,C,D} */
+        "movd %0, %%mm5\n\t"
+        "movd %1, %%mm6\n\t"
+        "punpckldq %%mm5, %%mm5\n\t"
+        "punpckldq %%mm6, %%mm6\n\t"
+        "pxor %%mm7, %%mm7\n\t"
+        :: "r"(AB), "r"(CD));
+
+    asm volatile(
+        /* mm0 = src[0,1,1,2] */
+        "movd %0, %%mm0\n\t"
+        "punpcklbw %%mm7, %%mm0\n\t"
+        "pshufw $0x94, %%mm0, %%mm0\n\t"
+        :: "m"(src[0]));
+
+    for(i=0; i<h; i++) {
+        asm volatile(
+            /* mm1 = A * src[0,1] + B * src[1,2] */
+            "movq    %%mm0, %%mm1\n\t"
+            "pmaddwd %%mm5, %%mm1\n\t"
+            ::);
+
+        src += stride;
+        asm volatile(
+            /* mm0 = src[0,1,1,2] */
+            "movd %0, %%mm0\n\t"
+            "punpcklbw %%mm7, %%mm0\n\t"
+            "pshufw $0x94, %%mm0, %%mm0\n\t"
+            :: "m"(src[0]));
+
+        asm volatile(
+            /* mm1 += C * src[0,1] + D * src[1,2] */
+            "movq    %%mm0, %%mm2\n\t"
+            "pmaddwd %%mm6, %%mm2\n\t"
+            "paddw   %%mm2, %%mm1\n\t"
+            ::);
+
+        asm volatile(
+            /* dst[0,1] = pack((mm1 + 32) >> 6) */
+            "paddw %1, %%mm1\n\t"
+            "psrlw $6, %%mm1\n\t"
+            "packssdw %%mm7, %%mm1\n\t"
+            "packuswb %%mm7, %%mm1\n\t"
+            /* writes garbage to the right of dst.
+             * ok because partitions are processed from left to right. */
+            H264_CHROMA_OP4(%0, %%mm1, %%mm3)
+            "movd %%mm1, %0\n\t"
+            : "=m" (dst[0]) : "m" (ff_pw_32));
+        dst += stride;
+    }
+}
+#endif
+
diff --git a/src/libffmpeg/libavcodec/i386/dsputil_mmx.c b/src/libffmpeg/libavcodec/i386/dsputil_mmx.c
index 7d69859a6..ec6b2ad1a 100644
--- a/src/libffmpeg/libavcodec/i386/dsputil_mmx.c
+++ b/src/libffmpeg/libavcodec/i386/dsputil_mmx.c
@@ -23,6 +23,7 @@
 #include "../dsputil.h"
 #include "../simple_idct.h"
 #include "../mpegvideo.h"
+#include "x86_cpu.h"
 #include "mmx.h"
 
 //#undef NDEBUG
@@ -186,6 +187,11 @@ static const uint64_t ff_pb_FC attribute_used __attribute__ ((aligned(8))) = 0xF
 #undef DEF
 #undef PAVGB
 
+#define SBUTTERFLY(a,b,t,n)\
+    "movq " #a ", " #t "              \n\t" /* abcd */\
+    "punpckl" #n " " #b ", " #a "     \n\t" /* aebf */\
+    "punpckh" #n " " #b ", " #t "     \n\t" /* cgdh */\
+
 /***********************************/
 /* standard MMX */
 
@@ -1522,11 +1528,6 @@ static void sub_hfyu_median_prediction_mmx2(uint8_t *dst, uint8_t *src1, uint8_t
     "pmaxsw " #z ", " #a "            \n\t"\
     "paddusw " #a ", " #sum "         \n\t"
 
-#define SBUTTERFLY(a,b,t,n)\
-    "movq " #a ", " #t "              \n\t" /* abcd */\
-    "punpckl" #n " " #b ", " #a "     \n\t" /* aebf */\
-    "punpckh" #n " " #b ", " #t "     \n\t" /* cgdh */\
-
 #define TRANSPOSE4(a,b,c,d,t)\
     SBUTTERFLY(a,b,t,wd) /* a=aebf t=cgdh */\
     SBUTTERFLY(c,d,b,wd) /* c=imjn b=kolp */\
@@ -2403,6 +2404,124 @@ static void just_return() { return; }
     c->put_no_rnd_ ## postfix1 = put_no_rnd_ ## postfix2;\
     c->avg_ ## postfix1 = avg_ ## postfix2;
 
+static void gmc_mmx(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
+                    int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height){
+    const int w = 8;
+    const int ix = ox>>(16+shift);
+    const int iy = oy>>(16+shift);
+    const int oxs = ox>>4;
+    const int oys = oy>>4;
+    const int dxxs = dxx>>4;
+    const int dxys = dxy>>4;
+    const int dyxs = dyx>>4;
+    const int dyys = dyy>>4;
+    const uint16_t r4[4] = {r,r,r,r};
+    const uint16_t dxy4[4] = {dxys,dxys,dxys,dxys};
+    const uint16_t dyy4[4] = {dyys,dyys,dyys,dyys};
+    const uint64_t shift2 = 2*shift;
+    uint8_t edge_buf[(h+1)*stride];
+    int x, y;
+
+    const int dxw = (dxx-(1<<(16+shift)))*(w-1);
+    const int dyh = (dyy-(1<<(16+shift)))*(h-1);
+    const int dxh = dxy*(h-1);
+    const int dyw = dyx*(w-1);
+    if( // non-constant fullpel offset (3% of blocks)
+        (ox^(ox+dxw) | ox^(ox+dxh) | ox^(ox+dxw+dxh) |
+         oy^(oy+dyw) | oy^(oy+dyh) | oy^(oy+dyw+dyh)) >> (16+shift)
+        // uses more than 16 bits of subpel mv (only at huge resolution)
+        || (dxx|dxy|dyx|dyy)&15 )
+    {
+        //FIXME could still use mmx for some of the rows
+        ff_gmc_c(dst, src, stride, h, ox, oy, dxx, dxy, dyx, dyy, shift, r, width, height);
+        return;
+    }
+
+    src += ix + iy*stride;
+    if( (unsigned)ix >= width-w ||
+        (unsigned)iy >= height-h )
+    {
+        ff_emulated_edge_mc(edge_buf, src, stride, w+1, h+1, ix, iy, width, height);
+        src = edge_buf;
+    }
+
+    asm volatile(
+        "movd         %0, %%mm6 \n\t"
+        "pxor      %%mm7, %%mm7 \n\t"
+        "punpcklwd %%mm6, %%mm6 \n\t"
+        "punpcklwd %%mm6, %%mm6 \n\t"
+        :: "r"(1<<shift)
+    );
+
+    for(x=0; x<w; x+=4){
+        uint16_t dx4[4] = { oxs - dxys + dxxs*(x+0),
+                            oxs - dxys + dxxs*(x+1),
+                            oxs - dxys + dxxs*(x+2),
+                            oxs - dxys + dxxs*(x+3) };
+        uint16_t dy4[4] = { oys - dyys + dyxs*(x+0),
+                            oys - dyys + dyxs*(x+1),
+                            oys - dyys + dyxs*(x+2),
+                            oys - dyys + dyxs*(x+3) };
+
+        for(y=0; y<h; y++){
+            asm volatile(
+                "movq   %0,  %%mm4 \n\t"
+                "movq   %1,  %%mm5 \n\t"
+                "paddw  %2,  %%mm4 \n\t"
+                "paddw  %3,  %%mm5 \n\t"
+                "movq   %%mm4, %0  \n\t"
+                "movq   %%mm5, %1  \n\t"
+                "psrlw  $12, %%mm4 \n\t"
+                "psrlw  $12, %%mm5 \n\t"
+                : "+m"(*dx4), "+m"(*dy4)
+                : "m"(*dxy4), "m"(*dyy4)
+            );
+
+            asm volatile(
+                "movq   %%mm6, %%mm2 \n\t"
+                "movq   %%mm6, %%mm1 \n\t"
+                "psubw  %%mm4, %%mm2 \n\t"
+                "psubw  %%mm5, %%mm1 \n\t"
+                "movq   %%mm2, %%mm0 \n\t"
+                "movq   %%mm4, %%mm3 \n\t"
+                "pmullw %%mm1, %%mm0 \n\t" // (s-dx)*(s-dy)
+                "pmullw %%mm5, %%mm3 \n\t" // dx*dy
+                "pmullw %%mm5, %%mm2 \n\t" // (s-dx)*dy
+                "pmullw %%mm4, %%mm1 \n\t" // dx*(s-dy)
+
+                "movd   %4,    %%mm5 \n\t"
+                "movd   %3,    %%mm4 \n\t"
+                "punpcklbw %%mm7, %%mm5 \n\t"
+                "punpcklbw %%mm7, %%mm4 \n\t"
+                "pmullw %%mm5, %%mm3 \n\t" // src[1,1] * dx*dy
+                "pmullw %%mm4, %%mm2 \n\t" // src[0,1] * (s-dx)*dy
+
+                "movd   %2,    %%mm5 \n\t"
+                "movd   %1,    %%mm4 \n\t"
+                "punpcklbw %%mm7, %%mm5 \n\t"
+                "punpcklbw %%mm7, %%mm4 \n\t"
+                "pmullw %%mm5, %%mm1 \n\t" // src[1,0] * dx*(s-dy)
+                "pmullw %%mm4, %%mm0 \n\t" // src[0,0] * (s-dx)*(s-dy)
+                "paddw  %5,    %%mm1 \n\t"
+                "paddw  %%mm3, %%mm2 \n\t"
+                "paddw  %%mm1, %%mm0 \n\t"
+                "paddw  %%mm2, %%mm0 \n\t"
+
+                "psrlw    %6,    %%mm0 \n\t"
+                "packuswb %%mm0, %%mm0 \n\t"
+                "movd     %%mm0, %0    \n\t"
+
+                : "=m"(dst[x+y*stride])
+                : "m"(src[0]), "m"(src[1]),
+                  "m"(src[stride]), "m"(src[stride+1]),
+                  "m"(*r4), "m"(shift2)
+            );
+            src += stride;
+        }
+        src += 4-h*stride;
+    }
+}
+
 static int try_8x8basis_mmx(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale){
     long i=0;
 
@@ -2489,8 +2608,36 @@ static void add_8x8basis_mmx(int16_t rem[64], int16_t basis[64], int scale){
     }
 }
 
+#define PREFETCH(name, op) \
+void name(void *mem, int stride, int h){\
+    const uint8_t *p= mem;\
+    do{\
+        asm volatile(#op" %0" :: "m"(*p));\
+        p+= stride;\
+    }while(--h);\
+}
+PREFETCH(prefetch_mmx2,  prefetcht0)
+PREFETCH(prefetch_3dnow, prefetch)
+#undef PREFETCH
+
 #include "h264dsp_mmx.c"
 
+/* AVS specific */
+void ff_cavsdsp_init_mmx2(DSPContext* c, AVCodecContext *avctx);
+
+void ff_put_cavs_qpel8_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride) {
+    put_pixels8_mmx(dst, src, stride, 8);
+}
+void ff_avg_cavs_qpel8_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride) {
+    avg_pixels8_mmx(dst, src, stride, 8);
+}
+void ff_put_cavs_qpel16_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride) {
+    put_pixels16_mmx(dst, src, stride, 16);
+}
+void ff_avg_cavs_qpel16_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride) {
+    avg_pixels16_mmx(dst, src, stride, 16);
+}
+
 /* external functions, from idct_mmx.c */
 void ff_mmx_idct(DCTELEM *block);
 void ff_mmxext_idct(DCTELEM *block);
@@ -2564,6 +2711,17 @@ static void ff_idct_xvid_mmx2_add(uint8_t *dest, int line_size, DCTELEM *block)
 }
 #endif
 
+#ifdef CONFIG_SNOW_ENCODER
+extern void ff_snow_horizontal_compose97i_sse2(DWTELEM *b, int width);
+extern void ff_snow_horizontal_compose97i_mmx(DWTELEM *b, int width);
+extern void ff_snow_vertical_compose97i_sse2(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, DWTELEM *b3, DWTELEM *b4, DWTELEM *b5, int width);
+extern void ff_snow_vertical_compose97i_mmx(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, DWTELEM *b3, DWTELEM *b4, DWTELEM *b5, int width);
+extern void ff_snow_inner_add_yblock_sse2(uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h,
+                           int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8);
+extern void ff_snow_inner_add_yblock_mmx(uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h,
+                          int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8);
+#endif
+
 void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
 {
     mm_flags = mm_support();
@@ -2622,6 +2780,7 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
                     c->idct    = ff_mmx_idct;
                 }
                 c->idct_permutation_type= FF_LIBMPEG2_IDCT_PERM;
+#if 0
             }else if(idct_algo==FF_IDCT_VP3){
                 if(mm_flags & MM_SSE2){
                     c->idct_put= ff_vp3_idct_put_sse2;
@@ -2635,6 +2794,9 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
                     c->idct    = ff_vp3_idct_mmx;
                     c->idct_permutation_type= FF_PARTTRANS_IDCT_PERM;
                 }
+#endif
+            }else if(idct_algo==FF_IDCT_CAVS){
+                    c->idct_permutation_type= FF_TRANSPOSE_IDCT_PERM;
 #ifdef CONFIG_GPL
             }else if(idct_algo==FF_IDCT_XVIDMMX){
                 if(mm_flags & MM_MMXEXT){
@@ -2702,6 +2864,8 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
         c->avg_no_rnd_pixels_tab[1][2] = avg_no_rnd_pixels8_y2_mmx;
         c->avg_no_rnd_pixels_tab[1][3] = avg_no_rnd_pixels8_xy2_mmx;
 
+        c->gmc= gmc_mmx;
+
         c->add_bytes= add_bytes_mmx;
 #ifdef CONFIG_ENCODERS
         c->diff_bytes= diff_bytes_mmx;
@@ -2732,7 +2896,14 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
         c->put_h264_chroma_pixels_tab[0]= put_h264_chroma_mc8_mmx;
         c->put_h264_chroma_pixels_tab[1]= put_h264_chroma_mc4_mmx;
 
+        c->h264_idct_dc_add=
+        c->h264_idct_add= ff_h264_idct_add_mmx;
+        c->h264_idct8_dc_add=
+        c->h264_idct8_add= ff_h264_idct8_add_mmx;
+
         if (mm_flags & MM_MMXEXT) {
+            c->prefetch = prefetch_mmx2;
+
             c->put_pixels_tab[0][1] = put_pixels16_x2_mmx2;
             c->put_pixels_tab[0][2] = put_pixels16_y2_mmx2;
 
@@ -2753,7 +2924,8 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
             c->vsad[4]= vsad_intra16_mmx2;
 #endif //CONFIG_ENCODERS
 
-            c->h264_idct_add= ff_h264_idct_add_mmx2;
+            c->h264_idct_dc_add= ff_h264_idct_dc_add_mmx2;
+            c->h264_idct8_dc_add= ff_h264_idct8_dc_add_mmx2;
 
             if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
                 c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_mmx2;
@@ -2831,6 +3003,8 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
 
             c->avg_h264_chroma_pixels_tab[0]= avg_h264_chroma_mc8_mmx2;
             c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_mmx2;
+            c->avg_h264_chroma_pixels_tab[2]= avg_h264_chroma_mc2_mmx2;
+            c->put_h264_chroma_pixels_tab[2]= put_h264_chroma_mc2_mmx2;
             c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_mmx2;
             c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_mmx2;
             c->h264_v_loop_filter_chroma= h264_v_loop_filter_chroma_mmx2;
@@ -2856,10 +3030,16 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
             c->biweight_h264_pixels_tab[6]= ff_h264_biweight_4x4_mmx2;
             c->biweight_h264_pixels_tab[7]= ff_h264_biweight_4x2_mmx2;
 
+#ifdef CONFIG_CAVS_DECODER
+            ff_cavsdsp_init_mmx2(c, avctx);
+#endif
+
 #ifdef CONFIG_ENCODERS
             c->sub_hfyu_median_prediction= sub_hfyu_median_prediction_mmx2;
 #endif //CONFIG_ENCODERS
         } else if (mm_flags & MM_3DNOW) {
+            c->prefetch = prefetch_3dnow;
+
             c->put_pixels_tab[0][1] = put_pixels16_x2_3dnow;
             c->put_pixels_tab[0][2] = put_pixels16_y2_3dnow;
 
@@ -2944,6 +3124,19 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
             c->avg_h264_chroma_pixels_tab[0]= avg_h264_chroma_mc8_3dnow;
             c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_3dnow;
         }
+
+#ifdef CONFIG_SNOW_ENCODER
+        if(mm_flags & MM_SSE2){
+            c->horizontal_compose97i = ff_snow_horizontal_compose97i_sse2;
+            c->vertical_compose97i = ff_snow_vertical_compose97i_sse2;
+            c->inner_add_yblock = ff_snow_inner_add_yblock_sse2;
+        }
+        else{
+            c->horizontal_compose97i = ff_snow_horizontal_compose97i_mmx;
+            c->vertical_compose97i = ff_snow_vertical_compose97i_mmx;
+            c->inner_add_yblock = ff_snow_inner_add_yblock_mmx;
+        }
+#endif
     }
 
 #ifdef CONFIG_ENCODERS
diff --git a/src/libffmpeg/libavcodec/i386/fft_sse.c b/src/libffmpeg/libavcodec/i386/fft_sse.c
index 54851fb94..631848265 100644
--- a/src/libffmpeg/libavcodec/i386/fft_sse.c
+++ b/src/libffmpeg/libavcodec/i386/fft_sse.c
@@ -23,14 +23,14 @@
 
 #include <xmmintrin.h>
 
-static const float p1p1p1m1[4] __attribute__((aligned(16))) =
-    { 1.0, 1.0, 1.0, -1.0 };
+static const int p1p1p1m1[4] __attribute__((aligned(16))) =
+    { 0, 0, 0, 1 << 31 };
 
-static const float p1p1m1p1[4] __attribute__((aligned(16))) =
-    { 1.0, 1.0, -1.0, 1.0 };
+static const int p1p1m1p1[4] __attribute__((aligned(16))) =
+    { 0, 0, 1 << 31, 0 };
 
-static const float p1p1m1m1[4] __attribute__((aligned(16))) =
-    { 1.0, 1.0, -1.0, -1.0 };
+static const int p1p1m1m1[4] __attribute__((aligned(16))) =
+    { 0, 0, 1 << 31, 1 << 31 };
 
 #if 0
 static void print_v4sf(const char *str, __m128 a)
@@ -58,7 +58,6 @@ void ff_fft_calc_sse(FFTContext *s, FFTComplex *z)
 
         r = (__m128 *)&z[0];
         c1 = *(__m128 *)p1p1m1m1;
-        c2 = *(__m128 *)p1p1p1m1;
         if (s->inverse)
             c2 = *(__m128 *)p1p1m1p1;
         else
@@ -68,19 +67,20 @@ void ff_fft_calc_sse(FFTContext *s, FFTComplex *z)
         do {
             a = r[0];
             b = _mm_shuffle_ps(a, a, _MM_SHUFFLE(1, 0, 3, 2));
-            a = _mm_mul_ps(a, c1);
+            a = _mm_xor_ps(a, c1);
             /* do the pass 0 butterfly */
             a = _mm_add_ps(a, b);
 
             a1 = r[1];
             b = _mm_shuffle_ps(a1, a1, _MM_SHUFFLE(1, 0, 3, 2));
-            a1 = _mm_mul_ps(a1, c1);
+            a1 = _mm_xor_ps(a1, c1);
             /* do the pass 0 butterfly */
             b = _mm_add_ps(a1, b);
 
             /* multiply third by -i */
+            /* by toggling the sign bit */
             b = _mm_shuffle_ps(b, b, _MM_SHUFFLE(2, 3, 1, 0));
-            b = _mm_mul_ps(b, c2);
+            b = _mm_xor_ps(b, c2);
 
             /* do the pass 1 butterfly */
             r[0] = _mm_add_ps(a, b);
diff --git a/src/libffmpeg/libavcodec/i386/h264dsp_mmx.c b/src/libffmpeg/libavcodec/i386/h264dsp_mmx.c
index 8ab58f389..ac4ad6401 100644
--- a/src/libffmpeg/libavcodec/i386/h264dsp_mmx.c
+++ b/src/libffmpeg/libavcodec/i386/h264dsp_mmx.c
@@ -46,11 +46,6 @@
     SUMSUBD2_AB( s13, d13, t )\
     SUMSUB_BADC( d13, s02, s13, d02 )
 
-#define SBUTTERFLY(a,b,t,n)\
-    "movq " #a ", " #t "                \n\t" /* abcd */\
-    "punpckl" #n " " #b ", " #a "       \n\t" /* aebf */\
-    "punpckh" #n " " #b ", " #t "       \n\t" /* cgdh */\
-
 #define TRANSPOSE4(a,b,c,d,t)\
     SBUTTERFLY(a,b,t,wd) /* a=aebf t=cgdh */\
     SBUTTERFLY(c,d,b,wd) /* c=imjn b=kolp */\
@@ -65,7 +60,7 @@
     "packuswb  "#z",    "#p" \n\t"\
     "movd      "#p",    (%0) \n\t"
 
-void ff_h264_idct_add_mmx2(uint8_t *dst, int16_t *block, int stride)
+static void ff_h264_idct_add_mmx(uint8_t *dst, int16_t *block, int stride)
 {
     /* Load dct coeffs */
     asm volatile(
@@ -104,6 +99,208 @@ void ff_h264_idct_add_mmx2(uint8_t *dst, int16_t *block, int stride)
     );
 }
 
+static inline void h264_idct8_1d(int16_t *block)
+{
+    asm volatile(
+        "movq 112(%0), %%mm7  \n\t"
+        "movq  80(%0), %%mm5  \n\t"
+        "movq  48(%0), %%mm3  \n\t"
+        "movq  16(%0), %%mm1  \n\t"
+
+        "movq   %%mm7, %%mm4  \n\t"
+        "movq   %%mm3, %%mm6  \n\t"
+        "movq   %%mm5, %%mm0  \n\t"
+        "movq   %%mm7, %%mm2  \n\t"
+        "psraw  $1,    %%mm4  \n\t"
+        "psraw  $1,    %%mm6  \n\t"
+        "psubw  %%mm7, %%mm0  \n\t"
+        "psubw  %%mm6, %%mm2  \n\t"
+        "psubw  %%mm4, %%mm0  \n\t"
+        "psubw  %%mm3, %%mm2  \n\t"
+        "psubw  %%mm3, %%mm0  \n\t"
+        "paddw  %%mm1, %%mm2  \n\t"
+
+        "movq   %%mm5, %%mm4  \n\t"
+        "movq   %%mm1, %%mm6  \n\t"
+        "psraw  $1,    %%mm4  \n\t"
+        "psraw  $1,    %%mm6  \n\t"
+        "paddw  %%mm5, %%mm4  \n\t"
+        "paddw  %%mm1, %%mm6  \n\t"
+        "paddw  %%mm7, %%mm4  \n\t"
+        "paddw  %%mm5, %%mm6  \n\t"
+        "psubw  %%mm1, %%mm4  \n\t"
+        "paddw  %%mm3, %%mm6  \n\t"
+
+        "movq   %%mm0, %%mm1  \n\t"
+        "movq   %%mm4, %%mm3  \n\t"
+        "movq   %%mm2, %%mm5  \n\t"
+        "movq   %%mm6, %%mm7  \n\t"
+        "psraw  $2,    %%mm6  \n\t"
+        "psraw  $2,    %%mm3  \n\t"
+        "psraw  $2,    %%mm5  \n\t"
+        "psraw  $2,    %%mm0  \n\t"
+        "paddw  %%mm6, %%mm1  \n\t"
+        "paddw  %%mm2, %%mm3  \n\t"
+        "psubw  %%mm4, %%mm5  \n\t"
+        "psubw  %%mm0, %%mm7  \n\t"
+
+        "movq  32(%0), %%mm2  \n\t"
+        "movq  96(%0), %%mm6  \n\t"
+        "movq   %%mm2, %%mm4  \n\t"
+        "movq   %%mm6, %%mm0  \n\t"
+        "psraw  $1,    %%mm4  \n\t"
+        "psraw  $1,    %%mm6  \n\t"
+        "psubw  %%mm0, %%mm4  \n\t"
+        "paddw  %%mm2, %%mm6  \n\t"
+
+        "movq    (%0), %%mm2  \n\t"
+        "movq  64(%0), %%mm0  \n\t"
+        SUMSUB_BA( %%mm0, %%mm2 )
+        SUMSUB_BA( %%mm6, %%mm0 )
+        SUMSUB_BA( %%mm4, %%mm2 )
+        SUMSUB_BA( %%mm7, %%mm6 )
+        SUMSUB_BA( %%mm5, %%mm4 )
+        SUMSUB_BA( %%mm3, %%mm2 )
+        SUMSUB_BA( %%mm1, %%mm0 )
+        :: "r"(block)
+    );
+}
+
+static void ff_h264_idct8_add_mmx(uint8_t *dst, int16_t *block, int stride)
+{
+    int i;
+    int16_t __attribute__ ((aligned(8))) b2[64];
+
+    block[0] += 32;
+
+    for(i=0; i<2; i++){
+        uint64_t tmp;
+
+        h264_idct8_1d(block+4*i);
+
+        asm volatile(
+            "movq   %%mm7,    %0   \n\t"
+            TRANSPOSE4( %%mm0, %%mm2, %%mm4, %%mm6, %%mm7 )
+            "movq   %%mm0,  8(%1)  \n\t"
+            "movq   %%mm6, 24(%1)  \n\t"
+            "movq   %%mm7, 40(%1)  \n\t"
+            "movq   %%mm4, 56(%1)  \n\t"
+            "movq    %0,    %%mm7  \n\t"
+            TRANSPOSE4( %%mm7, %%mm5, %%mm3, %%mm1, %%mm0 )
+            "movq   %%mm7,   (%1)  \n\t"
+            "movq   %%mm1, 16(%1)  \n\t"
+            "movq   %%mm0, 32(%1)  \n\t"
+            "movq   %%mm3, 48(%1)  \n\t"
+            : "=m"(tmp)
+            : "r"(b2+32*i)
+            : "memory"
+        );
+    }
+
+    for(i=0; i<2; i++){
+        h264_idct8_1d(b2+4*i);
+
+        asm volatile(
+            "psraw     $6, %%mm7  \n\t"
+            "psraw     $6, %%mm6  \n\t"
+            "psraw     $6, %%mm5  \n\t"
+            "psraw     $6, %%mm4  \n\t"
+            "psraw     $6, %%mm3  \n\t"
+            "psraw     $6, %%mm2  \n\t"
+            "psraw     $6, %%mm1  \n\t"
+            "psraw     $6, %%mm0  \n\t"
+
+            "movq   %%mm7,    (%0)  \n\t"
+            "movq   %%mm5,  16(%0)  \n\t"
+            "movq   %%mm3,  32(%0)  \n\t"
+            "movq   %%mm1,  48(%0)  \n\t"
+            "movq   %%mm0,  64(%0)  \n\t"
+            "movq   %%mm2,  80(%0)  \n\t"
+            "movq   %%mm4,  96(%0)  \n\t"
+            "movq   %%mm6, 112(%0)  \n\t"
+            :: "r"(b2+4*i)
+            : "memory"
+        );
+    }
+
+    add_pixels_clamped_mmx(b2, dst, stride);
+}
+
+static void ff_h264_idct_dc_add_mmx2(uint8_t *dst, int16_t *block, int stride)
+{
+    int dc = (block[0] + 32) >> 6;
+    asm volatile(
+        "movd          %0, %%mm0 \n\t"
+        "pshufw $0, %%mm0, %%mm0 \n\t"
+        "pxor       %%mm1, %%mm1 \n\t"
+        "psubw      %%mm0, %%mm1 \n\t"
+        "packuswb   %%mm0, %%mm0 \n\t"
+        "packuswb   %%mm1, %%mm1 \n\t"
+        ::"r"(dc)
+    );
+    asm volatile(
+        "movd          %0, %%mm2 \n\t"
+        "movd          %1, %%mm3 \n\t"
+        "movd          %2, %%mm4 \n\t"
+        "movd          %3, %%mm5 \n\t"
+        "paddusb    %%mm0, %%mm2 \n\t"
+        "paddusb    %%mm0, %%mm3 \n\t"
+        "paddusb    %%mm0, %%mm4 \n\t"
+        "paddusb    %%mm0, %%mm5 \n\t"
+        "psubusb    %%mm1, %%mm2 \n\t"
+        "psubusb    %%mm1, %%mm3 \n\t"
+        "psubusb    %%mm1, %%mm4 \n\t"
+        "psubusb    %%mm1, %%mm5 \n\t"
+        "movd       %%mm2, %0    \n\t"
+        "movd       %%mm3, %1    \n\t"
+        "movd       %%mm4, %2    \n\t"
+        "movd       %%mm5, %3    \n\t"
+        :"+m"(*(uint32_t*)(dst+0*stride)),
+         "+m"(*(uint32_t*)(dst+1*stride)),
+         "+m"(*(uint32_t*)(dst+2*stride)),
+         "+m"(*(uint32_t*)(dst+3*stride))
+    );
+}
+
+static void ff_h264_idct8_dc_add_mmx2(uint8_t *dst, int16_t *block, int stride)
+{
+    int dc = (block[0] + 32) >> 6;
+    int y;
+    asm volatile(
+        "movd          %0, %%mm0 \n\t"
+        "pshufw $0, %%mm0, %%mm0 \n\t"
+        "pxor       %%mm1, %%mm1 \n\t"
+        "psubw      %%mm0, %%mm1 \n\t"
+        "packuswb   %%mm0, %%mm0 \n\t"
+        "packuswb   %%mm1, %%mm1 \n\t"
+        ::"r"(dc)
+    );
+    for(y=2; y--; dst += 4*stride){
+    asm volatile(
+        "movq          %0, %%mm2 \n\t"
+        "movq          %1, %%mm3 \n\t"
+        "movq          %2, %%mm4 \n\t"
+        "movq          %3, %%mm5 \n\t"
+        "paddusb    %%mm0, %%mm2 \n\t"
+        "paddusb    %%mm0, %%mm3 \n\t"
+        "paddusb    %%mm0, %%mm4 \n\t"
+        "paddusb    %%mm0, %%mm5 \n\t"
+        "psubusb    %%mm1, %%mm2 \n\t"
+        "psubusb    %%mm1, %%mm3 \n\t"
+        "psubusb    %%mm1, %%mm4 \n\t"
+        "psubusb    %%mm1, %%mm5 \n\t"
+        "movq       %%mm2, %0    \n\t"
+        "movq       %%mm3, %1    \n\t"
+        "movq       %%mm4, %2    \n\t"
+        "movq       %%mm5, %3    \n\t"
+        :"+m"(*(uint64_t*)(dst+0*stride)),
+         "+m"(*(uint64_t*)(dst+1*stride)),
+         "+m"(*(uint64_t*)(dst+2*stride)),
+         "+m"(*(uint64_t*)(dst+3*stride))
+    );
+    }
+}
+
 
 /***********************************/
 /* deblocking */
@@ -441,6 +638,50 @@ static void OPNAME ## h264_qpel4_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, i
         : "memory"\
     );\
 }\
+static void OPNAME ## h264_qpel4_h_lowpass_l2_ ## MMX(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride){\
+    int h=4;\
+    asm volatile(\
+        "pxor %%mm7, %%mm7          \n\t"\
+        "movq %0, %%mm4             \n\t"\
+        "movq %1, %%mm5             \n\t"\
+        :: "m"(ff_pw_5), "m"(ff_pw_16)\
+    );\
+    do{\
+    asm volatile(\
+        "movd  -1(%0), %%mm1        \n\t"\
+        "movd    (%0), %%mm2        \n\t"\
+        "movd   1(%0), %%mm3        \n\t"\
+        "movd   2(%0), %%mm0        \n\t"\
+        "punpcklbw %%mm7, %%mm1     \n\t"\
+        "punpcklbw %%mm7, %%mm2     \n\t"\
+        "punpcklbw %%mm7, %%mm3     \n\t"\
+        "punpcklbw %%mm7, %%mm0     \n\t"\
+        "paddw %%mm0, %%mm1         \n\t"\
+        "paddw %%mm3, %%mm2         \n\t"\
+        "movd  -2(%0), %%mm0        \n\t"\
+        "movd   3(%0), %%mm3        \n\t"\
+        "punpcklbw %%mm7, %%mm0     \n\t"\
+        "punpcklbw %%mm7, %%mm3     \n\t"\
+        "paddw %%mm3, %%mm0         \n\t"\
+        "psllw $2, %%mm2            \n\t"\
+        "psubw %%mm1, %%mm2         \n\t"\
+        "pmullw %%mm4, %%mm2        \n\t"\
+        "paddw %%mm5, %%mm0         \n\t"\
+        "paddw %%mm2, %%mm0         \n\t"\
+        "movd   (%2), %%mm3         \n\t"\
+        "psraw $5, %%mm0            \n\t"\
+        "packuswb %%mm0, %%mm0      \n\t"\
+        PAVGB" %%mm3, %%mm0         \n\t"\
+        OP(%%mm0, (%1),%%mm6, d)\
+        "add %4, %0                 \n\t"\
+        "add %4, %1                 \n\t"\
+        "add %3, %2                 \n\t"\
+        : "+a"(src), "+c"(dst), "+d"(src2)\
+        : "D"((long)src2Stride), "S"((long)dstStride)\
+        : "memory"\
+    );\
+    }while(--h);\
+}\
 static void OPNAME ## h264_qpel4_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
     src -= 2*srcStride;\
     asm volatile(\
@@ -591,11 +832,74 @@ static void OPNAME ## h264_qpel8_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, i
     );\
 }\
 \
-static void OPNAME ## h264_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
-    int h= 2;\
+static void OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride){\
+    int h=8;\
+    asm volatile(\
+        "pxor %%mm7, %%mm7          \n\t"\
+        "movq %0, %%mm6             \n\t"\
+        :: "m"(ff_pw_5)\
+    );\
+    do{\
+    asm volatile(\
+        "movq    (%0), %%mm0        \n\t"\
+        "movq   1(%0), %%mm2        \n\t"\
+        "movq %%mm0, %%mm1          \n\t"\
+        "movq %%mm2, %%mm3          \n\t"\
+        "punpcklbw %%mm7, %%mm0     \n\t"\
+        "punpckhbw %%mm7, %%mm1     \n\t"\
+        "punpcklbw %%mm7, %%mm2     \n\t"\
+        "punpckhbw %%mm7, %%mm3     \n\t"\
+        "paddw %%mm2, %%mm0         \n\t"\
+        "paddw %%mm3, %%mm1         \n\t"\
+        "psllw $2, %%mm0            \n\t"\
+        "psllw $2, %%mm1            \n\t"\
+        "movq   -1(%0), %%mm2       \n\t"\
+        "movq    2(%0), %%mm4       \n\t"\
+        "movq %%mm2, %%mm3          \n\t"\
+        "movq %%mm4, %%mm5          \n\t"\
+        "punpcklbw %%mm7, %%mm2     \n\t"\
+        "punpckhbw %%mm7, %%mm3     \n\t"\
+        "punpcklbw %%mm7, %%mm4     \n\t"\
+        "punpckhbw %%mm7, %%mm5     \n\t"\
+        "paddw %%mm4, %%mm2         \n\t"\
+        "paddw %%mm3, %%mm5         \n\t"\
+        "psubw %%mm2, %%mm0         \n\t"\
+        "psubw %%mm5, %%mm1         \n\t"\
+        "pmullw %%mm6, %%mm0        \n\t"\
+        "pmullw %%mm6, %%mm1        \n\t"\
+        "movd   -2(%0), %%mm2       \n\t"\
+        "movd    7(%0), %%mm5       \n\t"\
+        "punpcklbw %%mm7, %%mm2     \n\t"\
+        "punpcklbw %%mm7, %%mm5     \n\t"\
+        "paddw %%mm3, %%mm2         \n\t"\
+        "paddw %%mm5, %%mm4         \n\t"\
+        "movq %5, %%mm5             \n\t"\
+        "paddw %%mm5, %%mm2         \n\t"\
+        "paddw %%mm5, %%mm4         \n\t"\
+        "paddw %%mm2, %%mm0         \n\t"\
+        "paddw %%mm4, %%mm1         \n\t"\
+        "psraw $5, %%mm0            \n\t"\
+        "psraw $5, %%mm1            \n\t"\
+        "movq (%2), %%mm4           \n\t"\
+        "packuswb %%mm1, %%mm0      \n\t"\
+        PAVGB" %%mm4, %%mm0         \n\t"\
+        OP(%%mm0, (%1),%%mm5, q)\
+        "add %4, %0                 \n\t"\
+        "add %4, %1                 \n\t"\
+        "add %3, %2                 \n\t"\
+        : "+a"(src), "+c"(dst), "+d"(src2)\
+        : "D"((long)src2Stride), "S"((long)dstStride),\
+          "m"(ff_pw_16)\
+        : "memory"\
+    );\
+    }while(--h);\
+}\
+\
+static inline void OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
+    int w= 2;\
     src -= 2*srcStride;\
     \
-    while(h--){\
+    while(w--){\
       asm volatile(\
         "pxor %%mm7, %%mm7          \n\t"\
         "movd (%0), %%mm0           \n\t"\
@@ -626,13 +930,29 @@ static void OPNAME ## h264_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, i
         : "S"((long)srcStride), "D"((long)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
         : "memory"\
      );\
-     src += 4-13*srcStride;\
-     dst +=  4-8*dstStride;\
+     if(h==16){\
+        asm volatile(\
+            QPEL_H264V(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP)\
+            QPEL_H264V(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP)\
+            QPEL_H264V(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, OP)\
+            QPEL_H264V(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, OP)\
+            QPEL_H264V(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP)\
+            QPEL_H264V(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP)\
+            QPEL_H264V(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP)\
+            QPEL_H264V(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP)\
+            \
+           : "+a"(src), "+c"(dst)\
+           : "S"((long)srcStride), "D"((long)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
+           : "memory"\
+        );\
+     }\
+     src += 4-(h+5)*srcStride;\
+     dst += 4-h*dstStride;\
    }\
 }\
-static void OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
-    int h=8;\
-    int w=4;\
+static inline void OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride, int size){\
+    int h = size;\
+    int w = (size+8)>>2;\
     src -= 2*srcStride+2;\
     while(w--){\
         asm volatile(\
@@ -652,23 +972,40 @@ static void OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp,
             "punpcklbw %%mm7, %%mm2 \n\t"\
             "punpcklbw %%mm7, %%mm3 \n\t"\
             "punpcklbw %%mm7, %%mm4 \n\t"\
-            QPEL_H264HV(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, 0*8*4)\
-            QPEL_H264HV(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, 1*8*4)\
-            QPEL_H264HV(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, 2*8*4)\
-            QPEL_H264HV(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, 3*8*4)\
-            QPEL_H264HV(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, 4*8*4)\
-            QPEL_H264HV(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, 5*8*4)\
-            QPEL_H264HV(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, 6*8*4)\
-            QPEL_H264HV(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, 7*8*4)\
-             \
+            QPEL_H264HV(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, 0*48)\
+            QPEL_H264HV(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, 1*48)\
+            QPEL_H264HV(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, 2*48)\
+            QPEL_H264HV(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, 3*48)\
+            QPEL_H264HV(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, 4*48)\
+            QPEL_H264HV(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, 5*48)\
+            QPEL_H264HV(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, 6*48)\
+            QPEL_H264HV(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, 7*48)\
             : "+a"(src)\
             : "c"(tmp), "S"((long)srcStride), "m"(ff_pw_5)\
             : "memory"\
         );\
+        if(size==16){\
+            asm volatile(\
+                QPEL_H264HV(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1,  8*48)\
+                QPEL_H264HV(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2,  9*48)\
+                QPEL_H264HV(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, 10*48)\
+                QPEL_H264HV(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, 11*48)\
+                QPEL_H264HV(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, 12*48)\
+                QPEL_H264HV(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, 13*48)\
+                QPEL_H264HV(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, 14*48)\
+                QPEL_H264HV(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, 15*48)\
+                : "+a"(src)\
+                : "c"(tmp), "S"((long)srcStride), "m"(ff_pw_5)\
+                : "memory"\
+            );\
+        }\
         tmp += 4;\
-        src += 4 - 13*srcStride;\
+        src += 4 - (size+5)*srcStride;\
     }\
-    tmp -= 4*4;\
+    tmp -= size+8;\
+    w = size>>4;\
+    do{\
+    h = size;\
     asm volatile(\
         "movq %4, %%mm6             \n\t"\
         "1:                         \n\t"\
@@ -702,7 +1039,7 @@ static void OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp,
         "psraw $6, %%mm3            \n\t"\
         "packuswb %%mm3, %%mm0      \n\t"\
         OP(%%mm0, (%1),%%mm7, q)\
-        "add $32, %0                \n\t"\
+        "add $48, %0                \n\t"\
         "add %3, %1                 \n\t"\
         "decl %2                    \n\t"\
         " jnz 1b                    \n\t"\
@@ -710,14 +1047,17 @@ static void OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp,
         : "S"((long)dstStride), "m"(ff_pw_32)\
         : "memory"\
     );\
+    tmp += 8 - size*24;\
+    dst += 8 - size*dstStride;\
+    }while(w--);\
+}\
+\
+static void OPNAME ## h264_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
+    OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 8);\
 }\
 static void OPNAME ## h264_qpel16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
-    OPNAME ## h264_qpel8_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
-    OPNAME ## h264_qpel8_v_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
-    src += 8*srcStride;\
-    dst += 8*dstStride;\
-    OPNAME ## h264_qpel8_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
-    OPNAME ## h264_qpel8_v_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
+    OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 16);\
+    OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
 }\
 \
 static void OPNAME ## h264_qpel16_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
@@ -729,14 +1069,88 @@ static void OPNAME ## h264_qpel16_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src,
     OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
 }\
 \
-static void OPNAME ## h264_qpel16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
-    OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(dst  , tmp  , src  , dstStride, tmpStride, srcStride);\
-    OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(dst+8, tmp  , src+8, dstStride, tmpStride, srcStride);\
-    src += 8*srcStride;\
+static void OPNAME ## h264_qpel16_h_lowpass_l2_ ## MMX(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride){\
+    OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
+    OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
+    src += 8*dstStride;\
     dst += 8*dstStride;\
-    OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(dst  , tmp  , src  , dstStride, tmpStride, srcStride);\
-    OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(dst+8, tmp  , src+8, dstStride, tmpStride, srcStride);\
+    src2 += 8*src2Stride;\
+    OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
+    OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
+}\
+\
+static void OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
+    OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst  , tmp  , src  , dstStride, tmpStride, srcStride, 8);\
+}\
+\
+static void OPNAME ## h264_qpel16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
+    OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst  , tmp  , src  , dstStride, tmpStride, srcStride, 16);\
+}\
+\
+static void OPNAME ## pixels4_l2_shift5_ ## MMX(uint8_t *dst, int16_t *src16, uint8_t *src8, int dstStride, int src8Stride, int h)\
+{\
+    asm volatile(\
+        "movq       %5,  %%mm6          \n\t"\
+        "movq      (%1), %%mm0          \n\t"\
+        "movq    24(%1), %%mm1          \n\t"\
+        "paddw    %%mm6, %%mm0          \n\t"\
+        "paddw    %%mm6, %%mm1          \n\t"\
+        "psraw      $5,  %%mm0          \n\t"\
+        "psraw      $5,  %%mm1          \n\t"\
+        "packuswb %%mm0, %%mm0          \n\t"\
+        "packuswb %%mm1, %%mm1          \n\t"\
+        PAVGB"     (%0), %%mm0          \n\t"\
+        PAVGB"  (%0,%3), %%mm1          \n\t"\
+        OP(%%mm0, (%2),    %%mm4, d)\
+        OP(%%mm1, (%2,%4), %%mm5, d)\
+        "lea  (%0,%3,2), %0             \n\t"\
+        "lea  (%2,%4,2), %2             \n\t"\
+        "movq    48(%1), %%mm0          \n\t"\
+        "movq    72(%1), %%mm1          \n\t"\
+        "paddw    %%mm6, %%mm0          \n\t"\
+        "paddw    %%mm6, %%mm1          \n\t"\
+        "psraw      $5,  %%mm0          \n\t"\
+        "psraw      $5,  %%mm1          \n\t"\
+        "packuswb %%mm0, %%mm0          \n\t"\
+        "packuswb %%mm1, %%mm1          \n\t"\
+        PAVGB"     (%0), %%mm0          \n\t"\
+        PAVGB"  (%0,%3), %%mm1          \n\t"\
+        OP(%%mm0, (%2),    %%mm4, d)\
+        OP(%%mm1, (%2,%4), %%mm5, d)\
+        :"+a"(src8), "+c"(src16), "+d"(dst)\
+        :"S"((long)src8Stride), "D"((long)dstStride), "m"(ff_pw_16)\
+        :"memory");\
+}\
+static void OPNAME ## pixels8_l2_shift5_ ## MMX(uint8_t *dst, int16_t *src16, uint8_t *src8, int dstStride, int src8Stride, int h)\
+{\
+    asm volatile(\
+        "movq       %0,  %%mm6          \n\t"\
+        ::"m"(ff_pw_16)\
+        );\
+    while(h--){\
+    asm volatile(\
+        "movq      (%1), %%mm0          \n\t"\
+        "movq     8(%1), %%mm1          \n\t"\
+        "paddw    %%mm6, %%mm0          \n\t"\
+        "paddw    %%mm6, %%mm1          \n\t"\
+        "psraw      $5,  %%mm0          \n\t"\
+        "psraw      $5,  %%mm1          \n\t"\
+        "packuswb %%mm1, %%mm0          \n\t"\
+        PAVGB"     (%0), %%mm0          \n\t"\
+        OP(%%mm0, (%2), %%mm5, q)\
+        ::"a"(src8), "c"(src16), "d"(dst)\
+        :"memory");\
+        src8 += src8Stride;\
+        src16 += 24;\
+        dst += dstStride;\
+    }\
 }\
+static void OPNAME ## pixels16_l2_shift5_ ## MMX(uint8_t *dst, int16_t *src16, uint8_t *src8, int dstStride, int src8Stride, int h)\
+{\
+    OPNAME ## pixels8_l2_shift5_ ## MMX(dst  , src16  , src8  , dstStride, src8Stride, h);\
+    OPNAME ## pixels8_l2_shift5_ ## MMX(dst+8, src16+8, src8+8, dstStride, src8Stride, h);\
+}\
+
 
 #define H264_MC(OPNAME, SIZE, MMX) \
 static void OPNAME ## h264_qpel ## SIZE ## _mc00_ ## MMX (uint8_t *dst, uint8_t *src, int stride){\
@@ -744,10 +1158,7 @@ static void OPNAME ## h264_qpel ## SIZE ## _mc00_ ## MMX (uint8_t *dst, uint8_t
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc10_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[SIZE*SIZE/8];\
-    uint8_t * const half= (uint8_t*)temp;\
-    put_h264_qpel ## SIZE ## _h_lowpass_ ## MMX(half, src, SIZE, stride);\
-    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src, half, stride, stride, SIZE);\
+    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, src, stride, stride);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc20_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
@@ -755,10 +1166,7 @@ static void OPNAME ## h264_qpel ## SIZE ## _mc20_ ## MMX(uint8_t *dst, uint8_t *
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc30_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[SIZE*SIZE/8];\
-    uint8_t * const half= (uint8_t*)temp;\
-    put_h264_qpel ## SIZE ## _h_lowpass_ ## MMX(half, src, SIZE, stride);\
-    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src+1, half, stride, stride, SIZE);\
+    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, src+1, stride, stride);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc01_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
@@ -780,89 +1188,72 @@ static void OPNAME ## h264_qpel ## SIZE ## _mc03_ ## MMX(uint8_t *dst, uint8_t *
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc11_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[SIZE*SIZE/4];\
-    uint8_t * const halfH= (uint8_t*)temp;\
-    uint8_t * const halfV= ((uint8_t*)temp) + SIZE*SIZE;\
-    put_h264_qpel ## SIZE ## _h_lowpass_ ## MMX(halfH, src, SIZE, stride);\
+    uint64_t temp[SIZE*SIZE/8];\
+    uint8_t * const halfV= (uint8_t*)temp;\
     put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(halfV, src, SIZE, stride);\
-    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, halfH, halfV, stride, SIZE, SIZE);\
+    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, halfV, stride, SIZE);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc31_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[SIZE*SIZE/4];\
-    uint8_t * const halfH= (uint8_t*)temp;\
-    uint8_t * const halfV= ((uint8_t*)temp) + SIZE*SIZE;\
-    put_h264_qpel ## SIZE ## _h_lowpass_ ## MMX(halfH, src, SIZE, stride);\
+    uint64_t temp[SIZE*SIZE/8];\
+    uint8_t * const halfV= (uint8_t*)temp;\
     put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(halfV, src+1, SIZE, stride);\
-    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, halfH, halfV, stride, SIZE, SIZE);\
+    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, halfV, stride, SIZE);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc13_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[SIZE*SIZE/4];\
-    uint8_t * const halfH= (uint8_t*)temp;\
-    uint8_t * const halfV= ((uint8_t*)temp) + SIZE*SIZE;\
-    put_h264_qpel ## SIZE ## _h_lowpass_ ## MMX(halfH, src + stride, SIZE, stride);\
+    uint64_t temp[SIZE*SIZE/8];\
+    uint8_t * const halfV= (uint8_t*)temp;\
     put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(halfV, src, SIZE, stride);\
-    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, halfH, halfV, stride, SIZE, SIZE);\
+    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, halfV, stride, SIZE);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc33_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[SIZE*SIZE/4];\
-    uint8_t * const halfH= (uint8_t*)temp;\
-    uint8_t * const halfV= ((uint8_t*)temp) + SIZE*SIZE;\
-    put_h264_qpel ## SIZE ## _h_lowpass_ ## MMX(halfH, src + stride, SIZE, stride);\
+    uint64_t temp[SIZE*SIZE/8];\
+    uint8_t * const halfV= (uint8_t*)temp;\
     put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(halfV, src+1, SIZE, stride);\
-    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, halfH, halfV, stride, SIZE, SIZE);\
+    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, halfV, stride, SIZE);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc22_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[SIZE*(SIZE+8)/4];\
+    uint64_t temp[SIZE*(SIZE<8?12:24)/4];\
     int16_t * const tmp= (int16_t*)temp;\
     OPNAME ## h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(dst, tmp, src, stride, SIZE, stride);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc21_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[SIZE*(SIZE+8)/4 + SIZE*SIZE/4];\
-    uint8_t * const halfH= (uint8_t*)temp;\
-    uint8_t * const halfHV= ((uint8_t*)temp) + SIZE*SIZE;\
-    int16_t * const tmp= ((int16_t*)temp) + SIZE*SIZE;\
-    put_h264_qpel ## SIZE ## _h_lowpass_ ## MMX(halfH, src, SIZE, stride);\
+    uint64_t temp[SIZE*(SIZE<8?12:24)/4 + SIZE*SIZE/8];\
+    uint8_t * const halfHV= (uint8_t*)temp;\
+    int16_t * const tmp= ((int16_t*)temp) + SIZE*SIZE/2;\
     put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, tmp, src, SIZE, SIZE, stride);\
-    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, halfH, halfHV, stride, SIZE, SIZE);\
+    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, halfHV, stride, SIZE);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc23_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[SIZE*(SIZE+8)/4 + SIZE*SIZE/4];\
-    uint8_t * const halfH= (uint8_t*)temp;\
-    uint8_t * const halfHV= ((uint8_t*)temp) + SIZE*SIZE;\
-    int16_t * const tmp= ((int16_t*)temp) + SIZE*SIZE;\
-    put_h264_qpel ## SIZE ## _h_lowpass_ ## MMX(halfH, src + stride, SIZE, stride);\
+    uint64_t temp[SIZE*(SIZE<8?12:24)/4 + SIZE*SIZE/8];\
+    uint8_t * const halfHV= (uint8_t*)temp;\
+    int16_t * const tmp= ((int16_t*)temp) + SIZE*SIZE/2;\
     put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, tmp, src, SIZE, SIZE, stride);\
-    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, halfH, halfHV, stride, SIZE, SIZE);\
+    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, halfHV, stride, SIZE);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc12_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[SIZE*(SIZE+8)/4 + SIZE*SIZE/4];\
-    uint8_t * const halfV= (uint8_t*)temp;\
-    uint8_t * const halfHV= ((uint8_t*)temp) + SIZE*SIZE;\
-    int16_t * const tmp= ((int16_t*)temp) + SIZE*SIZE;\
-    put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(halfV, src, SIZE, stride);\
-    put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, tmp, src, SIZE, SIZE, stride);\
-    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, halfV, halfHV, stride, SIZE, SIZE);\
+    uint64_t temp[SIZE*(SIZE<8?12:24)/4 + SIZE*SIZE/8];\
+    int16_t * const halfV= ((int16_t*)temp) + SIZE*SIZE/2;\
+    uint8_t * const halfHV= ((uint8_t*)temp);\
+    put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
+    OPNAME ## pixels ## SIZE ## _l2_shift5_ ## MMX(dst, halfV+2, halfHV, stride, SIZE, SIZE);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[SIZE*(SIZE+8)/4 + SIZE*SIZE/4];\
-    uint8_t * const halfV= (uint8_t*)temp;\
-    uint8_t * const halfHV= ((uint8_t*)temp) + SIZE*SIZE;\
-    int16_t * const tmp= ((int16_t*)temp) + SIZE*SIZE;\
-    put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(halfV, src+1, SIZE, stride);\
-    put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, tmp, src, SIZE, SIZE, stride);\
-    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, halfV, halfHV, stride, SIZE, SIZE);\
+    uint64_t temp[SIZE*(SIZE<8?12:24)/4 + SIZE*SIZE/8];\
+    int16_t * const halfV= ((int16_t*)temp) + SIZE*SIZE/2;\
+    uint8_t * const halfHV= ((uint8_t*)temp);\
+    put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
+    OPNAME ## pixels ## SIZE ## _l2_shift5_ ## MMX(dst, halfV+3, halfHV, stride, SIZE, SIZE);\
 }\
 
 
-#define PUT_OP(a,b,temp, size) "mov" #size " " #a ", " #b "    \n\t"
 #define AVG_3DNOW_OP(a,b,temp, size) \
 "mov" #size " " #b ", " #temp "   \n\t"\
 "pavgusb " #temp ", " #a "        \n\t"\
@@ -872,10 +1263,14 @@ static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## MMX(uint8_t *dst, uint8_t *
 "pavgb " #temp ", " #a "          \n\t"\
 "mov" #size " " #a ", " #b "      \n\t"
 
+#define PAVGB "pavgusb"
 QPEL_H264(put_,       PUT_OP, 3dnow)
 QPEL_H264(avg_, AVG_3DNOW_OP, 3dnow)
+#undef PAVGB
+#define PAVGB "pavgb"
 QPEL_H264(put_,       PUT_OP, mmx2)
 QPEL_H264(avg_,  AVG_MMX2_OP, mmx2)
+#undef PAVGB
 
 H264_MC(put_, 4, 3dnow)
 H264_MC(put_, 8, 3dnow)
@@ -895,12 +1290,14 @@ H264_MC(avg_, 16,mmx2)
 #define H264_CHROMA_OP4(S,D,T)
 #define H264_CHROMA_MC8_TMPL put_h264_chroma_mc8_mmx
 #define H264_CHROMA_MC4_TMPL put_h264_chroma_mc4_mmx
+#define H264_CHROMA_MC2_TMPL put_h264_chroma_mc2_mmx2
 #define H264_CHROMA_MC8_MV0 put_pixels8_mmx
 #include "dsputil_h264_template_mmx.c"
 #undef H264_CHROMA_OP
 #undef H264_CHROMA_OP4
 #undef H264_CHROMA_MC8_TMPL
 #undef H264_CHROMA_MC4_TMPL
+#undef H264_CHROMA_MC2_TMPL
 #undef H264_CHROMA_MC8_MV0
 
 #define H264_CHROMA_OP(S,D) "pavgb " #S ", " #D " \n\t"
@@ -908,12 +1305,14 @@ H264_MC(avg_, 16,mmx2)
                                "pavgb " #T ", " #D " \n\t"
 #define H264_CHROMA_MC8_TMPL avg_h264_chroma_mc8_mmx2
 #define H264_CHROMA_MC4_TMPL avg_h264_chroma_mc4_mmx2
+#define H264_CHROMA_MC2_TMPL avg_h264_chroma_mc2_mmx2
 #define H264_CHROMA_MC8_MV0 avg_pixels8_mmx2
 #include "dsputil_h264_template_mmx.c"
 #undef H264_CHROMA_OP
 #undef H264_CHROMA_OP4
 #undef H264_CHROMA_MC8_TMPL
 #undef H264_CHROMA_MC4_TMPL
+#undef H264_CHROMA_MC2_TMPL
 #undef H264_CHROMA_MC8_MV0
 
 #define H264_CHROMA_OP(S,D) "pavgusb " #S ", " #D " \n\t"
diff --git a/src/libffmpeg/libavcodec/i386/idct_mmx_xvid.c b/src/libffmpeg/libavcodec/i386/idct_mmx_xvid.c
index 7bc6f5f78..a55d4ea07 100644
--- a/src/libffmpeg/libavcodec/i386/idct_mmx_xvid.c
+++ b/src/libffmpeg/libavcodec/i386/idct_mmx_xvid.c
@@ -20,7 +20,7 @@
 // *  Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 //
 // *
-// * $Id: idct_mmx_xvid.c,v 1.2 2006/02/05 14:11:36 miguelfreitas Exp $
+// * $Id: idct_mmx_xvid.c,v 1.3 2006/08/02 07:02:41 tmmm Exp $
 // *
 // ***************************************************************************/
 
diff --git a/src/libffmpeg/libavcodec/i386/mmx.h b/src/libffmpeg/libavcodec/i386/mmx.h
index df4620e0a..eab051341 100644
--- a/src/libffmpeg/libavcodec/i386/mmx.h
+++ b/src/libffmpeg/libavcodec/i386/mmx.h
@@ -5,22 +5,6 @@
 #ifndef AVCODEC_I386MMX_H
 #define AVCODEC_I386MMX_H
 
-#ifdef ARCH_X86_64
-#  define REG_a "rax"
-#  define REG_b "rbx"
-#  define REG_c "rcx"
-#  define REG_d "rdx"
-#  define REG_D "rdi"
-#  define REG_S "rsi"
-#else
-#  define REG_a "eax"
-#  define REG_b "ebx"
-#  define REG_c "ecx"
-#  define REG_d "edx"
-#  define REG_D "edi"
-#  define REG_S "esi"
-#endif
-
 /*
  * The type of an value that fits in an MMX register (note that long
  * long constant values MUST be suffixed by LL and unsigned long long
diff --git a/src/libffmpeg/libavcodec/i386/motion_est_mmx.c b/src/libffmpeg/libavcodec/i386/motion_est_mmx.c
index c14b79384..edcabcf38 100644
--- a/src/libffmpeg/libavcodec/i386/motion_est_mmx.c
+++ b/src/libffmpeg/libavcodec/i386/motion_est_mmx.c
@@ -20,7 +20,7 @@
  * mostly by Michael Niedermayer <michaelni@gmx.at>
  */
 #include "../dsputil.h"
-#include "mmx.h"
+#include "x86_cpu.h"
 
 static const __attribute__ ((aligned(8))) uint64_t round_tab[3]={
 0x0000000000000000ULL,
diff --git a/src/libffmpeg/libavcodec/i386/mpegvideo_mmx.c b/src/libffmpeg/libavcodec/i386/mpegvideo_mmx.c
index f83df3a19..c00a602bd 100644
--- a/src/libffmpeg/libavcodec/i386/mpegvideo_mmx.c
+++ b/src/libffmpeg/libavcodec/i386/mpegvideo_mmx.c
@@ -23,7 +23,7 @@
 #include "../dsputil.h"
 #include "../mpegvideo.h"
 #include "../avcodec.h"
-#include "mmx.h"
+#include "x86_cpu.h"
 
 extern uint8_t zigzag_direct_noperm[64];
 extern uint16_t inv_zigzag_direct16[64];
@@ -699,7 +699,8 @@ void MPV_common_init_mmx(MpegEncContext *s)
         s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_mmx;
         s->dct_unquantize_mpeg1_intra = dct_unquantize_mpeg1_intra_mmx;
         s->dct_unquantize_mpeg1_inter = dct_unquantize_mpeg1_inter_mmx;
-        s->dct_unquantize_mpeg2_intra = dct_unquantize_mpeg2_intra_mmx;
+        if(!(s->flags & CODEC_FLAG_BITEXACT))
+            s->dct_unquantize_mpeg2_intra = dct_unquantize_mpeg2_intra_mmx;
         s->dct_unquantize_mpeg2_inter = dct_unquantize_mpeg2_inter_mmx;
 
         draw_edges = draw_edges_mmx;
diff --git a/src/libffmpeg/libavcodec/i386/mpegvideo_mmx_template.c b/src/libffmpeg/libavcodec/i386/mpegvideo_mmx_template.c
index 2c50df232..de2ef08e5 100644
--- a/src/libffmpeg/libavcodec/i386/mpegvideo_mmx_template.c
+++ b/src/libffmpeg/libavcodec/i386/mpegvideo_mmx_template.c
@@ -19,6 +19,7 @@
  */
 #undef SPREADW
 #undef PMAXW
+#undef PMAX
 #ifdef HAVE_MMX2
 #define SPREADW(a) "pshufw $0, " #a ", " #a " \n\t"
 #define PMAXW(a,b) "pmaxsw " #a ", " #b "     \n\t"