Merge remote-tracking branch 'qatar/master'
[ffmpeg.git] / libavcodec / x86 / dsputil_mmx.c
index 976008a..c887d6d 100644 (file)
@@ -41,129 +41,129 @@ DECLARE_ALIGNED(8,  const uint64_t, ff_bone) = 0x0101010101010101ULL;
 DECLARE_ALIGNED(8,  const uint64_t, ff_wtwo) = 0x0002000200020002ULL;
 
 DECLARE_ALIGNED(16, const uint64_t, ff_pdw_80000000)[2] =
-{0x8000000080000000ULL, 0x8000000080000000ULL};
-
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_1  ) = {0x0001000100010001ULL, 0x0001000100010001ULL};
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_2  ) = {0x0002000200020002ULL, 0x0002000200020002ULL};
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_3  ) = {0x0003000300030003ULL, 0x0003000300030003ULL};
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_4  ) = {0x0004000400040004ULL, 0x0004000400040004ULL};
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_5  ) = {0x0005000500050005ULL, 0x0005000500050005ULL};
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_8  ) = {0x0008000800080008ULL, 0x0008000800080008ULL};
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_9  ) = {0x0009000900090009ULL, 0x0009000900090009ULL};
-DECLARE_ALIGNED(8,  const uint64_t, ff_pw_15 ) = 0x000F000F000F000FULL;
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_16 ) = {0x0010001000100010ULL, 0x0010001000100010ULL};
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_17 ) = {0x0011001100110011ULL, 0x0011001100110011ULL};
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_18 ) = {0x0012001200120012ULL, 0x0012001200120012ULL};
-DECLARE_ALIGNED(8,  const uint64_t, ff_pw_20 ) = 0x0014001400140014ULL;
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_27 ) = {0x001B001B001B001BULL, 0x001B001B001B001BULL};
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_28 ) = {0x001C001C001C001CULL, 0x001C001C001C001CULL};
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_32 ) = {0x0020002000200020ULL, 0x0020002000200020ULL};
-DECLARE_ALIGNED(8,  const uint64_t, ff_pw_42 ) = 0x002A002A002A002AULL;
-DECLARE_ALIGNED(8,  const uint64_t, ff_pw_53 ) = 0x0035003500350035ULL;
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_63 ) = {0x003F003F003F003FULL, 0x003F003F003F003FULL};
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_64 ) = {0x0040004000400040ULL, 0x0040004000400040ULL};
-DECLARE_ALIGNED(8,  const uint64_t, ff_pw_96 ) = 0x0060006000600060ULL;
-DECLARE_ALIGNED(8,  const uint64_t, ff_pw_128) = 0x0080008000800080ULL;
-DECLARE_ALIGNED(8,  const uint64_t, ff_pw_255) = 0x00ff00ff00ff00ffULL;
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_512) = {0x0200020002000200ULL, 0x0200020002000200ULL};
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_1019)= {0x03FB03FB03FB03FBULL, 0x03FB03FB03FB03FBULL};
-
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pb_0  ) = {0x0000000000000000ULL, 0x0000000000000000ULL};
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pb_1  ) = {0x0101010101010101ULL, 0x0101010101010101ULL};
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pb_3  ) = {0x0303030303030303ULL, 0x0303030303030303ULL};
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pb_4  ) = {0x0404040404040404ULL, 0x0404040404040404ULL};
-DECLARE_ALIGNED(8,  const uint64_t, ff_pb_7  ) = 0x0707070707070707ULL;
-DECLARE_ALIGNED(8,  const uint64_t, ff_pb_1F ) = 0x1F1F1F1F1F1F1F1FULL;
-DECLARE_ALIGNED(8,  const uint64_t, ff_pb_3F ) = 0x3F3F3F3F3F3F3F3FULL;
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pb_80 ) = {0x8080808080808080ULL, 0x8080808080808080ULL};
-DECLARE_ALIGNED(8,  const uint64_t, ff_pb_81 ) = 0x8181818181818181ULL;
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pb_A1 ) = {0xA1A1A1A1A1A1A1A1ULL, 0xA1A1A1A1A1A1A1A1ULL};
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pb_F8 ) = {0xF8F8F8F8F8F8F8F8ULL, 0xF8F8F8F8F8F8F8F8ULL};
-DECLARE_ALIGNED(8,  const uint64_t, ff_pb_FC ) = 0xFCFCFCFCFCFCFCFCULL;
-DECLARE_ALIGNED(16, const xmm_reg,  ff_pb_FE ) = {0xFEFEFEFEFEFEFEFEULL, 0xFEFEFEFEFEFEFEFEULL};
+    { 0x8000000080000000ULL, 0x8000000080000000ULL };
+
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_1)    = { 0x0001000100010001ULL, 0x0001000100010001ULL };
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_2)    = { 0x0002000200020002ULL, 0x0002000200020002ULL };
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_3)    = { 0x0003000300030003ULL, 0x0003000300030003ULL };
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_4)    = { 0x0004000400040004ULL, 0x0004000400040004ULL };
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_5)    = { 0x0005000500050005ULL, 0x0005000500050005ULL };
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_8)    = { 0x0008000800080008ULL, 0x0008000800080008ULL };
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_9)    = { 0x0009000900090009ULL, 0x0009000900090009ULL };
+DECLARE_ALIGNED(8,  const uint64_t, ff_pw_15)   =   0x000F000F000F000FULL;
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_16)   = { 0x0010001000100010ULL, 0x0010001000100010ULL };
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_17)   = { 0x0011001100110011ULL, 0x0011001100110011ULL };
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_18)   = { 0x0012001200120012ULL, 0x0012001200120012ULL };
+DECLARE_ALIGNED(8,  const uint64_t, ff_pw_20)   =   0x0014001400140014ULL;
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_27)   = { 0x001B001B001B001BULL, 0x001B001B001B001BULL };
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_28)   = { 0x001C001C001C001CULL, 0x001C001C001C001CULL };
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_32)   = { 0x0020002000200020ULL, 0x0020002000200020ULL };
+DECLARE_ALIGNED(8,  const uint64_t, ff_pw_42)   =   0x002A002A002A002AULL;
+DECLARE_ALIGNED(8,  const uint64_t, ff_pw_53)   =   0x0035003500350035ULL;
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_63)   = { 0x003F003F003F003FULL, 0x003F003F003F003FULL };
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_64)   = { 0x0040004000400040ULL, 0x0040004000400040ULL };
+DECLARE_ALIGNED(8,  const uint64_t, ff_pw_96)   =   0x0060006000600060ULL;
+DECLARE_ALIGNED(8,  const uint64_t, ff_pw_128)  =   0x0080008000800080ULL;
+DECLARE_ALIGNED(8,  const uint64_t, ff_pw_255)  =   0x00ff00ff00ff00ffULL;
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_512)  = { 0x0200020002000200ULL, 0x0200020002000200ULL };
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_1019) = { 0x03FB03FB03FB03FBULL, 0x03FB03FB03FB03FBULL };
+
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pb_0)    = { 0x0000000000000000ULL, 0x0000000000000000ULL };
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pb_1)    = { 0x0101010101010101ULL, 0x0101010101010101ULL };
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pb_3)    = { 0x0303030303030303ULL, 0x0303030303030303ULL };
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pb_4)    = { 0x0404040404040404ULL, 0x0404040404040404ULL };
+DECLARE_ALIGNED(8,  const uint64_t, ff_pb_7)    =   0x0707070707070707ULL;
+DECLARE_ALIGNED(8,  const uint64_t, ff_pb_1F)   =   0x1F1F1F1F1F1F1F1FULL;
+DECLARE_ALIGNED(8,  const uint64_t, ff_pb_3F)   =   0x3F3F3F3F3F3F3F3FULL;
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pb_80)   = { 0x8080808080808080ULL, 0x8080808080808080ULL };
+DECLARE_ALIGNED(8,  const uint64_t, ff_pb_81)   =   0x8181818181818181ULL;
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pb_A1)   = { 0xA1A1A1A1A1A1A1A1ULL, 0xA1A1A1A1A1A1A1A1ULL };
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pb_F8)   = { 0xF8F8F8F8F8F8F8F8ULL, 0xF8F8F8F8F8F8F8F8ULL };
+DECLARE_ALIGNED(8,  const uint64_t, ff_pb_FC)   =   0xFCFCFCFCFCFCFCFCULL;
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pb_FE)   = { 0xFEFEFEFEFEFEFEFEULL, 0xFEFEFEFEFEFEFEFEULL };
 
 DECLARE_ALIGNED(16, const double, ff_pd_1)[2] = { 1.0, 1.0 };
 DECLARE_ALIGNED(16, const double, ff_pd_2)[2] = { 2.0, 2.0 };
 
-#define JUMPALIGN() __asm__ volatile (".p2align 3"::)
-#define MOVQ_ZERO(regd)  __asm__ volatile ("pxor %%" #regd ", %%" #regd ::)
+#define JUMPALIGN()     __asm__ volatile (".p2align 3"::)
+#define MOVQ_ZERO(regd) __asm__ volatile ("pxor %%"#regd", %%"#regd ::)
 
-#define MOVQ_BFE(regd) \
-    __asm__ volatile ( \
-    "pcmpeqd %%" #regd ", %%" #regd " \n\t"\
-    "paddb %%" #regd ", %%" #regd " \n\t" ::)
+#define MOVQ_BFE(regd)                                  \
+    __asm__ volatile (                                  \
+        "pcmpeqd %%"#regd", %%"#regd"   \n\t"           \
+        "paddb   %%"#regd", %%"#regd"   \n\t" ::)
 
 #ifndef PIC
-#define MOVQ_BONE(regd)  __asm__ volatile ("movq %0, %%" #regd " \n\t" ::"m"(ff_bone))
-#define MOVQ_WTWO(regd)  __asm__ volatile ("movq %0, %%" #regd " \n\t" ::"m"(ff_wtwo))
+#define MOVQ_BONE(regd) __asm__ volatile ("movq %0, %%"#regd" \n\t" :: "m"(ff_bone))
+#define MOVQ_WTWO(regd) __asm__ volatile ("movq %0, %%"#regd" \n\t" :: "m"(ff_wtwo))
 #else
 // for shared library it's better to use this way for accessing constants
 // pcmpeqd -> -1
-#define MOVQ_BONE(regd) \
-    __asm__ volatile ( \
-    "pcmpeqd %%" #regd ", %%" #regd " \n\t" \
-    "psrlw $15, %%" #regd " \n\t" \
-    "packuswb %%" #regd ", %%" #regd " \n\t" ::)
-
-#define MOVQ_WTWO(regd) \
-    __asm__ volatile ( \
-    "pcmpeqd %%" #regd ", %%" #regd " \n\t" \
-    "psrlw $15, %%" #regd " \n\t" \
-    "psllw $1, %%" #regd " \n\t"::)
+#define MOVQ_BONE(regd)                                 \
+    __asm__ volatile (                                  \
+        "pcmpeqd  %%"#regd", %%"#regd"  \n\t"           \
+        "psrlw          $15, %%"#regd"  \n\t"           \
+        "packuswb %%"#regd", %%"#regd"  \n\t" ::)
+
+#define MOVQ_WTWO(regd)                                 \
+    __asm__ volatile (                                  \
+        "pcmpeqd %%"#regd", %%"#regd"   \n\t"           \
+        "psrlw         $15, %%"#regd"   \n\t"           \
+        "psllw          $1, %%"#regd"   \n\t"::)
 
 #endif
 
 // using regr as temporary and for the output result
 // first argument is unmodifed and second is trashed
 // regfe is supposed to contain 0xfefefefefefefefe
-#define PAVGB_MMX_NO_RND(rega, regb, regr, regfe) \
-    "movq " #rega ", " #regr "  \n\t"\
-    "pand " #regb ", " #regr "  \n\t"\
-    "pxor " #rega ", " #regb "  \n\t"\
-    "pand " #regfe "," #regb "  \n\t"\
-    "psrlq $1, " #regb "        \n\t"\
-    "paddb " #regb ", " #regr " \n\t"
-
-#define PAVGB_MMX(rega, regb, regr, regfe) \
-    "movq " #rega ", " #regr "  \n\t"\
-    "por  " #regb ", " #regr "  \n\t"\
-    "pxor " #rega ", " #regb "  \n\t"\
-    "pand " #regfe "," #regb "  \n\t"\
-    "psrlq $1, " #regb "        \n\t"\
-    "psubb " #regb ", " #regr " \n\t"
+#define PAVGB_MMX_NO_RND(rega, regb, regr, regfe)                \
+    "movq   "#rega", "#regr"            \n\t"                    \
+    "pand   "#regb", "#regr"            \n\t"                    \
+    "pxor   "#rega", "#regb"            \n\t"                    \
+    "pand  "#regfe", "#regb"            \n\t"                    \
+    "psrlq       $1, "#regb"            \n\t"                    \
+    "paddb  "#regb", "#regr"            \n\t"
+
+#define PAVGB_MMX(rega, regb, regr, regfe)                       \
+    "movq   "#rega", "#regr"            \n\t"                    \
+    "por    "#regb", "#regr"            \n\t"                    \
+    "pxor   "#rega", "#regb"            \n\t"                    \
+    "pand  "#regfe", "#regb"            \n\t"                    \
+    "psrlq       $1, "#regb"            \n\t"                    \
+    "psubb  "#regb", "#regr"            \n\t"
 
 // mm6 is supposed to contain 0xfefefefefefefefe
-#define PAVGBP_MMX_NO_RND(rega, regb, regr,  regc, regd, regp) \
-    "movq " #rega ", " #regr "  \n\t"\
-    "movq " #regc ", " #regp "  \n\t"\
-    "pand " #regb ", " #regr "  \n\t"\
-    "pand " #regd ", " #regp "  \n\t"\
-    "pxor " #rega ", " #regb "  \n\t"\
-    "pxor " #regc ", " #regd "  \n\t"\
-    "pand %%mm6, " #regb "      \n\t"\
-    "pand %%mm6, " #regd "      \n\t"\
-    "psrlq $1, " #regb "        \n\t"\
-    "psrlq $1, " #regd "        \n\t"\
-    "paddb " #regb ", " #regr " \n\t"\
-    "paddb " #regd ", " #regp " \n\t"
-
-#define PAVGBP_MMX(rega, regb, regr, regc, regd, regp) \
-    "movq " #rega ", " #regr "  \n\t"\
-    "movq " #regc ", " #regp "  \n\t"\
-    "por  " #regb ", " #regr "  \n\t"\
-    "por  " #regd ", " #regp "  \n\t"\
-    "pxor " #rega ", " #regb "  \n\t"\
-    "pxor " #regc ", " #regd "  \n\t"\
-    "pand %%mm6, " #regb "      \n\t"\
-    "pand %%mm6, " #regd "      \n\t"\
-    "psrlq $1, " #regd "        \n\t"\
-    "psrlq $1, " #regb "        \n\t"\
-    "psubb " #regb ", " #regr " \n\t"\
-    "psubb " #regd ", " #regp " \n\t"
+#define PAVGBP_MMX_NO_RND(rega, regb, regr,  regc, regd, regp)   \
+    "movq  "#rega", "#regr"             \n\t"                    \
+    "movq  "#regc", "#regp"             \n\t"                    \
+    "pand  "#regb", "#regr"             \n\t"                    \
+    "pand  "#regd", "#regp"             \n\t"                    \
+    "pxor  "#rega", "#regb"             \n\t"                    \
+    "pxor  "#regc", "#regd"             \n\t"                    \
+    "pand    %%mm6, "#regb"             \n\t"                    \
+    "pand    %%mm6, "#regd"             \n\t"                    \
+    "psrlq      $1, "#regb"             \n\t"                    \
+    "psrlq      $1, "#regd"             \n\t"                    \
+    "paddb "#regb", "#regr"             \n\t"                    \
+    "paddb "#regd", "#regp"             \n\t"
+
+#define PAVGBP_MMX(rega, regb, regr, regc, regd, regp)           \
+    "movq  "#rega", "#regr"             \n\t"                    \
+    "movq  "#regc", "#regp"             \n\t"                    \
+    "por   "#regb", "#regr"             \n\t"                    \
+    "por   "#regd", "#regp"             \n\t"                    \
+    "pxor  "#rega", "#regb"             \n\t"                    \
+    "pxor  "#regc", "#regd"             \n\t"                    \
+    "pand    %%mm6, "#regb"             \n\t"                    \
+    "pand    %%mm6, "#regd"             \n\t"                    \
+    "psrlq      $1, "#regd"             \n\t"                    \
+    "psrlq      $1, "#regb"             \n\t"                    \
+    "psubb "#regb", "#regr"             \n\t"                    \
+    "psubb "#regd", "#regp"             \n\t"
 
 /***********************************/
 /* MMX no rounding */
-#define DEF(x, y) x ## _no_rnd_ ## y ##_mmx
+#define DEF(x, y) x ## _no_rnd_ ## y ## _mmx
 #define SET_RND  MOVQ_WONE
 #define PAVGBP(a, b, c, d, e, f)        PAVGBP_MMX_NO_RND(a, b, c, d, e, f)
 #define PAVGB(a, b, c, e)               PAVGB_MMX_NO_RND(a, b, c, e)
@@ -178,7 +178,7 @@ DECLARE_ALIGNED(16, const double, ff_pd_2)[2] = { 2.0, 2.0 };
 /***********************************/
 /* MMX rounding */
 
-#define DEF(x, y) x ## _ ## y ##_mmx
+#define DEF(x, y) x ## _ ## y ## _mmx
 #define SET_RND  MOVQ_WTWO
 #define PAVGBP(a, b, c, d, e, f)        PAVGBP_MMX(a, b, c, d, e, f)
 #define PAVGB(a, b, c, e)               PAVGB_MMX(a, b, c, e)
@@ -235,537 +235,552 @@ DECLARE_ALIGNED(16, const double, ff_pd_2)[2] = { 2.0, 2.0 };
 /***********************************/
 /* standard MMX */
 
-void ff_put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size)
+void ff_put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels,
+                               int line_size)
 {
     const DCTELEM *p;
     uint8_t *pix;
 
     /* read the pixels */
-    p = block;
+    p   = block;
     pix = pixels;
     /* unrolled loop */
-        __asm__ volatile(
-                "movq   %3, %%mm0               \n\t"
-                "movq   8%3, %%mm1              \n\t"
-                "movq   16%3, %%mm2             \n\t"
-                "movq   24%3, %%mm3             \n\t"
-                "movq   32%3, %%mm4             \n\t"
-                "movq   40%3, %%mm5             \n\t"
-                "movq   48%3, %%mm6             \n\t"
-                "movq   56%3, %%mm7             \n\t"
-                "packuswb %%mm1, %%mm0          \n\t"
-                "packuswb %%mm3, %%mm2          \n\t"
-                "packuswb %%mm5, %%mm4          \n\t"
-                "packuswb %%mm7, %%mm6          \n\t"
-                "movq   %%mm0, (%0)             \n\t"
-                "movq   %%mm2, (%0, %1)         \n\t"
-                "movq   %%mm4, (%0, %1, 2)      \n\t"
-                "movq   %%mm6, (%0, %2)         \n\t"
-                ::"r" (pix), "r" ((x86_reg)line_size), "r" ((x86_reg)line_size*3), "m"(*p)
-                :"memory");
-        pix += line_size*4;
-        p += 32;
+    __asm__ volatile (
+        "movq        %3, %%mm0          \n\t"
+        "movq       8%3, %%mm1          \n\t"
+        "movq      16%3, %%mm2          \n\t"
+        "movq      24%3, %%mm3          \n\t"
+        "movq      32%3, %%mm4          \n\t"
+        "movq      40%3, %%mm5          \n\t"
+        "movq      48%3, %%mm6          \n\t"
+        "movq      56%3, %%mm7          \n\t"
+        "packuswb %%mm1, %%mm0          \n\t"
+        "packuswb %%mm3, %%mm2          \n\t"
+        "packuswb %%mm5, %%mm4          \n\t"
+        "packuswb %%mm7, %%mm6          \n\t"
+        "movq     %%mm0, (%0)           \n\t"
+        "movq     %%mm2, (%0, %1)       \n\t"
+        "movq     %%mm4, (%0, %1, 2)    \n\t"
+        "movq     %%mm6, (%0, %2)       \n\t"
+        :: "r"(pix), "r"((x86_reg)line_size), "r"((x86_reg)line_size * 3),
+           "m"(*p)
+        : "memory");
+    pix += line_size * 4;
+    p   += 32;
 
     // if here would be an exact copy of the code above
     // compiler would generate some very strange code
     // thus using "r"
-    __asm__ volatile(
-            "movq       (%3), %%mm0             \n\t"
-            "movq       8(%3), %%mm1            \n\t"
-            "movq       16(%3), %%mm2           \n\t"
-            "movq       24(%3), %%mm3           \n\t"
-            "movq       32(%3), %%mm4           \n\t"
-            "movq       40(%3), %%mm5           \n\t"
-            "movq       48(%3), %%mm6           \n\t"
-            "movq       56(%3), %%mm7           \n\t"
-            "packuswb %%mm1, %%mm0              \n\t"
-            "packuswb %%mm3, %%mm2              \n\t"
-            "packuswb %%mm5, %%mm4              \n\t"
-            "packuswb %%mm7, %%mm6              \n\t"
-            "movq       %%mm0, (%0)             \n\t"
-            "movq       %%mm2, (%0, %1)         \n\t"
-            "movq       %%mm4, (%0, %1, 2)      \n\t"
-            "movq       %%mm6, (%0, %2)         \n\t"
-            ::"r" (pix), "r" ((x86_reg)line_size), "r" ((x86_reg)line_size*3), "r"(p)
-            :"memory");
-}
-
-#define put_signed_pixels_clamped_mmx_half(off) \
-            "movq    "#off"(%2), %%mm1          \n\t"\
-            "movq 16+"#off"(%2), %%mm2          \n\t"\
-            "movq 32+"#off"(%2), %%mm3          \n\t"\
-            "movq 48+"#off"(%2), %%mm4          \n\t"\
-            "packsswb  8+"#off"(%2), %%mm1      \n\t"\
-            "packsswb 24+"#off"(%2), %%mm2      \n\t"\
-            "packsswb 40+"#off"(%2), %%mm3      \n\t"\
-            "packsswb 56+"#off"(%2), %%mm4      \n\t"\
-            "paddb %%mm0, %%mm1                 \n\t"\
-            "paddb %%mm0, %%mm2                 \n\t"\
-            "paddb %%mm0, %%mm3                 \n\t"\
-            "paddb %%mm0, %%mm4                 \n\t"\
-            "movq %%mm1, (%0)                   \n\t"\
-            "movq %%mm2, (%0, %3)               \n\t"\
-            "movq %%mm3, (%0, %3, 2)            \n\t"\
-            "movq %%mm4, (%0, %1)               \n\t"
-
-void ff_put_signed_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size)
+    __asm__ volatile (
+        "movq       (%3), %%mm0         \n\t"
+        "movq      8(%3), %%mm1         \n\t"
+        "movq     16(%3), %%mm2         \n\t"
+        "movq     24(%3), %%mm3         \n\t"
+        "movq     32(%3), %%mm4         \n\t"
+        "movq     40(%3), %%mm5         \n\t"
+        "movq     48(%3), %%mm6         \n\t"
+        "movq     56(%3), %%mm7         \n\t"
+        "packuswb  %%mm1, %%mm0         \n\t"
+        "packuswb  %%mm3, %%mm2         \n\t"
+        "packuswb  %%mm5, %%mm4         \n\t"
+        "packuswb  %%mm7, %%mm6         \n\t"
+        "movq      %%mm0, (%0)          \n\t"
+        "movq      %%mm2, (%0, %1)      \n\t"
+        "movq      %%mm4, (%0, %1, 2)   \n\t"
+        "movq      %%mm6, (%0, %2)      \n\t"
+        :: "r"(pix), "r"((x86_reg)line_size), "r"((x86_reg)line_size * 3), "r"(p)
+        : "memory");
+}
+
+#define put_signed_pixels_clamped_mmx_half(off)             \
+    "movq          "#off"(%2), %%mm1        \n\t"           \
+    "movq     16 + "#off"(%2), %%mm2        \n\t"           \
+    "movq     32 + "#off"(%2), %%mm3        \n\t"           \
+    "movq     48 + "#off"(%2), %%mm4        \n\t"           \
+    "packsswb  8 + "#off"(%2), %%mm1        \n\t"           \
+    "packsswb 24 + "#off"(%2), %%mm2        \n\t"           \
+    "packsswb 40 + "#off"(%2), %%mm3        \n\t"           \
+    "packsswb 56 + "#off"(%2), %%mm4        \n\t"           \
+    "paddb              %%mm0, %%mm1        \n\t"           \
+    "paddb              %%mm0, %%mm2        \n\t"           \
+    "paddb              %%mm0, %%mm3        \n\t"           \
+    "paddb              %%mm0, %%mm4        \n\t"           \
+    "movq               %%mm1, (%0)         \n\t"           \
+    "movq               %%mm2, (%0, %3)     \n\t"           \
+    "movq               %%mm3, (%0, %3, 2)  \n\t"           \
+    "movq               %%mm4, (%0, %1)     \n\t"
+
+void ff_put_signed_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels,
+                                      int line_size)
 {
     x86_reg line_skip = line_size;
     x86_reg line_skip3;
 
     __asm__ volatile (
-            "movq "MANGLE(ff_pb_80)", %%mm0 \n\t"
-            "lea (%3, %3, 2), %1                \n\t"
-            put_signed_pixels_clamped_mmx_half(0)
-            "lea (%0, %3, 4), %0                \n\t"
-            put_signed_pixels_clamped_mmx_half(64)
-            :"+&r" (pixels), "=&r" (line_skip3)
-            :"r" (block), "r"(line_skip)
-            :"memory");
-}
-
-void ff_add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size)
+        "movq "MANGLE(ff_pb_80)", %%mm0     \n\t"
+        "lea         (%3, %3, 2), %1        \n\t"
+        put_signed_pixels_clamped_mmx_half(0)
+        "lea         (%0, %3, 4), %0        \n\t"
+        put_signed_pixels_clamped_mmx_half(64)
+        : "+&r"(pixels), "=&r"(line_skip3)
+        : "r"(block), "r"(line_skip)
+        : "memory");
+}
+
+void ff_add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels,
+                               int line_size)
 {
     const DCTELEM *p;
     uint8_t *pix;
     int i;
 
     /* read the pixels */
-    p = block;
+    p   = block;
     pix = pixels;
     MOVQ_ZERO(mm7);
     i = 4;
     do {
-        __asm__ volatile(
-                "movq   (%2), %%mm0     \n\t"
-                "movq   8(%2), %%mm1    \n\t"
-                "movq   16(%2), %%mm2   \n\t"
-                "movq   24(%2), %%mm3   \n\t"
-                "movq   %0, %%mm4       \n\t"
-                "movq   %1, %%mm6       \n\t"
-                "movq   %%mm4, %%mm5    \n\t"
-                "punpcklbw %%mm7, %%mm4 \n\t"
-                "punpckhbw %%mm7, %%mm5 \n\t"
-                "paddsw %%mm4, %%mm0    \n\t"
-                "paddsw %%mm5, %%mm1    \n\t"
-                "movq   %%mm6, %%mm5    \n\t"
-                "punpcklbw %%mm7, %%mm6 \n\t"
-                "punpckhbw %%mm7, %%mm5 \n\t"
-                "paddsw %%mm6, %%mm2    \n\t"
-                "paddsw %%mm5, %%mm3    \n\t"
-                "packuswb %%mm1, %%mm0  \n\t"
-                "packuswb %%mm3, %%mm2  \n\t"
-                "movq   %%mm0, %0       \n\t"
-                "movq   %%mm2, %1       \n\t"
-                :"+m"(*pix), "+m"(*(pix+line_size))
-                :"r"(p)
-                :"memory");
-        pix += line_size*2;
-        p += 16;
+        __asm__ volatile (
+            "movq        (%2), %%mm0    \n\t"
+            "movq       8(%2), %%mm1    \n\t"
+            "movq      16(%2), %%mm2    \n\t"
+            "movq      24(%2), %%mm3    \n\t"
+            "movq          %0, %%mm4    \n\t"
+            "movq          %1, %%mm6    \n\t"
+            "movq       %%mm4, %%mm5    \n\t"
+            "punpcklbw  %%mm7, %%mm4    \n\t"
+            "punpckhbw  %%mm7, %%mm5    \n\t"
+            "paddsw     %%mm4, %%mm0    \n\t"
+            "paddsw     %%mm5, %%mm1    \n\t"
+            "movq       %%mm6, %%mm5    \n\t"
+            "punpcklbw  %%mm7, %%mm6    \n\t"
+            "punpckhbw  %%mm7, %%mm5    \n\t"
+            "paddsw     %%mm6, %%mm2    \n\t"
+            "paddsw     %%mm5, %%mm3    \n\t"
+            "packuswb   %%mm1, %%mm0    \n\t"
+            "packuswb   %%mm3, %%mm2    \n\t"
+            "movq       %%mm0, %0       \n\t"
+            "movq       %%mm2, %1       \n\t"
+            : "+m"(*pix), "+m"(*(pix + line_size))
+            "r"(p)
+            "memory");
+        pix += line_size * 2;
+        p   += 16;
     } while (--i);
 }
 
-static void put_pixels4_mmx(uint8_t *block, const uint8_t *pixels, int line_size, int h)
+static void put_pixels4_mmx(uint8_t *block, const uint8_t *pixels,
+                            int line_size, int h)
 {
-    __asm__ volatile(
-         "lea (%3, %3), %%"REG_a"       \n\t"
-         ".p2align 3                    \n\t"
-         "1:                            \n\t"
-         "movd (%1), %%mm0              \n\t"
-         "movd (%1, %3), %%mm1          \n\t"
-         "movd %%mm0, (%2)              \n\t"
-         "movd %%mm1, (%2, %3)          \n\t"
-         "add %%"REG_a", %1             \n\t"
-         "add %%"REG_a", %2             \n\t"
-         "movd (%1), %%mm0              \n\t"
-         "movd (%1, %3), %%mm1          \n\t"
-         "movd %%mm0, (%2)              \n\t"
-         "movd %%mm1, (%2, %3)          \n\t"
-         "add %%"REG_a", %1             \n\t"
-         "add %%"REG_a", %2             \n\t"
-         "subl $4, %0                   \n\t"
-         "jnz 1b                        \n\t"
-         : "+g"(h), "+r" (pixels),  "+r" (block)
-         : "r"((x86_reg)line_size)
-         : "%"REG_a, "memory"
+    __asm__ volatile (
+        "lea   (%3, %3), %%"REG_a"      \n\t"
+        ".p2align     3                 \n\t"
+        "1:                             \n\t"
+        "movd  (%1    ), %%mm0          \n\t"
+        "movd  (%1, %3), %%mm1          \n\t"
+        "movd     %%mm0, (%2)           \n\t"
+        "movd     %%mm1, (%2, %3)       \n\t"
+        "add  %%"REG_a", %1             \n\t"
+        "add  %%"REG_a", %2             \n\t"
+        "movd  (%1    ), %%mm0          \n\t"
+        "movd  (%1, %3), %%mm1          \n\t"
+        "movd     %%mm0, (%2)           \n\t"
+        "movd     %%mm1, (%2, %3)       \n\t"
+        "add  %%"REG_a", %1             \n\t"
+        "add  %%"REG_a", %2             \n\t"
+        "subl        $4, %0             \n\t"
+        "jnz         1b                 \n\t"
+        : "+g"(h), "+r"(pixels),  "+r"(block)
+        : "r"((x86_reg)line_size)
+        : "%"REG_a, "memory"
         );
 }
 
-static void put_pixels8_mmx(uint8_t *block, const uint8_t *pixels, int line_size, int h)
+static void put_pixels8_mmx(uint8_t *block, const uint8_t *pixels,
+                            int line_size, int h)
 {
-    __asm__ volatile(
-         "lea (%3, %3), %%"REG_a"       \n\t"
-         ".p2align 3                    \n\t"
-         "1:                            \n\t"
-         "movq (%1), %%mm0              \n\t"
-         "movq (%1, %3), %%mm1          \n\t"
-         "movq %%mm0, (%2)              \n\t"
-         "movq %%mm1, (%2, %3)          \n\t"
-         "add %%"REG_a", %1             \n\t"
-         "add %%"REG_a", %2             \n\t"
-         "movq (%1), %%mm0              \n\t"
-         "movq (%1, %3), %%mm1          \n\t"
-         "movq %%mm0, (%2)              \n\t"
-         "movq %%mm1, (%2, %3)          \n\t"
-         "add %%"REG_a", %1             \n\t"
-         "add %%"REG_a", %2             \n\t"
-         "subl $4, %0                   \n\t"
-         "jnz 1b                        \n\t"
-         : "+g"(h), "+r" (pixels),  "+r" (block)
-         : "r"((x86_reg)line_size)
-         : "%"REG_a, "memory"
+    __asm__ volatile (
+        "lea   (%3, %3), %%"REG_a"      \n\t"
+        ".p2align     3                 \n\t"
+        "1:                             \n\t"
+        "movq  (%1    ), %%mm0          \n\t"
+        "movq  (%1, %3), %%mm1          \n\t"
+        "movq     %%mm0, (%2)           \n\t"
+        "movq     %%mm1, (%2, %3)       \n\t"
+        "add  %%"REG_a", %1             \n\t"
+        "add  %%"REG_a", %2             \n\t"
+        "movq  (%1    ), %%mm0          \n\t"
+        "movq  (%1, %3), %%mm1          \n\t"
+        "movq     %%mm0, (%2)           \n\t"
+        "movq     %%mm1, (%2, %3)       \n\t"
+        "add  %%"REG_a", %1             \n\t"
+        "add  %%"REG_a", %2             \n\t"
+        "subl        $4, %0             \n\t"
+        "jnz         1b                 \n\t"
+        : "+g"(h), "+r"(pixels),  "+r"(block)
+        : "r"((x86_reg)line_size)
+        : "%"REG_a, "memory"
         );
 }
 
-static void put_pixels16_mmx(uint8_t *block, const uint8_t *pixels, int line_size, int h)
+static void put_pixels16_mmx(uint8_t *block, const uint8_t *pixels,
+                             int line_size, int h)
 {
-    __asm__ volatile(
-         "lea (%3, %3), %%"REG_a"       \n\t"
-         ".p2align 3                    \n\t"
-         "1:                            \n\t"
-         "movq (%1), %%mm0              \n\t"
-         "movq 8(%1), %%mm4             \n\t"
-         "movq (%1, %3), %%mm1          \n\t"
-         "movq 8(%1, %3), %%mm5         \n\t"
-         "movq %%mm0, (%2)              \n\t"
-         "movq %%mm4, 8(%2)             \n\t"
-         "movq %%mm1, (%2, %3)          \n\t"
-         "movq %%mm5, 8(%2, %3)         \n\t"
-         "add %%"REG_a", %1             \n\t"
-         "add %%"REG_a", %2             \n\t"
-         "movq (%1), %%mm0              \n\t"
-         "movq 8(%1), %%mm4             \n\t"
-         "movq (%1, %3), %%mm1          \n\t"
-         "movq 8(%1, %3), %%mm5         \n\t"
-         "movq %%mm0, (%2)              \n\t"
-         "movq %%mm4, 8(%2)             \n\t"
-         "movq %%mm1, (%2, %3)          \n\t"
-         "movq %%mm5, 8(%2, %3)         \n\t"
-         "add %%"REG_a", %1             \n\t"
-         "add %%"REG_a", %2             \n\t"
-         "subl $4, %0                   \n\t"
-         "jnz 1b                        \n\t"
-         : "+g"(h), "+r" (pixels),  "+r" (block)
-         : "r"((x86_reg)line_size)
-         : "%"REG_a, "memory"
+    __asm__ volatile (
+        "lea   (%3, %3), %%"REG_a"      \n\t"
+        ".p2align     3                 \n\t"
+        "1:                             \n\t"
+        "movq  (%1    ), %%mm0          \n\t"
+        "movq 8(%1    ), %%mm4          \n\t"
+        "movq  (%1, %3), %%mm1          \n\t"
+        "movq 8(%1, %3), %%mm5          \n\t"
+        "movq     %%mm0,  (%2)          \n\t"
+        "movq     %%mm4, 8(%2)          \n\t"
+        "movq     %%mm1,  (%2, %3)      \n\t"
+        "movq     %%mm5, 8(%2, %3)      \n\t"
+        "add  %%"REG_a", %1             \n\t"
+        "add  %%"REG_a", %2             \n\t"
+        "movq  (%1    ), %%mm0          \n\t"
+        "movq 8(%1    ), %%mm4          \n\t"
+        "movq  (%1, %3), %%mm1          \n\t"
+        "movq 8(%1, %3), %%mm5          \n\t"
+        "movq     %%mm0,  (%2)          \n\t"
+        "movq     %%mm4, 8(%2)          \n\t"
+        "movq     %%mm1,  (%2, %3)      \n\t"
+        "movq     %%mm5, 8(%2, %3)      \n\t"
+        "add  %%"REG_a", %1             \n\t"
+        "add  %%"REG_a", %2             \n\t"
+        "subl        $4, %0             \n\t"
+        "jnz         1b                 \n\t"
+        : "+g"(h), "+r"(pixels),  "+r"(block)
+        : "r"((x86_reg)line_size)
+        : "%"REG_a, "memory"
         );
 }
 
-static void put_pixels16_sse2(uint8_t *block, const uint8_t *pixels, int line_size, int h)
+static void put_pixels16_sse2(uint8_t *block, const uint8_t *pixels,
+                              int line_size, int h)
 {
-    __asm__ volatile(
-         "1:                            \n\t"
-         "movdqu (%1), %%xmm0           \n\t"
-         "movdqu (%1,%3), %%xmm1        \n\t"
-         "movdqu (%1,%3,2), %%xmm2      \n\t"
-         "movdqu (%1,%4), %%xmm3        \n\t"
-         "lea (%1,%3,4), %1             \n\t"
-         "movdqa %%xmm0, (%2)           \n\t"
-         "movdqa %%xmm1, (%2,%3)        \n\t"
-         "movdqa %%xmm2, (%2,%3,2)      \n\t"
-         "movdqa %%xmm3, (%2,%4)        \n\t"
-         "subl $4, %0                   \n\t"
-         "lea (%2,%3,4), %2             \n\t"
-         "jnz 1b                        \n\t"
-         : "+g"(h), "+r" (pixels),  "+r" (block)
-         : "r"((x86_reg)line_size), "r"((x86_reg)3L*line_size)
-         : "memory"
+    __asm__ volatile (
+        "1:                              \n\t"
+        "movdqu (%1       ), %%xmm0      \n\t"
+        "movdqu (%1, %3   ), %%xmm1      \n\t"
+        "movdqu (%1, %3, 2), %%xmm2      \n\t"
+        "movdqu (%1, %4   ), %%xmm3      \n\t"
+        "lea    (%1, %3, 4), %1          \n\t"
+        "movdqa      %%xmm0, (%2)        \n\t"
+        "movdqa      %%xmm1, (%2, %3)    \n\t"
+        "movdqa      %%xmm2, (%2, %3, 2) \n\t"
+        "movdqa      %%xmm3, (%2, %4)    \n\t"
+        "subl            $4, %0          \n\t"
+        "lea    (%2, %3, 4), %2          \n\t"
+        "jnz             1b              \n\t"
+        : "+g"(h), "+r"(pixels),  "+r"(block)
+        : "r"((x86_reg)line_size), "r"((x86_reg)3L * line_size)
+        : "memory"
         );
 }
 
-static void avg_pixels16_sse2(uint8_t *block, const uint8_t *pixels, int line_size, int h)
+static void avg_pixels16_sse2(uint8_t *block, const uint8_t *pixels,
+                              int line_size, int h)
 {
-    __asm__ volatile(
-         "1:                            \n\t"
-         "movdqu (%1), %%xmm0           \n\t"
-         "movdqu (%1,%3), %%xmm1        \n\t"
-         "movdqu (%1,%3,2), %%xmm2      \n\t"
-         "movdqu (%1,%4), %%xmm3        \n\t"
-         "lea (%1,%3,4), %1             \n\t"
-         "pavgb  (%2), %%xmm0           \n\t"
-         "pavgb  (%2,%3), %%xmm1        \n\t"
-         "pavgb  (%2,%3,2), %%xmm2      \n\t"
-         "pavgb  (%2,%4), %%xmm3        \n\t"
-         "movdqa %%xmm0, (%2)           \n\t"
-         "movdqa %%xmm1, (%2,%3)        \n\t"
-         "movdqa %%xmm2, (%2,%3,2)      \n\t"
-         "movdqa %%xmm3, (%2,%4)        \n\t"
-         "subl $4, %0                   \n\t"
-         "lea (%2,%3,4), %2             \n\t"
-         "jnz 1b                        \n\t"
-         : "+g"(h), "+r" (pixels),  "+r" (block)
-         : "r"((x86_reg)line_size), "r"((x86_reg)3L*line_size)
-         : "memory"
+    __asm__ volatile (
+        "1:                                 \n\t"
+        "movdqu (%1       ), %%xmm0         \n\t"
+        "movdqu (%1, %3   ), %%xmm1         \n\t"
+        "movdqu (%1, %3, 2), %%xmm2         \n\t"
+        "movdqu (%1, %4   ), %%xmm3         \n\t"
+        "lea    (%1, %3, 4), %1             \n\t"
+        "pavgb  (%2       ), %%xmm0         \n\t"
+        "pavgb  (%2, %3   ), %%xmm1         \n\t"
+        "pavgb  (%2, %3, 2), %%xmm2         \n\t"
+        "pavgb     (%2, %4), %%xmm3         \n\t"
+        "movdqa      %%xmm0, (%2)           \n\t"
+        "movdqa      %%xmm1, (%2, %3)       \n\t"
+        "movdqa      %%xmm2, (%2, %3, 2)    \n\t"
+        "movdqa      %%xmm3, (%2, %4)       \n\t"
+        "subl            $4, %0             \n\t"
+        "lea    (%2, %3, 4), %2             \n\t"
+        "jnz             1b                 \n\t"
+        : "+g"(h), "+r"(pixels),  "+r"(block)
+        : "r"((x86_reg)line_size), "r"((x86_reg)3L * line_size)
+        : "memory"
         );
 }
 
-#define CLEAR_BLOCKS(name,n) \
-static void name(DCTELEM *blocks)\
-{\
-    __asm__ volatile(\
-                "pxor %%mm7, %%mm7              \n\t"\
-                "mov     %1, %%"REG_a"          \n\t"\
-                "1:                             \n\t"\
-                "movq %%mm7, (%0, %%"REG_a")    \n\t"\
-                "movq %%mm7, 8(%0, %%"REG_a")   \n\t"\
-                "movq %%mm7, 16(%0, %%"REG_a")  \n\t"\
-                "movq %%mm7, 24(%0, %%"REG_a")  \n\t"\
-                "add $32, %%"REG_a"             \n\t"\
-                " js 1b                         \n\t"\
-                : : "r" (((uint8_t *)blocks)+128*n),\
-                    "i" (-128*n)\
-                : "%"REG_a\
-        );\
+#define CLEAR_BLOCKS(name, n)                           \
+static void name(DCTELEM *blocks)                       \
+{                                                       \
+    __asm__ volatile (                                  \
+        "pxor %%mm7, %%mm7              \n\t"           \
+        "mov     %1,        %%"REG_a"   \n\t"           \
+        "1:                             \n\t"           \
+        "movq %%mm7,   (%0, %%"REG_a")  \n\t"           \
+        "movq %%mm7,  8(%0, %%"REG_a")  \n\t"           \
+        "movq %%mm7, 16(%0, %%"REG_a")  \n\t"           \
+        "movq %%mm7, 24(%0, %%"REG_a")  \n\t"           \
+        "add    $32, %%"REG_a"          \n\t"           \
+        "js      1b                     \n\t"           \
+        :: "r"(((uint8_t *)blocks) + 128 * n),          \
+           "i"(-128 * n)                                \
+        : "%"REG_a                                      \
+        );                                              \
 }
 CLEAR_BLOCKS(clear_blocks_mmx, 6)
 CLEAR_BLOCKS(clear_block_mmx, 1)
 
 static void clear_block_sse(DCTELEM *block)
 {
-    __asm__ volatile(
-        "xorps  %%xmm0, %%xmm0  \n"
-        "movaps %%xmm0,    (%0) \n"
-        "movaps %%xmm0,  16(%0) \n"
-        "movaps %%xmm0,  32(%0) \n"
-        "movaps %%xmm0,  48(%0) \n"
-        "movaps %%xmm0,  64(%0) \n"
-        "movaps %%xmm0,  80(%0) \n"
-        "movaps %%xmm0,  96(%0) \n"
-        "movaps %%xmm0, 112(%0) \n"
+    __asm__ volatile (
+        "xorps  %%xmm0, %%xmm0          \n"
+        "movaps %%xmm0,    (%0)         \n"
+        "movaps %%xmm0,  16(%0)         \n"
+        "movaps %%xmm0,  32(%0)         \n"
+        "movaps %%xmm0,  48(%0)         \n"
+        "movaps %%xmm0,  64(%0)         \n"
+        "movaps %%xmm0,  80(%0)         \n"
+        "movaps %%xmm0,  96(%0)         \n"
+        "movaps %%xmm0, 112(%0)         \n"
         :: "r"(block)
         : "memory"
     );
 }
 
 static void clear_blocks_sse(DCTELEM *blocks)
-{\
-    __asm__ volatile(
-        "xorps  %%xmm0, %%xmm0  \n"
-        "mov     %1, %%"REG_a"  \n"
-        "1:                     \n"
-        "movaps %%xmm0,    (%0, %%"REG_a") \n"
-        "movaps %%xmm0,  16(%0, %%"REG_a") \n"
-        "movaps %%xmm0,  32(%0, %%"REG_a") \n"
-        "movaps %%xmm0,  48(%0, %%"REG_a") \n"
-        "movaps %%xmm0,  64(%0, %%"REG_a") \n"
-        "movaps %%xmm0,  80(%0, %%"REG_a") \n"
-        "movaps %%xmm0,  96(%0, %%"REG_a") \n"
-        "movaps %%xmm0, 112(%0, %%"REG_a") \n"
-        "add $128, %%"REG_a"    \n"
-        " js 1b                 \n"
-        : : "r" (((uint8_t *)blocks)+128*6),
-            "i" (-128*6)
+{
+    __asm__ volatile (
+        "xorps  %%xmm0, %%xmm0              \n"
+        "mov        %1,         %%"REG_a"   \n"
+        "1:                                 \n"
+        "movaps %%xmm0,    (%0, %%"REG_a")  \n"
+        "movaps %%xmm0,  16(%0, %%"REG_a")  \n"
+        "movaps %%xmm0,  32(%0, %%"REG_a")  \n"
+        "movaps %%xmm0,  48(%0, %%"REG_a")  \n"
+        "movaps %%xmm0,  64(%0, %%"REG_a")  \n"
+        "movaps %%xmm0,  80(%0, %%"REG_a")  \n"
+        "movaps %%xmm0,  96(%0, %%"REG_a")  \n"
+        "movaps %%xmm0, 112(%0, %%"REG_a")  \n"
+        "add      $128,         %%"REG_a"   \n"
+        "js         1b                      \n"
+        :: "r"(((uint8_t *)blocks) + 128 * 6),
+           "i"(-128 * 6)
         : "%"REG_a
     );
 }
 
-static void add_bytes_mmx(uint8_t *dst, uint8_t *src, int w){
-    x86_reg i=0;
-    __asm__ volatile(
-        "jmp 2f                         \n\t"
+static void add_bytes_mmx(uint8_t *dst, uint8_t *src, int w)
+{
+    x86_reg i = 0;
+    __asm__ volatile (
+        "jmp          2f                \n\t"
         "1:                             \n\t"
-        "movq  (%1, %0), %%mm0          \n\t"
-        "movq  (%2, %0), %%mm1          \n\t"
-        "paddb %%mm0, %%mm1             \n\t"
-        "movq %%mm1, (%2, %0)           \n\t"
-        "movq 8(%1, %0), %%mm0          \n\t"
-        "movq 8(%2, %0), %%mm1          \n\t"
-        "paddb %%mm0, %%mm1             \n\t"
-        "movq %%mm1, 8(%2, %0)          \n\t"
-        "add $16, %0                    \n\t"
+        "movq   (%1, %0), %%mm0         \n\t"
+        "movq   (%2, %0), %%mm1         \n\t"
+        "paddb     %%mm0, %%mm1         \n\t"
+        "movq      %%mm1, (%2, %0)      \n\t"
+        "movq  8(%1, %0), %%mm0         \n\t"
+        "movq  8(%2, %0), %%mm1         \n\t"
+        "paddb     %%mm0, %%mm1         \n\t"
+        "movq      %%mm1, 8(%2, %0)     \n\t"
+        "add         $16, %0            \n\t"
         "2:                             \n\t"
-        "cmp %3, %0                     \n\t"
-        " js 1b                         \n\t"
-        : "+r" (i)
-        : "r"(src), "r"(dst), "r"((x86_reg)w-15)
+        "cmp          %3, %0            \n\t"
+        "js           1b                \n\t"
+        : "+r"(i)
+        : "r"(src), "r"(dst), "r"((x86_reg)w - 15)
     );
-    for(; i<w; i++)
-        dst[i+0] += src[i+0];
+    for ( ; i < w; i++)
+        dst[i + 0] += src[i + 0];
 }
 
 #if HAVE_7REGS
-static void add_hfyu_median_prediction_cmov(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top) {
+static void add_hfyu_median_prediction_cmov(uint8_t *dst, const uint8_t *top,
+                                            const uint8_t *diff, int w,
+                                            int *left, int *left_top)
+{
     x86_reg w2 = -w;
     x86_reg x;
-    int l = *left & 0xff;
+    int l  = *left     & 0xff;
     int tl = *left_top & 0xff;
     int t;
-    __asm__ volatile(
-        "mov    %7, %3 \n"
-        "1: \n"
-        "movzbl (%3,%4), %2 \n"
-        "mov    %2, %k3 \n"
-        "sub   %b1, %b3 \n"
-        "add   %b0, %b3 \n"
-        "mov    %2, %1 \n"
-        "cmp    %0, %2 \n"
-        "cmovg  %0, %2 \n"
-        "cmovg  %1, %0 \n"
-        "cmp   %k3, %0 \n"
-        "cmovg %k3, %0 \n"
-        "mov    %7, %3 \n"
-        "cmp    %2, %0 \n"
-        "cmovl  %2, %0 \n"
-        "add (%6,%4), %b0 \n"
-        "mov   %b0, (%5,%4) \n"
-        "inc    %4 \n"
-        "jl 1b \n"
-        :"+&q"(l), "+&q"(tl), "=&r"(t), "=&q"(x), "+&r"(w2)
-        :"r"(dst+w), "r"(diff+w), "rm"(top+w)
+    __asm__ volatile (
+        "mov          %7, %3            \n"
+        "1:                             \n"
+        "movzbl (%3, %4), %2            \n"
+        "mov          %2, %k3           \n"
+        "sub         %b1, %b3           \n"
+        "add         %b0, %b3           \n"
+        "mov          %2, %1            \n"
+        "cmp          %0, %2            \n"
+        "cmovg        %0, %2            \n"
+        "cmovg        %1, %0            \n"
+        "cmp         %k3, %0            \n"
+        "cmovg       %k3, %0            \n"
+        "mov          %7, %3            \n"
+        "cmp          %2, %0            \n"
+        "cmovl        %2, %0            \n"
+        "add    (%6, %4), %b0           \n"
+        "mov         %b0, (%5, %4)      \n"
+        "inc          %4                \n"
+        "jl           1b                \n"
+        : "+&q"(l), "+&q"(tl), "=&r"(t), "=&q"(x), "+&r"(w2)
+        : "r"(dst + w), "r"(diff + w), "rm"(top + w)
     );
-    *left = l;
+    *left     = l;
     *left_top = tl;
 }
 #endif
 
-#define H263_LOOP_FILTER \
-        "pxor %%mm7, %%mm7              \n\t"\
-        "movq  %0, %%mm0                \n\t"\
-        "movq  %0, %%mm1                \n\t"\
-        "movq  %3, %%mm2                \n\t"\
-        "movq  %3, %%mm3                \n\t"\
-        "punpcklbw %%mm7, %%mm0         \n\t"\
-        "punpckhbw %%mm7, %%mm1         \n\t"\
-        "punpcklbw %%mm7, %%mm2         \n\t"\
-        "punpckhbw %%mm7, %%mm3         \n\t"\
-        "psubw %%mm2, %%mm0             \n\t"\
-        "psubw %%mm3, %%mm1             \n\t"\
-        "movq  %1, %%mm2                \n\t"\
-        "movq  %1, %%mm3                \n\t"\
-        "movq  %2, %%mm4                \n\t"\
-        "movq  %2, %%mm5                \n\t"\
-        "punpcklbw %%mm7, %%mm2         \n\t"\
-        "punpckhbw %%mm7, %%mm3         \n\t"\
-        "punpcklbw %%mm7, %%mm4         \n\t"\
-        "punpckhbw %%mm7, %%mm5         \n\t"\
-        "psubw %%mm2, %%mm4             \n\t"\
-        "psubw %%mm3, %%mm5             \n\t"\
-        "psllw $2, %%mm4                \n\t"\
-        "psllw $2, %%mm5                \n\t"\
-        "paddw %%mm0, %%mm4             \n\t"\
-        "paddw %%mm1, %%mm5             \n\t"\
-        "pxor %%mm6, %%mm6              \n\t"\
-        "pcmpgtw %%mm4, %%mm6           \n\t"\
-        "pcmpgtw %%mm5, %%mm7           \n\t"\
-        "pxor %%mm6, %%mm4              \n\t"\
-        "pxor %%mm7, %%mm5              \n\t"\
-        "psubw %%mm6, %%mm4             \n\t"\
-        "psubw %%mm7, %%mm5             \n\t"\
-        "psrlw $3, %%mm4                \n\t"\
-        "psrlw $3, %%mm5                \n\t"\
-        "packuswb %%mm5, %%mm4          \n\t"\
-        "packsswb %%mm7, %%mm6          \n\t"\
-        "pxor %%mm7, %%mm7              \n\t"\
-        "movd %4, %%mm2                 \n\t"\
-        "punpcklbw %%mm2, %%mm2         \n\t"\
-        "punpcklbw %%mm2, %%mm2         \n\t"\
-        "punpcklbw %%mm2, %%mm2         \n\t"\
-        "psubusb %%mm4, %%mm2           \n\t"\
-        "movq %%mm2, %%mm3              \n\t"\
-        "psubusb %%mm4, %%mm3           \n\t"\
-        "psubb %%mm3, %%mm2             \n\t"\
-        "movq %1, %%mm3                 \n\t"\
-        "movq %2, %%mm4                 \n\t"\
-        "pxor %%mm6, %%mm3              \n\t"\
-        "pxor %%mm6, %%mm4              \n\t"\
-        "paddusb %%mm2, %%mm3           \n\t"\
-        "psubusb %%mm2, %%mm4           \n\t"\
-        "pxor %%mm6, %%mm3              \n\t"\
-        "pxor %%mm6, %%mm4              \n\t"\
-        "paddusb %%mm2, %%mm2           \n\t"\
-        "packsswb %%mm1, %%mm0          \n\t"\
-        "pcmpgtb %%mm0, %%mm7           \n\t"\
-        "pxor %%mm7, %%mm0              \n\t"\
-        "psubb %%mm7, %%mm0             \n\t"\
-        "movq %%mm0, %%mm1              \n\t"\
-        "psubusb %%mm2, %%mm0           \n\t"\
-        "psubb %%mm0, %%mm1             \n\t"\
-        "pand %5, %%mm1                 \n\t"\
-        "psrlw $2, %%mm1                \n\t"\
-        "pxor %%mm7, %%mm1              \n\t"\
-        "psubb %%mm7, %%mm1             \n\t"\
-        "movq %0, %%mm5                 \n\t"\
-        "movq %3, %%mm6                 \n\t"\
-        "psubb %%mm1, %%mm5             \n\t"\
-        "paddb %%mm1, %%mm6             \n\t"
-
-static void h263_v_loop_filter_mmx(uint8_t *src, int stride, int qscale){
-    if(CONFIG_H263_DECODER || CONFIG_H263_ENCODER) {
-    const int strength= ff_h263_loop_filter_strength[qscale];
-
-    __asm__ volatile(
-
-        H263_LOOP_FILTER
-
-        "movq %%mm3, %1                 \n\t"
-        "movq %%mm4, %2                 \n\t"
-        "movq %%mm5, %0                 \n\t"
-        "movq %%mm6, %3                 \n\t"
-        : "+m" (*(uint64_t*)(src - 2*stride)),
-          "+m" (*(uint64_t*)(src - 1*stride)),
-          "+m" (*(uint64_t*)(src + 0*stride)),
-          "+m" (*(uint64_t*)(src + 1*stride))
-        : "g" (2*strength), "m"(ff_pb_FC)
-    );
+#define H263_LOOP_FILTER                        \
+    "pxor      %%mm7, %%mm7             \n\t"   \
+    "movq         %0, %%mm0             \n\t"   \
+    "movq         %0, %%mm1             \n\t"   \
+    "movq         %3, %%mm2             \n\t"   \
+    "movq         %3, %%mm3             \n\t"   \
+    "punpcklbw %%mm7, %%mm0             \n\t"   \
+    "punpckhbw %%mm7, %%mm1             \n\t"   \
+    "punpcklbw %%mm7, %%mm2             \n\t"   \
+    "punpckhbw %%mm7, %%mm3             \n\t"   \
+    "psubw     %%mm2, %%mm0             \n\t"   \
+    "psubw     %%mm3, %%mm1             \n\t"   \
+    "movq         %1, %%mm2             \n\t"   \
+    "movq         %1, %%mm3             \n\t"   \
+    "movq         %2, %%mm4             \n\t"   \
+    "movq         %2, %%mm5             \n\t"   \
+    "punpcklbw %%mm7, %%mm2             \n\t"   \
+    "punpckhbw %%mm7, %%mm3             \n\t"   \
+    "punpcklbw %%mm7, %%mm4             \n\t"   \
+    "punpckhbw %%mm7, %%mm5             \n\t"   \
+    "psubw     %%mm2, %%mm4             \n\t"   \
+    "psubw     %%mm3, %%mm5             \n\t"   \
+    "psllw        $2, %%mm4             \n\t"   \
+    "psllw        $2, %%mm5             \n\t"   \
+    "paddw     %%mm0, %%mm4             \n\t"   \
+    "paddw     %%mm1, %%mm5             \n\t"   \
+    "pxor      %%mm6, %%mm6             \n\t"   \
+    "pcmpgtw   %%mm4, %%mm6             \n\t"   \
+    "pcmpgtw   %%mm5, %%mm7             \n\t"   \
+    "pxor      %%mm6, %%mm4             \n\t"   \
+    "pxor      %%mm7, %%mm5             \n\t"   \
+    "psubw     %%mm6, %%mm4             \n\t"   \
+    "psubw     %%mm7, %%mm5             \n\t"   \
+    "psrlw        $3, %%mm4             \n\t"   \
+    "psrlw        $3, %%mm5             \n\t"   \
+    "packuswb  %%mm5, %%mm4             \n\t"   \
+    "packsswb  %%mm7, %%mm6             \n\t"   \
+    "pxor      %%mm7, %%mm7             \n\t"   \
+    "movd         %4, %%mm2             \n\t"   \
+    "punpcklbw %%mm2, %%mm2             \n\t"   \
+    "punpcklbw %%mm2, %%mm2             \n\t"   \
+    "punpcklbw %%mm2, %%mm2             \n\t"   \
+    "psubusb   %%mm4, %%mm2             \n\t"   \
+    "movq      %%mm2, %%mm3             \n\t"   \
+    "psubusb   %%mm4, %%mm3             \n\t"   \
+    "psubb     %%mm3, %%mm2             \n\t"   \
+    "movq         %1, %%mm3             \n\t"   \
+    "movq         %2, %%mm4             \n\t"   \
+    "pxor      %%mm6, %%mm3             \n\t"   \
+    "pxor      %%mm6, %%mm4             \n\t"   \
+    "paddusb   %%mm2, %%mm3             \n\t"   \
+    "psubusb   %%mm2, %%mm4             \n\t"   \
+    "pxor      %%mm6, %%mm3             \n\t"   \
+    "pxor      %%mm6, %%mm4             \n\t"   \
+    "paddusb   %%mm2, %%mm2             \n\t"   \
+    "packsswb  %%mm1, %%mm0             \n\t"   \
+    "pcmpgtb   %%mm0, %%mm7             \n\t"   \
+    "pxor      %%mm7, %%mm0             \n\t"   \
+    "psubb     %%mm7, %%mm0             \n\t"   \
+    "movq      %%mm0, %%mm1             \n\t"   \
+    "psubusb   %%mm2, %%mm0             \n\t"   \
+    "psubb     %%mm0, %%mm1             \n\t"   \
+    "pand         %5, %%mm1             \n\t"   \
+    "psrlw        $2, %%mm1             \n\t"   \
+    "pxor      %%mm7, %%mm1             \n\t"   \
+    "psubb     %%mm7, %%mm1             \n\t"   \
+    "movq         %0, %%mm5             \n\t"   \
+    "movq         %3, %%mm6             \n\t"   \
+    "psubb     %%mm1, %%mm5             \n\t"   \
+    "paddb     %%mm1, %%mm6             \n\t"
+
+static void h263_v_loop_filter_mmx(uint8_t *src, int stride, int qscale)
+{
+    if (CONFIG_H263_DECODER || CONFIG_H263_ENCODER) {
+        const int strength = ff_h263_loop_filter_strength[qscale];
+
+        __asm__ volatile (
+            H263_LOOP_FILTER
+
+            "movq %%mm3, %1             \n\t"
+            "movq %%mm4, %2             \n\t"
+            "movq %%mm5, %0             \n\t"
+            "movq %%mm6, %3             \n\t"
+            : "+m"(*(uint64_t*)(src - 2 * stride)),
+              "+m"(*(uint64_t*)(src - 1 * stride)),
+              "+m"(*(uint64_t*)(src + 0 * stride)),
+              "+m"(*(uint64_t*)(src + 1 * stride))
+            : "g"(2 * strength), "m"(ff_pb_FC)
+            );
     }
 }
 
-static void h263_h_loop_filter_mmx(uint8_t *src, int stride, int qscale){
-    if(CONFIG_H263_DECODER || CONFIG_H263_ENCODER) {
-    const int strength= ff_h263_loop_filter_strength[qscale];
-    DECLARE_ALIGNED(8, uint64_t, temp)[4];
-    uint8_t *btemp= (uint8_t*)temp;
-
-    src -= 2;
-
-    transpose4x4(btemp  , src           , 8, stride);
-    transpose4x4(btemp+4, src + 4*stride, 8, stride);
-    __asm__ volatile(
-        H263_LOOP_FILTER // 5 3 4 6
-
-        : "+m" (temp[0]),
-          "+m" (temp[1]),
-          "+m" (temp[2]),
-          "+m" (temp[3])
-        : "g" (2*strength), "m"(ff_pb_FC)
-    );
+static void h263_h_loop_filter_mmx(uint8_t *src, int stride, int qscale)
+{
+    if (CONFIG_H263_DECODER || CONFIG_H263_ENCODER) {
+        const int strength = ff_h263_loop_filter_strength[qscale];
+        DECLARE_ALIGNED(8, uint64_t, temp)[4];
+        uint8_t *btemp = (uint8_t*)temp;
+
+        src -= 2;
+
+        transpose4x4(btemp,     src,              8, stride);
+        transpose4x4(btemp + 4, src + 4 * stride, 8, stride);
+        __asm__ volatile (
+            H263_LOOP_FILTER // 5 3 4 6
+
+            : "+m"(temp[0]),
+              "+m"(temp[1]),
+              "+m"(temp[2]),
+              "+m"(temp[3])
+            : "g"(2 * strength), "m"(ff_pb_FC)
+            );
 
-    __asm__ volatile(
-        "movq %%mm5, %%mm1              \n\t"
-        "movq %%mm4, %%mm0              \n\t"
-        "punpcklbw %%mm3, %%mm5         \n\t"
-        "punpcklbw %%mm6, %%mm4         \n\t"
-        "punpckhbw %%mm3, %%mm1         \n\t"
-        "punpckhbw %%mm6, %%mm0         \n\t"
-        "movq %%mm5, %%mm3              \n\t"
-        "movq %%mm1, %%mm6              \n\t"
-        "punpcklwd %%mm4, %%mm5         \n\t"
-        "punpcklwd %%mm0, %%mm1         \n\t"
-        "punpckhwd %%mm4, %%mm3         \n\t"
-        "punpckhwd %%mm0, %%mm6         \n\t"
-        "movd %%mm5, (%0)               \n\t"
-        "punpckhdq %%mm5, %%mm5         \n\t"
-        "movd %%mm5, (%0,%2)            \n\t"
-        "movd %%mm3, (%0,%2,2)          \n\t"
-        "punpckhdq %%mm3, %%mm3         \n\t"
-        "movd %%mm3, (%0,%3)            \n\t"
-        "movd %%mm1, (%1)               \n\t"
-        "punpckhdq %%mm1, %%mm1         \n\t"
-        "movd %%mm1, (%1,%2)            \n\t"
-        "movd %%mm6, (%1,%2,2)          \n\t"
-        "punpckhdq %%mm6, %%mm6         \n\t"
-        "movd %%mm6, (%1,%3)            \n\t"
-        :: "r" (src),
-           "r" (src + 4*stride),
-           "r" ((x86_reg)   stride ),
-           "r" ((x86_reg)(3*stride))
-    );
+        __asm__ volatile (
+            "movq      %%mm5, %%mm1         \n\t"
+            "movq      %%mm4, %%mm0         \n\t"
+            "punpcklbw %%mm3, %%mm5         \n\t"
+            "punpcklbw %%mm6, %%mm4         \n\t"
+            "punpckhbw %%mm3, %%mm1         \n\t"
+            "punpckhbw %%mm6, %%mm0         \n\t"
+            "movq      %%mm5, %%mm3         \n\t"
+            "movq      %%mm1, %%mm6         \n\t"
+            "punpcklwd %%mm4, %%mm5         \n\t"
+            "punpcklwd %%mm0, %%mm1         \n\t"
+            "punpckhwd %%mm4, %%mm3         \n\t"
+            "punpckhwd %%mm0, %%mm6         \n\t"
+            "movd      %%mm5, (%0)          \n\t"
+            "punpckhdq %%mm5, %%mm5         \n\t"
+            "movd      %%mm5, (%0, %2)      \n\t"
+            "movd      %%mm3, (%0, %2, 2)   \n\t"
+            "punpckhdq %%mm3, %%mm3         \n\t"
+            "movd      %%mm3, (%0, %3)      \n\t"
+            "movd      %%mm1, (%1)          \n\t"
+            "punpckhdq %%mm1, %%mm1         \n\t"
+            "movd      %%mm1, (%1, %2)      \n\t"
+            "movd      %%mm6, (%1, %2, 2)   \n\t"
+            "punpckhdq %%mm6, %%mm6         \n\t"
+            "movd      %%mm6, (%1, %3)      \n\t"
+            :: "r"(src),
+               "r"(src + 4 * stride),
+               "r"((x86_reg)stride),
+               "r"((x86_reg)(3 * stride))
+            );
     }
 }
 
-/* draw the edges of width 'w' of an image of size width, height
-   this mmx version can only handle w==8 || w==16 */
-static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides)
+/* Draw the edges of width 'w' of an image of size width, height
+ * this MMX version can only handle w == 8 || w == 16. */
+static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height,
+                           int w, int h, int sides)
 {
     uint8_t *ptr, *last_line;
     int i;
@@ -773,794 +788,1000 @@ static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w,
     last_line = buf + (height - 1) * wrap;
     /* left and right */
     ptr = buf;
-    if(w==8)
-    {
-        __asm__ volatile(
-                "1:                             \n\t"
-                "movd (%0), %%mm0               \n\t"
-                "punpcklbw %%mm0, %%mm0         \n\t"
-                "punpcklwd %%mm0, %%mm0         \n\t"
-                "punpckldq %%mm0, %%mm0         \n\t"
-                "movq %%mm0, -8(%0)             \n\t"
-                "movq -8(%0, %2), %%mm1         \n\t"
-                "punpckhbw %%mm1, %%mm1         \n\t"
-                "punpckhwd %%mm1, %%mm1         \n\t"
-                "punpckhdq %%mm1, %%mm1         \n\t"
-                "movq %%mm1, (%0, %2)           \n\t"
-                "add %1, %0                     \n\t"
-                "cmp %3, %0                     \n\t"
-                " jb 1b                         \n\t"
-                : "+r" (ptr)
-                : "r" ((x86_reg)wrap), "r" ((x86_reg)width), "r" (ptr + wrap*height)
-        );
-    }
-    else
-    {
-        __asm__ volatile(
-                "1:                             \n\t"
-                "movd (%0), %%mm0               \n\t"
-                "punpcklbw %%mm0, %%mm0         \n\t"
-                "punpcklwd %%mm0, %%mm0         \n\t"
-                "punpckldq %%mm0, %%mm0         \n\t"
-                "movq %%mm0, -8(%0)             \n\t"
-                "movq %%mm0, -16(%0)            \n\t"
-                "movq -8(%0, %2), %%mm1         \n\t"
-                "punpckhbw %%mm1, %%mm1         \n\t"
-                "punpckhwd %%mm1, %%mm1         \n\t"
-                "punpckhdq %%mm1, %%mm1         \n\t"
-                "movq %%mm1, (%0, %2)           \n\t"
-                "movq %%mm1, 8(%0, %2)          \n\t"
-                "add %1, %0                     \n\t"
-                "cmp %3, %0                     \n\t"
-                " jb 1b                         \n\t"
-                : "+r" (ptr)
-                : "r" ((x86_reg)wrap), "r" ((x86_reg)width), "r" (ptr + wrap*height)
-        );
+    if (w == 8) {
+        __asm__ volatile (
+            "1:                             \n\t"
+            "movd            (%0), %%mm0    \n\t"
+            "punpcklbw      %%mm0, %%mm0    \n\t"
+            "punpcklwd      %%mm0, %%mm0    \n\t"
+            "punpckldq      %%mm0, %%mm0    \n\t"
+            "movq           %%mm0, -8(%0)   \n\t"
+            "movq      -8(%0, %2), %%mm1    \n\t"
+            "punpckhbw      %%mm1, %%mm1    \n\t"
+            "punpckhwd      %%mm1, %%mm1    \n\t"
+            "punpckhdq      %%mm1, %%mm1    \n\t"
+            "movq           %%mm1, (%0, %2) \n\t"
+            "add               %1, %0       \n\t"
+            "cmp               %3, %0       \n\t"
+            "jb                1b           \n\t"
+            : "+r"(ptr)
+            : "r"((x86_reg)wrap), "r"((x86_reg)width), "r"(ptr + wrap * height)
+            );
+    } else {
+        __asm__ volatile (
+            "1:                                 \n\t"
+            "movd            (%0), %%mm0        \n\t"
+            "punpcklbw      %%mm0, %%mm0        \n\t"
+            "punpcklwd      %%mm0, %%mm0        \n\t"
+            "punpckldq      %%mm0, %%mm0        \n\t"
+            "movq           %%mm0, -8(%0)       \n\t"
+            "movq           %%mm0, -16(%0)      \n\t"
+            "movq      -8(%0, %2), %%mm1        \n\t"
+            "punpckhbw      %%mm1, %%mm1        \n\t"
+            "punpckhwd      %%mm1, %%mm1        \n\t"
+            "punpckhdq      %%mm1, %%mm1        \n\t"
+            "movq           %%mm1,  (%0, %2)    \n\t"
+            "movq           %%mm1, 8(%0, %2)    \n\t"
+            "add               %1, %0           \n\t"
+            "cmp               %3, %0           \n\t"
+            "jb                1b               \n\t"
+            : "+r"(ptr)
+            : "r"((x86_reg)wrap), "r"((x86_reg)width), "r"(ptr + wrap * height)
+            );
     }
 
     /* top and bottom (and hopefully also the corners) */
-    if (sides&EDGE_TOP) {
-        for(i = 0; i < h; i += 4) {
-            ptr= buf - (i + 1) * wrap - w;
-            __asm__ volatile(
-                    "1:                             \n\t"
-                    "movq (%1, %0), %%mm0           \n\t"
-                    "movq %%mm0, (%0)               \n\t"
-                    "movq %%mm0, (%0, %2)           \n\t"
-                    "movq %%mm0, (%0, %2, 2)        \n\t"
-                    "movq %%mm0, (%0, %3)           \n\t"
-                    "add $8, %0                     \n\t"
-                    "cmp %4, %0                     \n\t"
-                    " jb 1b                         \n\t"
-                    : "+r" (ptr)
-                    : "r" ((x86_reg)buf - (x86_reg)ptr - w), "r" ((x86_reg)-wrap), "r" ((x86_reg)-wrap*3), "r" (ptr+width+2*w)
-            );
+    if (sides & EDGE_TOP) {
+        for (i = 0; i < h; i += 4) {
+            ptr = buf - (i + 1) * wrap - w;
+            __asm__ volatile (
+                "1:                             \n\t"
+                "movq (%1, %0), %%mm0           \n\t"
+                "movq    %%mm0, (%0)            \n\t"
+                "movq    %%mm0, (%0, %2)        \n\t"
+                "movq    %%mm0, (%0, %2, 2)     \n\t"
+                "movq    %%mm0, (%0, %3)        \n\t"
+                "add        $8, %0              \n\t"
+                "cmp        %4, %0              \n\t"
+                "jb         1b                  \n\t"
+                : "+r"(ptr)
+                : "r"((x86_reg)buf - (x86_reg)ptr - w), "r"((x86_reg) -wrap),
+                  "r"((x86_reg) -wrap * 3), "r"(ptr + width + 2 * w)
+                );
         }
     }
 
-    if (sides&EDGE_BOTTOM) {
-        for(i = 0; i < h; i += 4) {
-            ptr= last_line + (i + 1) * wrap - w;
-            __asm__ volatile(
-                    "1:                             \n\t"
-                    "movq (%1, %0), %%mm0           \n\t"
-                    "movq %%mm0, (%0)               \n\t"
-                    "movq %%mm0, (%0, %2)           \n\t"
-                    "movq %%mm0, (%0, %2, 2)        \n\t"
-                    "movq %%mm0, (%0, %3)           \n\t"
-                    "add $8, %0                     \n\t"
-                    "cmp %4, %0                     \n\t"
-                    " jb 1b                         \n\t"
-                    : "+r" (ptr)
-                    : "r" ((x86_reg)last_line - (x86_reg)ptr - w), "r" ((x86_reg)wrap), "r" ((x86_reg)wrap*3), "r" (ptr+width+2*w)
-            );
+    if (sides & EDGE_BOTTOM) {
+        for (i = 0; i < h; i += 4) {
+            ptr = last_line + (i + 1) * wrap - w;
+            __asm__ volatile (
+                "1:                             \n\t"
+                "movq (%1, %0), %%mm0           \n\t"
+                "movq    %%mm0, (%0)            \n\t"
+                "movq    %%mm0, (%0, %2)        \n\t"
+                "movq    %%mm0, (%0, %2, 2)     \n\t"
+                "movq    %%mm0, (%0, %3)        \n\t"
+                "add        $8, %0              \n\t"
+                "cmp        %4, %0              \n\t"
+                "jb         1b                  \n\t"
+                : "+r"(ptr)
+                : "r"((x86_reg)last_line - (x86_reg)ptr - w),
+                  "r"((x86_reg)wrap), "r"((x86_reg)wrap * 3),
+                  "r"(ptr + width + 2 * w)
+                );
         }
     }
 }
 
-#define QPEL_V_LOW(m3,m4,m5,m6, pw_20, pw_3, rnd, in0, in1, in2, in7, out, OP)\
-        "paddw " #m4 ", " #m3 "           \n\t" /* x1 */\
-        "movq "MANGLE(ff_pw_20)", %%mm4   \n\t" /* 20 */\
-        "pmullw " #m3 ", %%mm4            \n\t" /* 20x1 */\
-        "movq "#in7", " #m3 "             \n\t" /* d */\
-        "movq "#in0", %%mm5               \n\t" /* D */\
-        "paddw " #m3 ", %%mm5             \n\t" /* x4 */\
-        "psubw %%mm5, %%mm4               \n\t" /* 20x1 - x4 */\
-        "movq "#in1", %%mm5               \n\t" /* C */\
-        "movq "#in2", %%mm6               \n\t" /* B */\
-        "paddw " #m6 ", %%mm5             \n\t" /* x3 */\
-        "paddw " #m5 ", %%mm6             \n\t" /* x2 */\
-        "paddw %%mm6, %%mm6               \n\t" /* 2x2 */\
-        "psubw %%mm6, %%mm5               \n\t" /* -2x2 + x3 */\
-        "pmullw "MANGLE(ff_pw_3)", %%mm5  \n\t" /* -6x2 + 3x3 */\
-        "paddw " #rnd ", %%mm4            \n\t" /* x2 */\
-        "paddw %%mm4, %%mm5               \n\t" /* 20x1 - 6x2 + 3x3 - x4 */\
-        "psraw $5, %%mm5                  \n\t"\
-        "packuswb %%mm5, %%mm5            \n\t"\
-        OP(%%mm5, out, %%mm7, d)
-
-#define QPEL_BASE(OPNAME, ROUNDER, RND, OP_MMX2, OP_3DNOW)\
-static void OPNAME ## mpeg4_qpel16_h_lowpass_mmx2(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
-    uint64_t temp;\
-\
-    __asm__ volatile(\
-        "pxor %%mm7, %%mm7                \n\t"\
-        "1:                               \n\t"\
-        "movq  (%0), %%mm0                \n\t" /* ABCDEFGH */\
-        "movq %%mm0, %%mm1                \n\t" /* ABCDEFGH */\
-        "movq %%mm0, %%mm2                \n\t" /* ABCDEFGH */\
-        "punpcklbw %%mm7, %%mm0           \n\t" /* 0A0B0C0D */\
-        "punpckhbw %%mm7, %%mm1           \n\t" /* 0E0F0G0H */\
-        "pshufw $0x90, %%mm0, %%mm5       \n\t" /* 0A0A0B0C */\
-        "pshufw $0x41, %%mm0, %%mm6       \n\t" /* 0B0A0A0B */\
-        "movq %%mm2, %%mm3                \n\t" /* ABCDEFGH */\
-        "movq %%mm2, %%mm4                \n\t" /* ABCDEFGH */\
-        "psllq $8, %%mm2                  \n\t" /* 0ABCDEFG */\
-        "psllq $16, %%mm3                 \n\t" /* 00ABCDEF */\
-        "psllq $24, %%mm4                 \n\t" /* 000ABCDE */\
-        "punpckhbw %%mm7, %%mm2           \n\t" /* 0D0E0F0G */\
-        "punpckhbw %%mm7, %%mm3           \n\t" /* 0C0D0E0F */\
-        "punpckhbw %%mm7, %%mm4           \n\t" /* 0B0C0D0E */\
-        "paddw %%mm3, %%mm5               \n\t" /* b */\
-        "paddw %%mm2, %%mm6               \n\t" /* c */\
-        "paddw %%mm5, %%mm5               \n\t" /* 2b */\
-        "psubw %%mm5, %%mm6               \n\t" /* c - 2b */\
-        "pshufw $0x06, %%mm0, %%mm5       \n\t" /* 0C0B0A0A */\
-        "pmullw "MANGLE(ff_pw_3)", %%mm6  \n\t" /* 3c - 6b */\
-        "paddw %%mm4, %%mm0               \n\t" /* a */\
-        "paddw %%mm1, %%mm5               \n\t" /* d */\
-        "pmullw "MANGLE(ff_pw_20)", %%mm0 \n\t" /* 20a */\
-        "psubw %%mm5, %%mm0               \n\t" /* 20a - d */\
-        "paddw %6, %%mm6                  \n\t"\
-        "paddw %%mm6, %%mm0               \n\t" /* 20a - 6b + 3c - d */\
-        "psraw $5, %%mm0                  \n\t"\
-        "movq %%mm0, %5                   \n\t"\
-        /* mm1=EFGH, mm2=DEFG, mm3=CDEF, mm4=BCDE, mm7=0 */\
-        \
-        "movq 5(%0), %%mm0                \n\t" /* FGHIJKLM */\
-        "movq %%mm0, %%mm5                \n\t" /* FGHIJKLM */\
-        "movq %%mm0, %%mm6                \n\t" /* FGHIJKLM */\
-        "psrlq $8, %%mm0                  \n\t" /* GHIJKLM0 */\
-        "psrlq $16, %%mm5                 \n\t" /* HIJKLM00 */\
-        "punpcklbw %%mm7, %%mm0           \n\t" /* 0G0H0I0J */\
-        "punpcklbw %%mm7, %%mm5           \n\t" /* 0H0I0J0K */\
-        "paddw %%mm0, %%mm2               \n\t" /* b */\
-        "paddw %%mm5, %%mm3               \n\t" /* c */\
-        "paddw %%mm2, %%mm2               \n\t" /* 2b */\
-        "psubw %%mm2, %%mm3               \n\t" /* c - 2b */\
-        "movq %%mm6, %%mm2                \n\t" /* FGHIJKLM */\
-        "psrlq $24, %%mm6                 \n\t" /* IJKLM000 */\
-        "punpcklbw %%mm7, %%mm2           \n\t" /* 0F0G0H0I */\
-        "punpcklbw %%mm7, %%mm6           \n\t" /* 0I0J0K0L */\
-        "pmullw "MANGLE(ff_pw_3)", %%mm3  \n\t" /* 3c - 6b */\
-        "paddw %%mm2, %%mm1               \n\t" /* a */\
-        "paddw %%mm6, %%mm4               \n\t" /* d */\
-        "pmullw "MANGLE(ff_pw_20)", %%mm1 \n\t" /* 20a */\
-        "psubw %%mm4, %%mm3               \n\t" /* - 6b +3c - d */\
-        "paddw %6, %%mm1                  \n\t"\
-        "paddw %%mm1, %%mm3               \n\t" /* 20a - 6b +3c - d */\
-        "psraw $5, %%mm3                  \n\t"\
-        "movq %5, %%mm1                   \n\t"\
-        "packuswb %%mm3, %%mm1            \n\t"\
-        OP_MMX2(%%mm1, (%1),%%mm4, q)\
-        /* mm0= GHIJ, mm2=FGHI, mm5=HIJK, mm6=IJKL, mm7=0 */\
-        \
-        "movq 9(%0), %%mm1                \n\t" /* JKLMNOPQ */\
-        "movq %%mm1, %%mm4                \n\t" /* JKLMNOPQ */\
-        "movq %%mm1, %%mm3                \n\t" /* JKLMNOPQ */\
-        "psrlq $8, %%mm1                  \n\t" /* KLMNOPQ0 */\
-        "psrlq $16, %%mm4                 \n\t" /* LMNOPQ00 */\
-        "punpcklbw %%mm7, %%mm1           \n\t" /* 0K0L0M0N */\
-        "punpcklbw %%mm7, %%mm4           \n\t" /* 0L0M0N0O */\
-        "paddw %%mm1, %%mm5               \n\t" /* b */\
-        "paddw %%mm4, %%mm0               \n\t" /* c */\
-        "paddw %%mm5, %%mm5               \n\t" /* 2b */\
-        "psubw %%mm5, %%mm0               \n\t" /* c - 2b */\
-        "movq %%mm3, %%mm5                \n\t" /* JKLMNOPQ */\
-        "psrlq $24, %%mm3                 \n\t" /* MNOPQ000 */\
-        "pmullw "MANGLE(ff_pw_3)", %%mm0  \n\t" /* 3c - 6b */\
-        "punpcklbw %%mm7, %%mm3           \n\t" /* 0M0N0O0P */\
-        "paddw %%mm3, %%mm2               \n\t" /* d */\
-        "psubw %%mm2, %%mm0               \n\t" /* -6b + 3c - d */\
-        "movq %%mm5, %%mm2                \n\t" /* JKLMNOPQ */\
-        "punpcklbw %%mm7, %%mm2           \n\t" /* 0J0K0L0M */\
-        "punpckhbw %%mm7, %%mm5           \n\t" /* 0N0O0P0Q */\
-        "paddw %%mm2, %%mm6               \n\t" /* a */\
-        "pmullw "MANGLE(ff_pw_20)", %%mm6 \n\t" /* 20a */\
-        "paddw %6, %%mm0                  \n\t"\
-        "paddw %%mm6, %%mm0               \n\t" /* 20a - 6b + 3c - d */\
-        "psraw $5, %%mm0                  \n\t"\
-        /* mm1=KLMN, mm2=JKLM, mm3=MNOP, mm4=LMNO, mm5=NOPQ mm7=0 */\
-        \
-        "paddw %%mm5, %%mm3               \n\t" /* a */\
-        "pshufw $0xF9, %%mm5, %%mm6       \n\t" /* 0O0P0Q0Q */\
-        "paddw %%mm4, %%mm6               \n\t" /* b */\
-        "pshufw $0xBE, %%mm5, %%mm4       \n\t" /* 0P0Q0Q0P */\
-        "pshufw $0x6F, %%mm5, %%mm5       \n\t" /* 0Q0Q0P0O */\
-        "paddw %%mm1, %%mm4               \n\t" /* c */\
-        "paddw %%mm2, %%mm5               \n\t" /* d */\
-        "paddw %%mm6, %%mm6               \n\t" /* 2b */\
-        "psubw %%mm6, %%mm4               \n\t" /* c - 2b */\
-        "pmullw "MANGLE(ff_pw_20)", %%mm3 \n\t" /* 20a */\
-        "pmullw "MANGLE(ff_pw_3)", %%mm4  \n\t" /* 3c - 6b */\
-        "psubw %%mm5, %%mm3               \n\t" /* -6b + 3c - d */\
-        "paddw %6, %%mm4                  \n\t"\
-        "paddw %%mm3, %%mm4               \n\t" /* 20a - 6b + 3c - d */\
-        "psraw $5, %%mm4                  \n\t"\
-        "packuswb %%mm4, %%mm0            \n\t"\
-        OP_MMX2(%%mm0, 8(%1), %%mm4, q)\
-        \
-        "add %3, %0                       \n\t"\
-        "add %4, %1                       \n\t"\
-        "decl %2                          \n\t"\
-        " jnz 1b                          \n\t"\
-        : "+a"(src), "+c"(dst), "+D"(h)\
-        : "d"((x86_reg)srcStride), "S"((x86_reg)dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(temp), "m"(ROUNDER)\
-        : "memory"\
-    );\
-}\
-\
-static void OPNAME ## mpeg4_qpel16_h_lowpass_3dnow(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
-    int i;\
-    int16_t temp[16];\
-    /* quick HACK, XXX FIXME MUST be optimized */\
-    for(i=0; i<h; i++)\
-    {\
-        temp[ 0]= (src[ 0]+src[ 1])*20 - (src[ 0]+src[ 2])*6 + (src[ 1]+src[ 3])*3 - (src[ 2]+src[ 4]);\
-        temp[ 1]= (src[ 1]+src[ 2])*20 - (src[ 0]+src[ 3])*6 + (src[ 0]+src[ 4])*3 - (src[ 1]+src[ 5]);\
-        temp[ 2]= (src[ 2]+src[ 3])*20 - (src[ 1]+src[ 4])*6 + (src[ 0]+src[ 5])*3 - (src[ 0]+src[ 6]);\
-        temp[ 3]= (src[ 3]+src[ 4])*20 - (src[ 2]+src[ 5])*6 + (src[ 1]+src[ 6])*3 - (src[ 0]+src[ 7]);\
-        temp[ 4]= (src[ 4]+src[ 5])*20 - (src[ 3]+src[ 6])*6 + (src[ 2]+src[ 7])*3 - (src[ 1]+src[ 8]);\
-        temp[ 5]= (src[ 5]+src[ 6])*20 - (src[ 4]+src[ 7])*6 + (src[ 3]+src[ 8])*3 - (src[ 2]+src[ 9]);\
-        temp[ 6]= (src[ 6]+src[ 7])*20 - (src[ 5]+src[ 8])*6 + (src[ 4]+src[ 9])*3 - (src[ 3]+src[10]);\
-        temp[ 7]= (src[ 7]+src[ 8])*20 - (src[ 6]+src[ 9])*6 + (src[ 5]+src[10])*3 - (src[ 4]+src[11]);\
-        temp[ 8]= (src[ 8]+src[ 9])*20 - (src[ 7]+src[10])*6 + (src[ 6]+src[11])*3 - (src[ 5]+src[12]);\
-        temp[ 9]= (src[ 9]+src[10])*20 - (src[ 8]+src[11])*6 + (src[ 7]+src[12])*3 - (src[ 6]+src[13]);\
-        temp[10]= (src[10]+src[11])*20 - (src[ 9]+src[12])*6 + (src[ 8]+src[13])*3 - (src[ 7]+src[14]);\
-        temp[11]= (src[11]+src[12])*20 - (src[10]+src[13])*6 + (src[ 9]+src[14])*3 - (src[ 8]+src[15]);\
-        temp[12]= (src[12]+src[13])*20 - (src[11]+src[14])*6 + (src[10]+src[15])*3 - (src[ 9]+src[16]);\
-        temp[13]= (src[13]+src[14])*20 - (src[12]+src[15])*6 + (src[11]+src[16])*3 - (src[10]+src[16]);\
-        temp[14]= (src[14]+src[15])*20 - (src[13]+src[16])*6 + (src[12]+src[16])*3 - (src[11]+src[15]);\
-        temp[15]= (src[15]+src[16])*20 - (src[14]+src[16])*6 + (src[13]+src[15])*3 - (src[12]+src[14]);\
-        __asm__ volatile(\
-            "movq (%0), %%mm0               \n\t"\
-            "movq 8(%0), %%mm1              \n\t"\
-            "paddw %2, %%mm0                \n\t"\
-            "paddw %2, %%mm1                \n\t"\
-            "psraw $5, %%mm0                \n\t"\
-            "psraw $5, %%mm1                \n\t"\
-            "packuswb %%mm1, %%mm0          \n\t"\
-            OP_3DNOW(%%mm0, (%1), %%mm1, q)\
-            "movq 16(%0), %%mm0             \n\t"\
-            "movq 24(%0), %%mm1             \n\t"\
-            "paddw %2, %%mm0                \n\t"\
-            "paddw %2, %%mm1                \n\t"\
-            "psraw $5, %%mm0                \n\t"\
-            "psraw $5, %%mm1                \n\t"\
-            "packuswb %%mm1, %%mm0          \n\t"\
-            OP_3DNOW(%%mm0, 8(%1), %%mm1, q)\
-            :: "r"(temp), "r"(dst), "m"(ROUNDER)\
-            : "memory"\
-        );\
-        dst+=dstStride;\
-        src+=srcStride;\
-    }\
-}\
-\
-static void OPNAME ## mpeg4_qpel8_h_lowpass_mmx2(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
-    __asm__ volatile(\
-        "pxor %%mm7, %%mm7                \n\t"\
-        "1:                               \n\t"\
-        "movq  (%0), %%mm0                \n\t" /* ABCDEFGH */\
-        "movq %%mm0, %%mm1                \n\t" /* ABCDEFGH */\
-        "movq %%mm0, %%mm2                \n\t" /* ABCDEFGH */\
-        "punpcklbw %%mm7, %%mm0           \n\t" /* 0A0B0C0D */\
-        "punpckhbw %%mm7, %%mm1           \n\t" /* 0E0F0G0H */\
-        "pshufw $0x90, %%mm0, %%mm5       \n\t" /* 0A0A0B0C */\
-        "pshufw $0x41, %%mm0, %%mm6       \n\t" /* 0B0A0A0B */\
-        "movq %%mm2, %%mm3                \n\t" /* ABCDEFGH */\
-        "movq %%mm2, %%mm4                \n\t" /* ABCDEFGH */\
-        "psllq $8, %%mm2                  \n\t" /* 0ABCDEFG */\
-        "psllq $16, %%mm3                 \n\t" /* 00ABCDEF */\
-        "psllq $24, %%mm4                 \n\t" /* 000ABCDE */\
-        "punpckhbw %%mm7, %%mm2           \n\t" /* 0D0E0F0G */\
-        "punpckhbw %%mm7, %%mm3           \n\t" /* 0C0D0E0F */\
-        "punpckhbw %%mm7, %%mm4           \n\t" /* 0B0C0D0E */\
-        "paddw %%mm3, %%mm5               \n\t" /* b */\
-        "paddw %%mm2, %%mm6               \n\t" /* c */\
-        "paddw %%mm5, %%mm5               \n\t" /* 2b */\
-        "psubw %%mm5, %%mm6               \n\t" /* c - 2b */\
-        "pshufw $0x06, %%mm0, %%mm5       \n\t" /* 0C0B0A0A */\
-        "pmullw "MANGLE(ff_pw_3)", %%mm6  \n\t" /* 3c - 6b */\
-        "paddw %%mm4, %%mm0               \n\t" /* a */\
-        "paddw %%mm1, %%mm5               \n\t" /* d */\
-        "pmullw "MANGLE(ff_pw_20)", %%mm0 \n\t" /* 20a */\
-        "psubw %%mm5, %%mm0               \n\t" /* 20a - d */\
-        "paddw %5, %%mm6                  \n\t"\
-        "paddw %%mm6, %%mm0               \n\t" /* 20a - 6b + 3c - d */\
-        "psraw $5, %%mm0                  \n\t"\
-        /* mm1=EFGH, mm2=DEFG, mm3=CDEF, mm4=BCDE, mm7=0 */\
-        \
-        "movd 5(%0), %%mm5                \n\t" /* FGHI */\
-        "punpcklbw %%mm7, %%mm5           \n\t" /* 0F0G0H0I */\
-        "pshufw $0xF9, %%mm5, %%mm6       \n\t" /* 0G0H0I0I */\
-        "paddw %%mm5, %%mm1               \n\t" /* a */\
-        "paddw %%mm6, %%mm2               \n\t" /* b */\
-        "pshufw $0xBE, %%mm5, %%mm6       \n\t" /* 0H0I0I0H */\
-        "pshufw $0x6F, %%mm5, %%mm5       \n\t" /* 0I0I0H0G */\
-        "paddw %%mm6, %%mm3               \n\t" /* c */\
-        "paddw %%mm5, %%mm4               \n\t" /* d */\
-        "paddw %%mm2, %%mm2               \n\t" /* 2b */\
-        "psubw %%mm2, %%mm3               \n\t" /* c - 2b */\
-        "pmullw "MANGLE(ff_pw_20)", %%mm1 \n\t" /* 20a */\
-        "pmullw "MANGLE(ff_pw_3)", %%mm3  \n\t" /* 3c - 6b */\
-        "psubw %%mm4, %%mm3               \n\t" /* -6b + 3c - d */\
-        "paddw %5, %%mm1                  \n\t"\
-        "paddw %%mm1, %%mm3               \n\t" /* 20a - 6b + 3c - d */\
-        "psraw $5, %%mm3                  \n\t"\
-        "packuswb %%mm3, %%mm0            \n\t"\
-        OP_MMX2(%%mm0, (%1), %%mm4, q)\
-        \
-        "add %3, %0                       \n\t"\
-        "add %4, %1                       \n\t"\
-        "decl %2                          \n\t"\
-        " jnz 1b                          \n\t"\
-        : "+a"(src), "+c"(dst), "+d"(h)\
-        : "S"((x86_reg)srcStride), "D"((x86_reg)dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(ROUNDER)\
-        : "memory"\
-    );\
-}\
-\
-static void OPNAME ## mpeg4_qpel8_h_lowpass_3dnow(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
-    int i;\
-    int16_t temp[8];\
-    /* quick HACK, XXX FIXME MUST be optimized */\
-    for(i=0; i<h; i++)\
-    {\
-        temp[ 0]= (src[ 0]+src[ 1])*20 - (src[ 0]+src[ 2])*6 + (src[ 1]+src[ 3])*3 - (src[ 2]+src[ 4]);\
-        temp[ 1]= (src[ 1]+src[ 2])*20 - (src[ 0]+src[ 3])*6 + (src[ 0]+src[ 4])*3 - (src[ 1]+src[ 5]);\
-        temp[ 2]= (src[ 2]+src[ 3])*20 - (src[ 1]+src[ 4])*6 + (src[ 0]+src[ 5])*3 - (src[ 0]+src[ 6]);\
-        temp[ 3]= (src[ 3]+src[ 4])*20 - (src[ 2]+src[ 5])*6 + (src[ 1]+src[ 6])*3 - (src[ 0]+src[ 7]);\
-        temp[ 4]= (src[ 4]+src[ 5])*20 - (src[ 3]+src[ 6])*6 + (src[ 2]+src[ 7])*3 - (src[ 1]+src[ 8]);\
-        temp[ 5]= (src[ 5]+src[ 6])*20 - (src[ 4]+src[ 7])*6 + (src[ 3]+src[ 8])*3 - (src[ 2]+src[ 8]);\
-        temp[ 6]= (src[ 6]+src[ 7])*20 - (src[ 5]+src[ 8])*6 + (src[ 4]+src[ 8])*3 - (src[ 3]+src[ 7]);\
-        temp[ 7]= (src[ 7]+src[ 8])*20 - (src[ 6]+src[ 8])*6 + (src[ 5]+src[ 7])*3 - (src[ 4]+src[ 6]);\
-        __asm__ volatile(\
-            "movq (%0), %%mm0           \n\t"\
-            "movq 8(%0), %%mm1          \n\t"\
-            "paddw %2, %%mm0            \n\t"\
-            "paddw %2, %%mm1            \n\t"\
-            "psraw $5, %%mm0            \n\t"\
-            "psraw $5, %%mm1            \n\t"\
-            "packuswb %%mm1, %%mm0      \n\t"\
-            OP_3DNOW(%%mm0, (%1), %%mm1, q)\
-            :: "r"(temp), "r"(dst), "m"(ROUNDER)\
-            :"memory"\
-        );\
-        dst+=dstStride;\
-        src+=srcStride;\
-    }\
-}
-
-#define QPEL_OP(OPNAME, ROUNDER, RND, OP, MMX)\
-\
-static void OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
-    uint64_t temp[17*4];\
-    uint64_t *temp_ptr= temp;\
-    int count= 17;\
-\
-    /*FIXME unroll */\
-    __asm__ volatile(\
-        "pxor %%mm7, %%mm7              \n\t"\
-        "1:                             \n\t"\
-        "movq (%0), %%mm0               \n\t"\
-        "movq (%0), %%mm1               \n\t"\
-        "movq 8(%0), %%mm2              \n\t"\
-        "movq 8(%0), %%mm3              \n\t"\
-        "punpcklbw %%mm7, %%mm0         \n\t"\
-        "punpckhbw %%mm7, %%mm1         \n\t"\
-        "punpcklbw %%mm7, %%mm2         \n\t"\
-        "punpckhbw %%mm7, %%mm3         \n\t"\
-        "movq %%mm0, (%1)               \n\t"\
-        "movq %%mm1, 17*8(%1)           \n\t"\
-        "movq %%mm2, 2*17*8(%1)         \n\t"\
-        "movq %%mm3, 3*17*8(%1)         \n\t"\
-        "add $8, %1                     \n\t"\
-        "add %3, %0                     \n\t"\
-        "decl %2                        \n\t"\
-        " jnz 1b                        \n\t"\
-        : "+r" (src), "+r" (temp_ptr), "+r"(count)\
-        : "r" ((x86_reg)srcStride)\
-        : "memory"\
-    );\
-    \
-    temp_ptr= temp;\
-    count=4;\
-    \
-/*FIXME reorder for speed */\
-    __asm__ volatile(\
-        /*"pxor %%mm7, %%mm7              \n\t"*/\
-        "1:                             \n\t"\
-        "movq (%0), %%mm0               \n\t"\
-        "movq 8(%0), %%mm1              \n\t"\
-        "movq 16(%0), %%mm2             \n\t"\
-        "movq 24(%0), %%mm3             \n\t"\
-        QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 16(%0),  8(%0),   (%0), 32(%0), (%1), OP)\
-        QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5,  8(%0),   (%0),   (%0), 40(%0), (%1, %3), OP)\
-        "add %4, %1                     \n\t"\
-        QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5,   (%0),   (%0),  8(%0), 48(%0), (%1), OP)\
-        \
-        QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5,   (%0),  8(%0), 16(%0), 56(%0), (%1, %3), OP)\
-        "add %4, %1                     \n\t"\
-        QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5,  8(%0), 16(%0), 24(%0), 64(%0), (%1), OP)\
-        QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 16(%0), 24(%0), 32(%0), 72(%0), (%1, %3), OP)\
-        "add %4, %1                     \n\t"\
-        QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 24(%0), 32(%0), 40(%0), 80(%0), (%1), OP)\
-        QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 32(%0), 40(%0), 48(%0), 88(%0), (%1, %3), OP)\
-        "add %4, %1                     \n\t"\
-        QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 40(%0), 48(%0), 56(%0), 96(%0), (%1), OP)\
-        QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 48(%0), 56(%0), 64(%0),104(%0), (%1, %3), OP)\
-        "add %4, %1                     \n\t"\
-        QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 56(%0), 64(%0), 72(%0),112(%0), (%1), OP)\
-        QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 64(%0), 72(%0), 80(%0),120(%0), (%1, %3), OP)\
-        "add %4, %1                     \n\t"\
-        QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 72(%0), 80(%0), 88(%0),128(%0), (%1), OP)\
-        \
-        QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 80(%0), 88(%0), 96(%0),128(%0), (%1, %3), OP)\
-        "add %4, %1                     \n\t"  \
-        QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 88(%0), 96(%0),104(%0),120(%0), (%1), OP)\
-        QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 96(%0),104(%0),112(%0),112(%0), (%1, %3), OP)\
-        \
-        "add $136, %0                   \n\t"\
-        "add %6, %1                     \n\t"\
-        "decl %2                        \n\t"\
-        " jnz 1b                        \n\t"\
-        \
-        : "+r"(temp_ptr), "+r"(dst), "+g"(count)\
-        : "r"((x86_reg)dstStride), "r"(2*(x86_reg)dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(ROUNDER), "g"(4-14*(x86_reg)dstStride)\
-        :"memory"\
-    );\
-}\
-\
-static void OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
-    uint64_t temp[9*2];\
-    uint64_t *temp_ptr= temp;\
-    int count= 9;\
-\
-    /*FIXME unroll */\
-    __asm__ volatile(\
-        "pxor %%mm7, %%mm7              \n\t"\
-        "1:                             \n\t"\
-        "movq (%0), %%mm0               \n\t"\
-        "movq (%0), %%mm1               \n\t"\
-        "punpcklbw %%mm7, %%mm0         \n\t"\
-        "punpckhbw %%mm7, %%mm1         \n\t"\
-        "movq %%mm0, (%1)               \n\t"\
-        "movq %%mm1, 9*8(%1)            \n\t"\
-        "add $8, %1                     \n\t"\
-        "add %3, %0                     \n\t"\
-        "decl %2                        \n\t"\
-        " jnz 1b                        \n\t"\
-        : "+r" (src), "+r" (temp_ptr), "+r"(count)\
-        : "r" ((x86_reg)srcStride)\
-        : "memory"\
-    );\
-    \
-    temp_ptr= temp;\
-    count=2;\
-    \
-/*FIXME reorder for speed */\
-    __asm__ volatile(\
-        /*"pxor %%mm7, %%mm7              \n\t"*/\
-        "1:                             \n\t"\
-        "movq (%0), %%mm0               \n\t"\
-        "movq 8(%0), %%mm1              \n\t"\
-        "movq 16(%0), %%mm2             \n\t"\
-        "movq 24(%0), %%mm3             \n\t"\
-        QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 16(%0),  8(%0),   (%0), 32(%0), (%1), OP)\
-        QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5,  8(%0),   (%0),   (%0), 40(%0), (%1, %3), OP)\
-        "add %4, %1                     \n\t"\
-        QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5,   (%0),   (%0),  8(%0), 48(%0), (%1), OP)\
-        \
-        QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5,   (%0),  8(%0), 16(%0), 56(%0), (%1, %3), OP)\
-        "add %4, %1                     \n\t"\
-        QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5,  8(%0), 16(%0), 24(%0), 64(%0), (%1), OP)\
-        \
-        QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 16(%0), 24(%0), 32(%0), 64(%0), (%1, %3), OP)\
-        "add %4, %1                     \n\t"\
-        QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 24(%0), 32(%0), 40(%0), 56(%0), (%1), OP)\
-        QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 32(%0), 40(%0), 48(%0), 48(%0), (%1, %3), OP)\
-                \
-        "add $72, %0                    \n\t"\
-        "add %6, %1                     \n\t"\
-        "decl %2                        \n\t"\
-        " jnz 1b                        \n\t"\
-         \
-        : "+r"(temp_ptr), "+r"(dst), "+g"(count)\
-        : "r"((x86_reg)dstStride), "r"(2*(x86_reg)dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(ROUNDER), "g"(4-6*(x86_reg)dstStride)\
-        : "memory"\
-   );\
-}\
-\
-static void OPNAME ## qpel8_mc00_ ## MMX (uint8_t *dst, uint8_t *src, int stride){\
-    OPNAME ## pixels8_ ## MMX(dst, src, stride, 8);\
-}\
-\
-static void OPNAME ## qpel8_mc10_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[8];\
-    uint8_t * const half= (uint8_t*)temp;\
-    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(half, src, 8, stride, 8);\
-    OPNAME ## pixels8_l2_ ## MMX(dst, src, half, stride, stride, 8);\
-}\
-\
-static void OPNAME ## qpel8_mc20_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    OPNAME ## mpeg4_qpel8_h_lowpass_ ## MMX(dst, src, stride, stride, 8);\
-}\
-\
-static void OPNAME ## qpel8_mc30_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[8];\
-    uint8_t * const half= (uint8_t*)temp;\
-    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(half, src, 8, stride, 8);\
-    OPNAME ## pixels8_l2_ ## MMX(dst, src+1, half, stride, stride, 8);\
-}\
-\
-static void OPNAME ## qpel8_mc01_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[8];\
-    uint8_t * const half= (uint8_t*)temp;\
-    put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(half, src, 8, stride);\
-    OPNAME ## pixels8_l2_ ## MMX(dst, src, half, stride, stride, 8);\
-}\
-\
-static void OPNAME ## qpel8_mc02_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(dst, src, stride, stride);\
-}\
-\
-static void OPNAME ## qpel8_mc03_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[8];\
-    uint8_t * const half= (uint8_t*)temp;\
-    put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(half, src, 8, stride);\
-    OPNAME ## pixels8_l2_ ## MMX(dst, src+stride, half, stride, stride, 8);\
-}\
-static void OPNAME ## qpel8_mc11_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[8 + 9];\
-    uint8_t * const halfH= ((uint8_t*)half) + 64;\
-    uint8_t * const halfHV= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, stride, 9);\
-    put ## RND ## pixels8_l2_ ## MMX(halfH, src, halfH, 8, stride, 9);\
-    put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\
-    OPNAME ## pixels8_l2_ ## MMX(dst, halfH, halfHV, stride, 8, 8);\
-}\
-static void OPNAME ## qpel8_mc31_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[8 + 9];\
-    uint8_t * const halfH= ((uint8_t*)half) + 64;\
-    uint8_t * const halfHV= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, stride, 9);\
-    put ## RND ## pixels8_l2_ ## MMX(halfH, src+1, halfH, 8, stride, 9);\
-    put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\
-    OPNAME ## pixels8_l2_ ## MMX(dst, halfH, halfHV, stride, 8, 8);\
-}\
-static void OPNAME ## qpel8_mc13_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[8 + 9];\
-    uint8_t * const halfH= ((uint8_t*)half) + 64;\
-    uint8_t * const halfHV= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, stride, 9);\
-    put ## RND ## pixels8_l2_ ## MMX(halfH, src, halfH, 8, stride, 9);\
-    put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\
-    OPNAME ## pixels8_l2_ ## MMX(dst, halfH+8, halfHV, stride, 8, 8);\
-}\
-static void OPNAME ## qpel8_mc33_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[8 + 9];\
-    uint8_t * const halfH= ((uint8_t*)half) + 64;\
-    uint8_t * const halfHV= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, stride, 9);\
-    put ## RND ## pixels8_l2_ ## MMX(halfH, src+1, halfH, 8, stride, 9);\
-    put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\
-    OPNAME ## pixels8_l2_ ## MMX(dst, halfH+8, halfHV, stride, 8, 8);\
-}\
-static void OPNAME ## qpel8_mc21_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[8 + 9];\
-    uint8_t * const halfH= ((uint8_t*)half) + 64;\
-    uint8_t * const halfHV= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, stride, 9);\
-    put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\
-    OPNAME ## pixels8_l2_ ## MMX(dst, halfH, halfHV, stride, 8, 8);\
-}\
-static void OPNAME ## qpel8_mc23_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[8 + 9];\
-    uint8_t * const halfH= ((uint8_t*)half) + 64;\
-    uint8_t * const halfHV= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, stride, 9);\
-    put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\
-    OPNAME ## pixels8_l2_ ## MMX(dst, halfH+8, halfHV, stride, 8, 8);\
-}\
-static void OPNAME ## qpel8_mc12_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[8 + 9];\
-    uint8_t * const halfH= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, stride, 9);\
-    put ## RND ## pixels8_l2_ ## MMX(halfH, src, halfH, 8, stride, 9);\
-    OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(dst, halfH, stride, 8);\
-}\
-static void OPNAME ## qpel8_mc32_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[8 + 9];\
-    uint8_t * const halfH= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, stride, 9);\
-    put ## RND ## pixels8_l2_ ## MMX(halfH, src+1, halfH, 8, stride, 9);\
-    OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(dst, halfH, stride, 8);\
-}\
-static void OPNAME ## qpel8_mc22_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[9];\
-    uint8_t * const halfH= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, stride, 9);\
-    OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(dst, halfH, stride, 8);\
-}\
-static void OPNAME ## qpel16_mc00_ ## MMX (uint8_t *dst, uint8_t *src, int stride){\
-    OPNAME ## pixels16_ ## MMX(dst, src, stride, 16);\
-}\
-\
-static void OPNAME ## qpel16_mc10_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[32];\
-    uint8_t * const half= (uint8_t*)temp;\
-    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(half, src, 16, stride, 16);\
-    OPNAME ## pixels16_l2_ ## MMX(dst, src, half, stride, stride, 16);\
-}\
-\
-static void OPNAME ## qpel16_mc20_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    OPNAME ## mpeg4_qpel16_h_lowpass_ ## MMX(dst, src, stride, stride, 16);\
-}\
-\
-static void OPNAME ## qpel16_mc30_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[32];\
-    uint8_t * const half= (uint8_t*)temp;\
-    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(half, src, 16, stride, 16);\
-    OPNAME ## pixels16_l2_ ## MMX(dst, src+1, half, stride, stride, 16);\
-}\
-\
-static void OPNAME ## qpel16_mc01_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[32];\
-    uint8_t * const half= (uint8_t*)temp;\
-    put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(half, src, 16, stride);\
-    OPNAME ## pixels16_l2_ ## MMX(dst, src, half, stride, stride, 16);\
-}\
-\
-static void OPNAME ## qpel16_mc02_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(dst, src, stride, stride);\
-}\
-\
-static void OPNAME ## qpel16_mc03_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t temp[32];\
-    uint8_t * const half= (uint8_t*)temp;\
-    put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(half, src, 16, stride);\
-    OPNAME ## pixels16_l2_ ## MMX(dst, src+stride, half, stride, stride, 16);\
-}\
-static void OPNAME ## qpel16_mc11_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[16*2 + 17*2];\
-    uint8_t * const halfH= ((uint8_t*)half) + 256;\
-    uint8_t * const halfHV= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, stride, 17);\
-    put ## RND ## pixels16_l2_ ## MMX(halfH, src, halfH, 16, stride, 17);\
-    put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, 16, 16);\
-    OPNAME ## pixels16_l2_ ## MMX(dst, halfH, halfHV, stride, 16, 16);\
-}\
-static void OPNAME ## qpel16_mc31_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[16*2 + 17*2];\
-    uint8_t * const halfH= ((uint8_t*)half) + 256;\
-    uint8_t * const halfHV= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, stride, 17);\
-    put ## RND ## pixels16_l2_ ## MMX(halfH, src+1, halfH, 16, stride, 17);\
-    put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, 16, 16);\
-    OPNAME ## pixels16_l2_ ## MMX(dst, halfH, halfHV, stride, 16, 16);\
-}\
-static void OPNAME ## qpel16_mc13_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[16*2 + 17*2];\
-    uint8_t * const halfH= ((uint8_t*)half) + 256;\
-    uint8_t * const halfHV= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, stride, 17);\
-    put ## RND ## pixels16_l2_ ## MMX(halfH, src, halfH, 16, stride, 17);\
-    put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, 16, 16);\
-    OPNAME ## pixels16_l2_ ## MMX(dst, halfH+16, halfHV, stride, 16, 16);\
-}\
-static void OPNAME ## qpel16_mc33_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[16*2 + 17*2];\
-    uint8_t * const halfH= ((uint8_t*)half) + 256;\
-    uint8_t * const halfHV= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, stride, 17);\
-    put ## RND ## pixels16_l2_ ## MMX(halfH, src+1, halfH, 16, stride, 17);\
-    put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, 16, 16);\
-    OPNAME ## pixels16_l2_ ## MMX(dst, halfH+16, halfHV, stride, 16, 16);\
-}\
-static void OPNAME ## qpel16_mc21_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[16*2 + 17*2];\
-    uint8_t * const halfH= ((uint8_t*)half) + 256;\
-    uint8_t * const halfHV= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, stride, 17);\
-    put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, 16, 16);\
-    OPNAME ## pixels16_l2_ ## MMX(dst, halfH, halfHV, stride, 16, 16);\
-}\
-static void OPNAME ## qpel16_mc23_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[16*2 + 17*2];\
-    uint8_t * const halfH= ((uint8_t*)half) + 256;\
-    uint8_t * const halfHV= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, stride, 17);\
-    put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, 16, 16);\
-    OPNAME ## pixels16_l2_ ## MMX(dst, halfH+16, halfHV, stride, 16, 16);\
-}\
-static void OPNAME ## qpel16_mc12_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[17*2];\
-    uint8_t * const halfH= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, stride, 17);\
-    put ## RND ## pixels16_l2_ ## MMX(halfH, src, halfH, 16, stride, 17);\
-    OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(dst, halfH, stride, 16);\
-}\
-static void OPNAME ## qpel16_mc32_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[17*2];\
-    uint8_t * const halfH= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, stride, 17);\
-    put ## RND ## pixels16_l2_ ## MMX(halfH, src+1, halfH, 16, stride, 17);\
-    OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(dst, halfH, stride, 16);\
-}\
-static void OPNAME ## qpel16_mc22_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    uint64_t half[17*2];\
-    uint8_t * const halfH= ((uint8_t*)half);\
-    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, stride, 17);\
-    OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(dst, halfH, stride, 16);\
-}
-
-#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"\
-"mov" #size " " #a ", " #b "      \n\t"
-#define AVG_MMX2_OP(a,b,temp, size) \
-"mov" #size " " #b ", " #temp "   \n\t"\
-"pavgb " #temp ", " #a "          \n\t"\
-"mov" #size " " #a ", " #b "      \n\t"
-
-QPEL_BASE(put_       , ff_pw_16, _       , PUT_OP, PUT_OP)
-QPEL_BASE(avg_       , ff_pw_16, _       , AVG_MMX2_OP, AVG_3DNOW_OP)
-QPEL_BASE(put_no_rnd_, ff_pw_15, _no_rnd_, PUT_OP, PUT_OP)
-QPEL_OP(put_       , ff_pw_16, _       , PUT_OP, 3dnow)
-QPEL_OP(avg_       , ff_pw_16, _       , AVG_3DNOW_OP, 3dnow)
-QPEL_OP(put_no_rnd_, ff_pw_15, _no_rnd_, PUT_OP, 3dnow)
-QPEL_OP(put_       , ff_pw_16, _       , PUT_OP, mmx2)
-QPEL_OP(avg_       , ff_pw_16, _       , AVG_MMX2_OP, mmx2)
-QPEL_OP(put_no_rnd_, ff_pw_15, _no_rnd_, PUT_OP, mmx2)
+#define QPEL_V_LOW(m3, m4, m5, m6, pw_20, pw_3, rnd,                      \
+                   in0, in1, in2, in7, out, OP)                           \
+    "paddw               "#m4", "#m3"   \n\t" /* x1 */                    \
+    "movq   "MANGLE(ff_pw_20)", %%mm4   \n\t" /* 20 */                    \
+    "pmullw              "#m3", %%mm4   \n\t" /* 20x1 */                  \
+    "movq               "#in7", "#m3"   \n\t" /* d */                     \
+    "movq               "#in0", %%mm5   \n\t" /* D */                     \
+    "paddw               "#m3", %%mm5   \n\t" /* x4 */                    \
+    "psubw               %%mm5, %%mm4   \n\t" /* 20x1 - x4 */             \
+    "movq               "#in1", %%mm5   \n\t" /* C */                     \
+    "movq               "#in2", %%mm6   \n\t" /* B */                     \
+    "paddw               "#m6", %%mm5   \n\t" /* x3 */                    \
+    "paddw               "#m5", %%mm6   \n\t" /* x2 */                    \
+    "paddw               %%mm6, %%mm6   \n\t" /* 2x2 */                   \
+    "psubw               %%mm6, %%mm5   \n\t" /* -2x2 + x3 */             \
+    "pmullw  "MANGLE(ff_pw_3)", %%mm5   \n\t" /* -6x2 + 3x3 */            \
+    "paddw              "#rnd", %%mm4   \n\t" /* x2 */                    \
+    "paddw               %%mm4, %%mm5   \n\t" /* 20x1 - 6x2 + 3x3 - x4 */ \
+    "psraw                  $5, %%mm5   \n\t"                             \
+    "packuswb            %%mm5, %%mm5   \n\t"                             \
+    OP(%%mm5, out, %%mm7, d)
+
+#define QPEL_BASE(OPNAME, ROUNDER, RND, OP_MMX2, OP_3DNOW)                \
+static void OPNAME ## mpeg4_qpel16_h_lowpass_mmx2(uint8_t *dst,           \
+                                                  uint8_t *src,           \
+                                                  int dstStride,          \
+                                                  int srcStride,          \
+                                                  int h)                  \
+{                                                                         \
+    uint64_t temp;                                                        \
+                                                                          \
+    __asm__ volatile (                                                    \
+        "pxor      %%mm7, %%mm7             \n\t"                         \
+        "1:                                 \n\t"                         \
+        "movq       (%0), %%mm0             \n\t" /* ABCDEFGH */          \
+        "movq      %%mm0, %%mm1             \n\t" /* ABCDEFGH */          \
+        "movq      %%mm0, %%mm2             \n\t" /* ABCDEFGH */          \
+        "punpcklbw %%mm7, %%mm0             \n\t" /* 0A0B0C0D */          \
+        "punpckhbw %%mm7, %%mm1             \n\t" /* 0E0F0G0H */          \
+        "pshufw    $0x90, %%mm0, %%mm5      \n\t" /* 0A0A0B0C */          \
+        "pshufw    $0x41, %%mm0, %%mm6      \n\t" /* 0B0A0A0B */          \
+        "movq      %%mm2, %%mm3             \n\t" /* ABCDEFGH */          \
+        "movq      %%mm2, %%mm4             \n\t" /* ABCDEFGH */          \
+        "psllq        $8, %%mm2             \n\t" /* 0ABCDEFG */          \
+        "psllq       $16, %%mm3             \n\t" /* 00ABCDEF */          \
+        "psllq       $24, %%mm4             \n\t" /* 000ABCDE */          \
+        "punpckhbw %%mm7, %%mm2             \n\t" /* 0D0E0F0G */          \
+        "punpckhbw %%mm7, %%mm3             \n\t" /* 0C0D0E0F */          \
+        "punpckhbw %%mm7, %%mm4             \n\t" /* 0B0C0D0E */          \
+        "paddw     %%mm3, %%mm5             \n\t" /* b */                 \
+        "paddw     %%mm2, %%mm6             \n\t" /* c */                 \
+        "paddw     %%mm5, %%mm5             \n\t" /* 2b */                \
+        "psubw     %%mm5, %%mm6             \n\t" /* c - 2b */            \
+        "pshufw    $0x06, %%mm0, %%mm5      \n\t" /* 0C0B0A0A */          \
+        "pmullw "MANGLE(ff_pw_3)", %%mm6    \n\t" /* 3c - 6b */           \
+        "paddw     %%mm4, %%mm0             \n\t" /* a */                 \
+        "paddw     %%mm1, %%mm5             \n\t" /* d */                 \
+        "pmullw "MANGLE(ff_pw_20)", %%mm0   \n\t" /* 20a */               \
+        "psubw     %%mm5, %%mm0             \n\t" /* 20a - d */           \
+        "paddw        %6, %%mm6             \n\t"                         \
+        "paddw     %%mm6, %%mm0             \n\t" /* 20a - 6b + 3c - d */ \
+        "psraw        $5, %%mm0             \n\t"                         \
+        "movq      %%mm0, %5                \n\t"                         \
+        /* mm1 = EFGH, mm2 = DEFG, mm3 = CDEF, mm4 = BCDE, mm7 = 0 */     \
+                                                                          \
+        "movq      5(%0), %%mm0             \n\t" /* FGHIJKLM */          \
+        "movq      %%mm0, %%mm5             \n\t" /* FGHIJKLM */          \
+        "movq      %%mm0, %%mm6             \n\t" /* FGHIJKLM */          \
+        "psrlq        $8, %%mm0             \n\t" /* GHIJKLM0 */          \
+        "psrlq       $16, %%mm5             \n\t" /* HIJKLM00 */          \
+        "punpcklbw %%mm7, %%mm0             \n\t" /* 0G0H0I0J */          \
+        "punpcklbw %%mm7, %%mm5             \n\t" /* 0H0I0J0K */          \
+        "paddw     %%mm0, %%mm2             \n\t" /* b */                 \
+        "paddw     %%mm5, %%mm3             \n\t" /* c */                 \
+        "paddw     %%mm2, %%mm2             \n\t" /* 2b */                \
+        "psubw     %%mm2, %%mm3             \n\t" /* c - 2b */            \
+        "movq      %%mm6, %%mm2             \n\t" /* FGHIJKLM */          \
+        "psrlq       $24, %%mm6             \n\t" /* IJKLM000 */          \
+        "punpcklbw %%mm7, %%mm2             \n\t" /* 0F0G0H0I */          \
+        "punpcklbw %%mm7, %%mm6             \n\t" /* 0I0J0K0L */          \
+        "pmullw "MANGLE(ff_pw_3)", %%mm3    \n\t" /* 3c - 6b */           \
+        "paddw     %%mm2, %%mm1             \n\t" /* a */                 \
+        "paddw     %%mm6, %%mm4             \n\t" /* d */                 \
+        "pmullw "MANGLE(ff_pw_20)", %%mm1   \n\t" /* 20a */               \
+        "psubw     %%mm4, %%mm3             \n\t" /* - 6b +3c - d */      \
+        "paddw        %6, %%mm1             \n\t"                         \
+        "paddw     %%mm1, %%mm3             \n\t" /* 20a - 6b +3c - d */  \
+        "psraw        $5, %%mm3             \n\t"                         \
+        "movq         %5, %%mm1             \n\t"                         \
+        "packuswb  %%mm3, %%mm1             \n\t"                         \
+        OP_MMX2(%%mm1, (%1), %%mm4, q)                                    \
+        /* mm0 = GHIJ, mm2 = FGHI, mm5 = HIJK, mm6 = IJKL, mm7 = 0 */     \
+                                                                          \
+        "movq      9(%0), %%mm1             \n\t" /* JKLMNOPQ */          \
+        "movq      %%mm1, %%mm4             \n\t" /* JKLMNOPQ */          \
+        "movq      %%mm1, %%mm3             \n\t" /* JKLMNOPQ */          \
+        "psrlq        $8, %%mm1             \n\t" /* KLMNOPQ0 */          \
+        "psrlq       $16, %%mm4             \n\t" /* LMNOPQ00 */          \
+        "punpcklbw %%mm7, %%mm1             \n\t" /* 0K0L0M0N */          \
+        "punpcklbw %%mm7, %%mm4             \n\t" /* 0L0M0N0O */          \
+        "paddw     %%mm1, %%mm5             \n\t" /* b */                 \
+        "paddw     %%mm4, %%mm0             \n\t" /* c */                 \
+        "paddw     %%mm5, %%mm5             \n\t" /* 2b */                \
+        "psubw     %%mm5, %%mm0             \n\t" /* c - 2b */            \
+        "movq      %%mm3, %%mm5             \n\t" /* JKLMNOPQ */          \
+        "psrlq       $24, %%mm3             \n\t" /* MNOPQ000 */          \
+        "pmullw "MANGLE(ff_pw_3)", %%mm0    \n\t" /* 3c - 6b */           \
+        "punpcklbw %%mm7, %%mm3             \n\t" /* 0M0N0O0P */          \
+        "paddw     %%mm3, %%mm2             \n\t" /* d */                 \
+        "psubw     %%mm2, %%mm0             \n\t" /* -6b + 3c - d */      \
+        "movq      %%mm5, %%mm2             \n\t" /* JKLMNOPQ */          \
+        "punpcklbw %%mm7, %%mm2             \n\t" /* 0J0K0L0M */          \
+        "punpckhbw %%mm7, %%mm5             \n\t" /* 0N0O0P0Q */          \
+        "paddw     %%mm2, %%mm6             \n\t" /* a */                 \
+        "pmullw "MANGLE(ff_pw_20)", %%mm6   \n\t" /* 20a */               \
+        "paddw        %6, %%mm0             \n\t"                         \
+        "paddw     %%mm6, %%mm0             \n\t" /* 20a - 6b + 3c - d */ \
+        "psraw        $5, %%mm0             \n\t"                         \
+        /* mm1 = KLMN, mm2 = JKLM, mm3 = MNOP, */                         \
+        /* mm4 = LMNO, mm5 = NOPQ mm7 = 0 */                              \
+                                                                          \
+        "paddw    %%mm5, %%mm3              \n\t" /* a */                 \
+        "pshufw   $0xF9, %%mm5, %%mm6       \n\t" /* 0O0P0Q0Q */          \
+        "paddw    %%mm4, %%mm6              \n\t" /* b */                 \
+        "pshufw   $0xBE, %%mm5, %%mm4       \n\t" /* 0P0Q0Q0P */          \
+        "pshufw   $0x6F, %%mm5, %%mm5       \n\t" /* 0Q0Q0P0O */          \
+        "paddw    %%mm1, %%mm4              \n\t" /* c */                 \
+        "paddw    %%mm2, %%mm5              \n\t" /* d */                 \
+        "paddw    %%mm6, %%mm6              \n\t" /* 2b */                \
+        "psubw    %%mm6, %%mm4              \n\t" /* c - 2b */            \
+        "pmullw "MANGLE(ff_pw_20)", %%mm3   \n\t" /* 20a */               \
+        "pmullw  "MANGLE(ff_pw_3)", %%mm4   \n\t" /* 3c - 6b */           \
+        "psubw    %%mm5, %%mm3              \n\t" /* -6b + 3c - d */      \
+        "paddw       %6, %%mm4              \n\t"                         \
+        "paddw    %%mm3, %%mm4              \n\t" /* 20a - 6b + 3c - d */ \
+        "psraw       $5, %%mm4              \n\t"                         \
+        "packuswb %%mm4, %%mm0              \n\t"                         \
+        OP_MMX2(%%mm0, 8(%1), %%mm4, q)                                   \
+                                                                          \
+        "add         %3, %0                 \n\t"                         \
+        "add         %4, %1                 \n\t"                         \
+        "decl        %2                     \n\t"                         \
+        "jnz         1b                     \n\t"                         \
+        : "+a"(src), "+c"(dst), "+D"(h)                                   \
+        : "d"((x86_reg)srcStride), "S"((x86_reg)dstStride),               \
+          /* "m"(ff_pw_20), "m"(ff_pw_3), */ "m"(temp), "m"(ROUNDER)      \
+        : "memory"                                                        \
+        );                                                                \
+}                                                                         \
+                                                                          \
+static void OPNAME ## mpeg4_qpel16_h_lowpass_3dnow(uint8_t *dst,          \
+                                                   uint8_t *src,          \
+                                                   int dstStride,         \
+                                                   int srcStride,         \
+                                                   int h)                 \
+{                                                                         \
+    int i;                                                                \
+    int16_t temp[16];                                                     \
+    /* quick HACK, XXX FIXME MUST be optimized */                         \
+    for (i = 0; i < h; i++) {                                             \
+        temp[ 0] = (src[ 0] + src[ 1]) * 20 - (src[ 0] + src[ 2]) * 6 +   \
+                   (src[ 1] + src[ 3]) *  3 - (src[ 2] + src[ 4]);        \
+        temp[ 1] = (src[ 1] + src[ 2]) * 20 - (src[ 0] + src[ 3]) * 6 +   \
+                   (src[ 0] + src[ 4]) *  3 - (src[ 1] + src[ 5]);        \
+        temp[ 2] = (src[ 2] + src[ 3]) * 20 - (src[ 1] + src[ 4]) * 6 +   \
+                   (src[ 0] + src[ 5]) *  3 - (src[ 0] + src[ 6]);        \
+        temp[ 3] = (src[ 3] + src[ 4]) * 20 - (src[ 2] + src[ 5]) * 6 +   \
+                   (src[ 1] + src[ 6]) *  3 - (src[ 0] + src[ 7]);        \
+        temp[ 4] = (src[ 4] + src[ 5]) * 20 - (src[ 3] + src[ 6]) * 6 +   \
+                   (src[ 2] + src[ 7]) *  3 - (src[ 1] + src[ 8]);        \
+        temp[ 5] = (src[ 5] + src[ 6]) * 20 - (src[ 4] + src[ 7]) * 6 +   \
+                   (src[ 3] + src[ 8]) *  3 - (src[ 2] + src[ 9]);        \
+        temp[ 6] = (src[ 6] + src[ 7]) * 20 - (src[ 5] + src[ 8]) * 6 +   \
+                   (src[ 4] + src[ 9]) *  3 - (src[ 3] + src[10]);        \
+        temp[ 7] = (src[ 7] + src[ 8]) * 20 - (src[ 6] + src[ 9]) * 6 +   \
+                   (src[ 5] + src[10]) *  3 - (src[ 4] + src[11]);        \
+        temp[ 8] = (src[ 8] + src[ 9]) * 20 - (src[ 7] + src[10]) * 6 +   \
+                   (src[ 6] + src[11]) *  3 - (src[ 5] + src[12]);        \
+        temp[ 9] = (src[ 9] + src[10]) * 20 - (src[ 8] + src[11]) * 6 +   \
+                   (src[ 7] + src[12]) *  3 - (src[ 6] + src[13]);        \
+        temp[10] = (src[10] + src[11]) * 20 - (src[ 9] + src[12]) * 6 +   \
+                   (src[ 8] + src[13]) *  3 - (src[ 7] + src[14]);        \
+        temp[11] = (src[11] + src[12]) * 20 - (src[10] + src[13]) * 6 +   \
+                   (src[ 9] + src[14]) *  3 - (src[ 8] + src[15]);        \
+        temp[12] = (src[12] + src[13]) * 20 - (src[11] + src[14]) * 6 +   \
+                   (src[10] + src[15]) *  3 - (src[ 9] + src[16]);        \
+        temp[13] = (src[13] + src[14]) * 20 - (src[12] + src[15]) * 6 +   \
+                   (src[11] + src[16]) *  3 - (src[10] + src[16]);        \
+        temp[14] = (src[14] + src[15]) * 20 - (src[13] + src[16]) * 6 +   \
+                   (src[12] + src[16]) *  3 - (src[11] + src[15]);        \
+        temp[15] = (src[15] + src[16]) * 20 - (src[14] + src[16]) * 6 +   \
+                   (src[13] + src[15]) *  3 - (src[12] + src[14]);        \
+        __asm__ volatile (                                                \
+            "movq      (%0), %%mm0          \n\t"                         \
+            "movq     8(%0), %%mm1          \n\t"                         \
+            "paddw       %2, %%mm0          \n\t"                         \
+            "paddw       %2, %%mm1          \n\t"                         \
+            "psraw       $5, %%mm0          \n\t"                         \
+            "psraw       $5, %%mm1          \n\t"                         \
+            "packuswb %%mm1, %%mm0          \n\t"                         \
+            OP_3DNOW(%%mm0, (%1), %%mm1, q)                               \
+            "movq    16(%0), %%mm0          \n\t"                         \
+            "movq    24(%0), %%mm1          \n\t"                         \
+            "paddw       %2, %%mm0          \n\t"                         \
+            "paddw       %2, %%mm1          \n\t"                         \
+            "psraw       $5, %%mm0          \n\t"                         \
+            "psraw       $5, %%mm1          \n\t"                         \
+            "packuswb %%mm1, %%mm0          \n\t"                         \
+            OP_3DNOW(%%mm0, 8(%1), %%mm1, q)                              \
+            :: "r"(temp), "r"(dst), "m"(ROUNDER)                          \
+            : "memory"                                                    \
+            );                                                            \
+        dst += dstStride;                                                 \
+        src += srcStride;                                                 \
+    }                                                                     \
+}                                                                         \
+                                                                          \
+static void OPNAME ## mpeg4_qpel8_h_lowpass_mmx2(uint8_t *dst,            \
+                                                 uint8_t *src,            \
+                                                 int dstStride,           \
+                                                 int srcStride,           \
+                                                 int h)                   \
+{                                                                         \
+    __asm__ volatile (                                                    \
+        "pxor      %%mm7, %%mm7             \n\t"                         \
+        "1:                                 \n\t"                         \
+        "movq       (%0), %%mm0             \n\t" /* ABCDEFGH */          \
+        "movq      %%mm0, %%mm1             \n\t" /* ABCDEFGH */          \
+        "movq      %%mm0, %%mm2             \n\t" /* ABCDEFGH */          \
+        "punpcklbw %%mm7, %%mm0             \n\t" /* 0A0B0C0D */          \
+        "punpckhbw %%mm7, %%mm1             \n\t" /* 0E0F0G0H */          \
+        "pshufw    $0x90, %%mm0, %%mm5      \n\t" /* 0A0A0B0C */          \
+        "pshufw    $0x41, %%mm0, %%mm6      \n\t" /* 0B0A0A0B */          \
+        "movq      %%mm2, %%mm3             \n\t" /* ABCDEFGH */          \
+        "movq      %%mm2, %%mm4             \n\t" /* ABCDEFGH */          \
+        "psllq        $8, %%mm2             \n\t" /* 0ABCDEFG */          \
+        "psllq       $16, %%mm3             \n\t" /* 00ABCDEF */          \
+        "psllq       $24, %%mm4             \n\t" /* 000ABCDE */          \
+        "punpckhbw %%mm7, %%mm2             \n\t" /* 0D0E0F0G */          \
+        "punpckhbw %%mm7, %%mm3             \n\t" /* 0C0D0E0F */          \
+        "punpckhbw %%mm7, %%mm4             \n\t" /* 0B0C0D0E */          \
+        "paddw     %%mm3, %%mm5             \n\t" /* b */                 \
+        "paddw     %%mm2, %%mm6             \n\t" /* c */                 \
+        "paddw     %%mm5, %%mm5             \n\t" /* 2b */                \
+        "psubw     %%mm5, %%mm6             \n\t" /* c - 2b */            \
+        "pshufw    $0x06, %%mm0, %%mm5      \n\t" /* 0C0B0A0A */          \
+        "pmullw "MANGLE(ff_pw_3)", %%mm6    \n\t" /* 3c - 6b */           \
+        "paddw     %%mm4, %%mm0             \n\t" /* a */                 \
+        "paddw     %%mm1, %%mm5             \n\t" /* d */                 \
+        "pmullw "MANGLE(ff_pw_20)", %%mm0   \n\t" /* 20a */               \
+        "psubw     %%mm5, %%mm0             \n\t" /* 20a - d */           \
+        "paddw        %5, %%mm6             \n\t"                         \
+        "paddw     %%mm6, %%mm0             \n\t" /* 20a - 6b + 3c - d */ \
+        "psraw        $5, %%mm0             \n\t"                         \
+        /* mm1 = EFGH, mm2 = DEFG, mm3 = CDEF, mm4 = BCDE, mm7 = 0 */     \
+                                                                          \
+        "movd      5(%0), %%mm5             \n\t" /* FGHI */              \
+        "punpcklbw %%mm7, %%mm5             \n\t" /* 0F0G0H0I */          \
+        "pshufw    $0xF9, %%mm5, %%mm6      \n\t" /* 0G0H0I0I */          \
+        "paddw     %%mm5, %%mm1             \n\t" /* a */                 \
+        "paddw     %%mm6, %%mm2             \n\t" /* b */                 \
+        "pshufw    $0xBE, %%mm5, %%mm6      \n\t" /* 0H0I0I0H */          \
+        "pshufw    $0x6F, %%mm5, %%mm5      \n\t" /* 0I0I0H0G */          \
+        "paddw     %%mm6, %%mm3             \n\t" /* c */                 \
+        "paddw     %%mm5, %%mm4             \n\t" /* d */                 \
+        "paddw     %%mm2, %%mm2             \n\t" /* 2b */                \
+        "psubw     %%mm2, %%mm3             \n\t" /* c - 2b */            \
+        "pmullw "MANGLE(ff_pw_20)", %%mm1   \n\t" /* 20a */               \
+        "pmullw  "MANGLE(ff_pw_3)", %%mm3   \n\t" /* 3c - 6b */           \
+        "psubw     %%mm4, %%mm3             \n\t" /* -6b + 3c - d */      \
+        "paddw        %5, %%mm1             \n\t"                         \
+        "paddw     %%mm1, %%mm3             \n\t" /* 20a - 6b + 3c - d */ \
+        "psraw        $5, %%mm3             \n\t"                         \
+        "packuswb  %%mm3, %%mm0             \n\t"                         \
+        OP_MMX2(%%mm0, (%1), %%mm4, q)                                    \
+                                                                          \
+        "add          %3, %0                \n\t"                         \
+        "add          %4, %1                \n\t"                         \
+        "decl         %2                    \n\t"                         \
+        "jnz          1b                    \n\t"                         \
+        : "+a"(src), "+c"(dst), "+d"(h)                                   \
+        : "S"((x86_reg)srcStride), "D"((x86_reg)dstStride),               \
+          /* "m"(ff_pw_20), "m"(ff_pw_3), */ "m"(ROUNDER)                 \
+        : "memory"                                                        \
+        );                                                                \
+}                                                                         \
+                                                                          \
+static void OPNAME ## mpeg4_qpel8_h_lowpass_3dnow(uint8_t *dst,           \
+                                                  uint8_t *src,           \
+                                                  int dstStride,          \
+                                                  int srcStride,          \
+                                                  int h)                  \
+{                                                                         \
+    int i;                                                                \
+    int16_t temp[8];                                                      \
+    /* quick HACK, XXX FIXME MUST be optimized */                         \
+    for (i = 0; i < h; i++) {                                             \
+        temp[0] = (src[0] + src[1]) * 20 - (src[0] + src[2]) * 6 +        \
+                  (src[1] + src[3]) *  3 - (src[2] + src[4]);             \
+        temp[1] = (src[1] + src[2]) * 20 - (src[0] + src[3]) * 6 +        \
+                  (src[0] + src[4]) *  3 - (src[1] + src[5]);             \
+        temp[2] = (src[2] + src[3]) * 20 - (src[1] + src[4]) * 6 +        \
+                  (src[0] + src[5]) *  3 - (src[0] + src[6]);             \
+        temp[3] = (src[3] + src[4]) * 20 - (src[2] + src[5]) * 6 +        \
+                  (src[1] + src[6]) *  3 - (src[0] + src[7]);             \
+        temp[4] = (src[4] + src[5]) * 20 - (src[3] + src[6]) * 6 +        \
+                  (src[2] + src[7]) *  3 - (src[1] + src[8]);             \
+        temp[5] = (src[5] + src[6]) * 20 - (src[4] + src[7]) * 6 +        \
+                  (src[3] + src[8]) *  3 - (src[2] + src[8]);             \
+        temp[6] = (src[6] + src[7]) * 20 - (src[5] + src[8]) * 6 +        \
+                  (src[4] + src[8]) *  3 - (src[3] + src[7]);             \
+        temp[7] = (src[7] + src[8]) * 20 - (src[6] + src[8]) * 6 +        \
+                  (src[5] + src[7]) *  3 - (src[4] + src[6]);             \
+        __asm__ volatile (                                                \
+            "movq      (%0), %%mm0      \n\t"                             \
+            "movq     8(%0), %%mm1      \n\t"                             \
+            "paddw       %2, %%mm0      \n\t"                             \
+            "paddw       %2, %%mm1      \n\t"                             \
+            "psraw       $5, %%mm0      \n\t"                             \
+            "psraw       $5, %%mm1      \n\t"                             \
+            "packuswb %%mm1, %%mm0      \n\t"                             \
+            OP_3DNOW(%%mm0, (%1), %%mm1, q)                               \
+            :: "r"(temp), "r"(dst), "m"(ROUNDER)                          \
+            : "memory"                                                    \
+            );                                                            \
+        dst += dstStride;                                                 \
+        src += srcStride;                                                 \
+    }                                                                     \
+}
+
+#define QPEL_OP(OPNAME, ROUNDER, RND, OP, MMX)                          \
+static void OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(uint8_t *dst,      \
+                                                     uint8_t *src,      \
+                                                     int dstStride,     \
+                                                     int srcStride)     \
+{                                                                       \
+    uint64_t temp[17 * 4];                                              \
+    uint64_t *temp_ptr = temp;                                          \
+    int count = 17;                                                     \
+                                                                        \
+    /* FIXME unroll */                                                  \
+    __asm__ volatile (                                                  \
+        "pxor      %%mm7, %%mm7             \n\t"                       \
+        "1:                                 \n\t"                       \
+        "movq       (%0), %%mm0             \n\t"                       \
+        "movq       (%0), %%mm1             \n\t"                       \
+        "movq      8(%0), %%mm2             \n\t"                       \
+        "movq      8(%0), %%mm3             \n\t"                       \
+        "punpcklbw %%mm7, %%mm0             \n\t"                       \
+        "punpckhbw %%mm7, %%mm1             \n\t"                       \
+        "punpcklbw %%mm7, %%mm2             \n\t"                       \
+        "punpckhbw %%mm7, %%mm3             \n\t"                       \
+        "movq      %%mm0, (%1)              \n\t"                       \
+        "movq      %%mm1, 17 * 8(%1)        \n\t"                       \
+        "movq      %%mm2, 2 * 17 * 8(%1)    \n\t"                       \
+        "movq      %%mm3, 3 * 17 * 8(%1)    \n\t"                       \
+        "add          $8, %1                \n\t"                       \
+        "add          %3, %0                \n\t"                       \
+        "decl         %2                    \n\t"                       \
+        "jnz          1b                    \n\t"                       \
+        : "+r"(src), "+r"(temp_ptr), "+r"(count)                        \
+        : "r"((x86_reg)srcStride)                                       \
+        : "memory"                                                      \
+        );                                                              \
+                                                                        \
+    temp_ptr = temp;                                                    \
+    count    = 4;                                                       \
+                                                                        \
+    /* FIXME reorder for speed */                                       \
+    __asm__ volatile (                                                  \
+        /* "pxor  %%mm7, %%mm7            \n\t" */                      \
+        "1:                             \n\t"                           \
+        "movq    (%0), %%mm0            \n\t"                           \
+        "movq   8(%0), %%mm1            \n\t"                           \
+        "movq  16(%0), %%mm2            \n\t"                           \
+        "movq  24(%0), %%mm3            \n\t"                           \
+        QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 16(%0),   8(%0),    (%0),  32(%0), (%1),     OP) \
+        QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5,  8(%0),    (%0),    (%0),  40(%0), (%1, %3), OP) \
+        "add       %4, %1               \n\t"                           \
+        QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5,   (%0),    (%0),   8(%0),  48(%0), (%1),     OP) \
+                                                                        \
+        QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5,   (%0),   8(%0),  16(%0),  56(%0), (%1, %3), OP) \
+        "add       %4, %1               \n\t"                           \
+        QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5,  8(%0),  16(%0),  24(%0),  64(%0), (%1),     OP) \
+        QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 16(%0),  24(%0),  32(%0),  72(%0), (%1, %3), OP) \
+        "add       %4, %1               \n\t"                           \
+        QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 24(%0),  32(%0),  40(%0),  80(%0), (%1),     OP) \
+        QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 32(%0),  40(%0),  48(%0),  88(%0), (%1, %3), OP) \
+        "add       %4, %1               \n\t"                           \
+        QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 40(%0),  48(%0),  56(%0),  96(%0), (%1),     OP) \
+        QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 48(%0),  56(%0),  64(%0), 104(%0), (%1, %3), OP) \
+        "add       %4, %1               \n\t"                           \
+        QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 56(%0),  64(%0),  72(%0), 112(%0), (%1),     OP) \
+        QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 64(%0),  72(%0),  80(%0), 120(%0), (%1, %3), OP) \
+        "add       %4, %1               \n\t"                           \
+        QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 72(%0),  80(%0),  88(%0), 128(%0), (%1),     OP) \
+                                                                        \
+        QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 80(%0),  88(%0),  96(%0), 128(%0), (%1, %3), OP) \
+        "add       %4, %1               \n\t"                           \
+        QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 88(%0),  96(%0), 104(%0), 120(%0), (%1),     OP) \
+        QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 96(%0), 104(%0), 112(%0), 112(%0), (%1, %3), OP) \
+                                                                        \
+        "add     $136, %0               \n\t"                           \
+        "add       %6, %1               \n\t"                           \
+        "decl      %2                   \n\t"                           \
+        "jnz       1b                   \n\t"                           \
+                                                                        \
+        : "+r"(temp_ptr), "+r"(dst), "+g"(count)                        \
+        : "r"((x86_reg)dstStride), "r"(2 * (x86_reg)dstStride),         \
+          /* "m"(ff_pw_20), "m"(ff_pw_3), */ "m"(ROUNDER),              \
+          "g"(4 - 14 * (x86_reg)dstStride)                              \
+        : "memory"                                                      \
+        );                                                              \
+}                                                                       \
+                                                                        \
+static void OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(uint8_t *dst,       \
+                                                    uint8_t *src,       \
+                                                    int dstStride,      \
+                                                    int srcStride)      \
+{                                                                       \
+    uint64_t temp[9 * 2];                                               \
+    uint64_t *temp_ptr = temp;                                          \
+    int count = 9;                                                      \
+                                                                        \
+    /* FIXME unroll */                                                  \
+    __asm__ volatile (                                                  \
+        "pxor      %%mm7, %%mm7         \n\t"                           \
+        "1:                             \n\t"                           \
+        "movq       (%0), %%mm0         \n\t"                           \
+        "movq       (%0), %%mm1         \n\t"                           \
+        "punpcklbw %%mm7, %%mm0         \n\t"                           \
+        "punpckhbw %%mm7, %%mm1         \n\t"                           \
+        "movq      %%mm0, (%1)          \n\t"                           \
+        "movq      %%mm1, 9*8(%1)       \n\t"                           \
+        "add          $8, %1            \n\t"                           \
+        "add          %3, %0            \n\t"                           \
+        "decl         %2                \n\t"                           \
+        "jnz          1b                \n\t"                           \
+        : "+r"(src), "+r"(temp_ptr), "+r"(count)                        \
+        : "r"((x86_reg)srcStride)                                       \
+        : "memory"                                                      \
+        );                                                              \
+                                                                        \
+    temp_ptr = temp;                                                    \
+    count    = 2;                                                       \
+                                                                        \
+    /* FIXME reorder for speed */                                       \
+    __asm__ volatile (                                                  \
+        /* "pxor  %%mm7, %%mm7            \n\t" */                      \
+        "1:                             \n\t"                           \
+        "movq    (%0), %%mm0            \n\t"                           \
+        "movq   8(%0), %%mm1            \n\t"                           \
+        "movq  16(%0), %%mm2            \n\t"                           \
+        "movq  24(%0), %%mm3            \n\t"                           \
+        QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 16(%0),  8(%0),   (%0), 32(%0), (%1), OP)     \
+        QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5,  8(%0),   (%0),   (%0), 40(%0), (%1, %3), OP) \
+        "add       %4, %1               \n\t"                           \
+        QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5,   (%0),   (%0),  8(%0), 48(%0), (%1), OP)     \
+                                                                        \
+        QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5,   (%0),  8(%0), 16(%0), 56(%0), (%1, %3), OP) \
+        "add       %4, %1               \n\t"                           \
+        QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5,  8(%0), 16(%0), 24(%0), 64(%0), (%1), OP)     \
+                                                                        \
+        QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 16(%0), 24(%0), 32(%0), 64(%0), (%1, %3), OP) \
+        "add       %4, %1               \n\t"                           \
+        QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 24(%0), 32(%0), 40(%0), 56(%0), (%1), OP)     \
+        QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 32(%0), 40(%0), 48(%0), 48(%0), (%1, %3), OP) \
+                                                                        \
+        "add      $72, %0               \n\t"                           \
+        "add       %6, %1               \n\t"                           \
+        "decl      %2                   \n\t"                           \
+        "jnz       1b                   \n\t"                           \
+                                                                        \
+        : "+r"(temp_ptr), "+r"(dst), "+g"(count)                        \
+        : "r"((x86_reg)dstStride), "r"(2 * (x86_reg)dstStride),         \
+          /* "m"(ff_pw_20), "m"(ff_pw_3), */ "m"(ROUNDER),              \
+          "g"(4 - 6 * (x86_reg)dstStride)                               \
+        : "memory"                                                      \
+        );                                                              \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel8_mc00_ ## MMX (uint8_t *dst, uint8_t *src,   \
+                                          int stride)                   \
+{                                                                       \
+    OPNAME ## pixels8_ ## MMX(dst, src, stride, 8);                     \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel8_mc10_ ## MMX(uint8_t *dst, uint8_t *src,    \
+                                         int stride)                    \
+{                                                                       \
+    uint64_t temp[8];                                                   \
+    uint8_t * const half = (uint8_t*)temp;                              \
+    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(half, src, 8,           \
+                                                stride, 8);             \
+    OPNAME ## pixels8_l2_ ## MMX(dst, src, half, stride, stride, 8);    \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel8_mc20_ ## MMX(uint8_t *dst, uint8_t *src,    \
+                                         int stride)                    \
+{                                                                       \
+    OPNAME ## mpeg4_qpel8_h_lowpass_ ## MMX(dst, src, stride,           \
+                                            stride, 8);                 \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel8_mc30_ ## MMX(uint8_t *dst, uint8_t *src,    \
+                                         int stride)                    \
+{                                                                       \
+    uint64_t temp[8];                                                   \
+    uint8_t * const half = (uint8_t*)temp;                              \
+    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(half, src, 8,           \
+                                                stride, 8);             \
+    OPNAME ## pixels8_l2_ ## MMX(dst, src + 1, half, stride,            \
+                                 stride, 8);                            \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel8_mc01_ ## MMX(uint8_t *dst, uint8_t *src,    \
+                                         int stride)                    \
+{                                                                       \
+    uint64_t temp[8];                                                   \
+    uint8_t * const half = (uint8_t*)temp;                              \
+    put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(half, src, 8, stride);  \
+    OPNAME ## pixels8_l2_ ## MMX(dst, src, half, stride, stride, 8);    \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel8_mc02_ ## MMX(uint8_t *dst, uint8_t *src,    \
+                                         int stride)                    \
+{                                                                       \
+    OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(dst, src, stride, stride);  \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel8_mc03_ ## MMX(uint8_t *dst, uint8_t *src,    \
+                                         int stride)                    \
+{                                                                       \
+    uint64_t temp[8];                                                   \
+    uint8_t * const half = (uint8_t*)temp;                              \
+    put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(half, src, 8, stride);  \
+    OPNAME ## pixels8_l2_ ## MMX(dst, src + stride, half, stride,       \
+                                 stride, 8);                            \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel8_mc11_ ## MMX(uint8_t *dst, uint8_t *src,    \
+                                         int stride)                    \
+{                                                                       \
+    uint64_t half[8 + 9];                                               \
+    uint8_t * const halfH  = ((uint8_t*)half) + 64;                     \
+    uint8_t * const halfHV = ((uint8_t*)half);                          \
+    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8,          \
+                                                stride, 9);             \
+    put ## RND ## pixels8_l2_ ## MMX(halfH, src, halfH, 8, stride, 9);  \
+    put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);   \
+    OPNAME ## pixels8_l2_ ## MMX(dst, halfH, halfHV, stride, 8, 8);     \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel8_mc31_ ## MMX(uint8_t *dst, uint8_t *src,    \
+                                         int stride)                    \
+{                                                                       \
+    uint64_t half[8 + 9];                                               \
+    uint8_t * const halfH  = ((uint8_t*)half) + 64;                     \
+    uint8_t * const halfHV = ((uint8_t*)half);                          \
+    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8,          \
+                                                stride, 9);             \
+    put ## RND ## pixels8_l2_ ## MMX(halfH, src + 1, halfH, 8,          \
+                                     stride, 9);                        \
+    put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);   \
+    OPNAME ## pixels8_l2_ ## MMX(dst, halfH, halfHV, stride, 8, 8);     \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel8_mc13_ ## MMX(uint8_t *dst, uint8_t *src,    \
+                                         int stride)                    \
+{                                                                       \
+    uint64_t half[8 + 9];                                               \
+    uint8_t * const halfH  = ((uint8_t*)half) + 64;                     \
+    uint8_t * const halfHV = ((uint8_t*)half);                          \
+    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8,          \
+                                                stride, 9);             \
+    put ## RND ## pixels8_l2_ ## MMX(halfH, src, halfH, 8, stride, 9);  \
+    put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);   \
+    OPNAME ## pixels8_l2_ ## MMX(dst, halfH + 8, halfHV, stride, 8, 8); \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel8_mc33_ ## MMX(uint8_t *dst, uint8_t *src,    \
+                                         int stride)                    \
+{                                                                       \
+    uint64_t half[8 + 9];                                               \
+    uint8_t * const halfH  = ((uint8_t*)half) + 64;                     \
+    uint8_t * const halfHV = ((uint8_t*)half);                          \
+    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8,          \
+                                                stride, 9);             \
+    put ## RND ## pixels8_l2_ ## MMX(halfH, src + 1, halfH, 8,          \
+                                     stride, 9);                        \
+    put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);   \
+    OPNAME ## pixels8_l2_ ## MMX(dst, halfH + 8, halfHV, stride, 8, 8); \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel8_mc21_ ## MMX(uint8_t *dst, uint8_t *src,    \
+                                         int stride)                    \
+{                                                                       \
+    uint64_t half[8 + 9];                                               \
+    uint8_t * const halfH  = ((uint8_t*)half) + 64;                     \
+    uint8_t * const halfHV = ((uint8_t*)half);                          \
+    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8,          \
+                                                stride, 9);             \
+    put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);   \
+    OPNAME ## pixels8_l2_ ## MMX(dst, halfH, halfHV, stride, 8, 8);     \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel8_mc23_ ## MMX(uint8_t *dst, uint8_t *src,    \
+                                         int stride)                    \
+{                                                                       \
+    uint64_t half[8 + 9];                                               \
+    uint8_t * const halfH  = ((uint8_t*)half) + 64;                     \
+    uint8_t * const halfHV = ((uint8_t*)half);                          \
+    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8,          \
+                                                stride, 9);             \
+    put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);   \
+    OPNAME ## pixels8_l2_ ## MMX(dst, halfH + 8, halfHV, stride, 8, 8); \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel8_mc12_ ## MMX(uint8_t *dst, uint8_t *src,    \
+                                         int stride)                    \
+{                                                                       \
+    uint64_t half[8 + 9];                                               \
+    uint8_t * const halfH = ((uint8_t*)half);                           \
+    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8,          \
+                                                stride, 9);             \
+    put ## RND ## pixels8_l2_ ## MMX(halfH, src, halfH, 8, stride, 9);  \
+    OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(dst, halfH, stride, 8);     \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel8_mc32_ ## MMX(uint8_t *dst, uint8_t *src,    \
+                                         int stride)                    \
+{                                                                       \
+    uint64_t half[8 + 9];                                               \
+    uint8_t * const halfH = ((uint8_t*)half);                           \
+    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8,          \
+                                                stride, 9);             \
+    put ## RND ## pixels8_l2_ ## MMX(halfH, src + 1, halfH, 8,          \
+                                     stride, 9);                        \
+    OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(dst, halfH, stride, 8);     \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel8_mc22_ ## MMX(uint8_t *dst, uint8_t *src,    \
+                                         int stride)                    \
+{                                                                       \
+    uint64_t half[9];                                                   \
+    uint8_t * const halfH = ((uint8_t*)half);                           \
+    put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8,          \
+                                                stride, 9);             \
+    OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(dst, halfH, stride, 8);     \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel16_mc00_ ## MMX (uint8_t *dst, uint8_t *src,  \
+                                           int stride)                  \
+{                                                                       \
+    OPNAME ## pixels16_ ## MMX(dst, src, stride, 16);                   \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel16_mc10_ ## MMX(uint8_t *dst, uint8_t *src,   \
+                                          int stride)                   \
+{                                                                       \
+    uint64_t temp[32];                                                  \
+    uint8_t * const half = (uint8_t*)temp;                              \
+    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(half, src, 16,         \
+                                                 stride, 16);           \
+    OPNAME ## pixels16_l2_ ## MMX(dst, src, half, stride, stride, 16);  \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel16_mc20_ ## MMX(uint8_t *dst, uint8_t *src,   \
+                                          int stride)                   \
+{                                                                       \
+    OPNAME ## mpeg4_qpel16_h_lowpass_ ## MMX(dst, src,                  \
+                                             stride, stride, 16);       \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel16_mc30_ ## MMX(uint8_t *dst, uint8_t *src,   \
+                                          int stride)                   \
+{                                                                       \
+    uint64_t temp[32];                                                  \
+    uint8_t * const half = (uint8_t*)temp;                              \
+    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(half, src, 16,         \
+                                                 stride, 16);           \
+    OPNAME ## pixels16_l2_ ## MMX(dst, src + 1, half,                   \
+                                  stride, stride, 16);                  \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel16_mc01_ ## MMX(uint8_t *dst, uint8_t *src,   \
+                                          int stride)                   \
+{                                                                       \
+    uint64_t temp[32];                                                  \
+    uint8_t * const half = (uint8_t*)temp;                              \
+    put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(half, src, 16,         \
+                                                 stride);               \
+    OPNAME ## pixels16_l2_ ## MMX(dst, src, half, stride, stride, 16);  \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel16_mc02_ ## MMX(uint8_t *dst, uint8_t *src,   \
+                                          int stride)                   \
+{                                                                       \
+    OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(dst, src, stride, stride); \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel16_mc03_ ## MMX(uint8_t *dst, uint8_t *src,   \
+                                          int stride)                   \
+{                                                                       \
+    uint64_t temp[32];                                                  \
+    uint8_t * const half = (uint8_t*)temp;                              \
+    put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(half, src, 16,         \
+                                                 stride);               \
+    OPNAME ## pixels16_l2_ ## MMX(dst, src+stride, half,                \
+                                  stride, stride, 16);                  \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel16_mc11_ ## MMX(uint8_t *dst, uint8_t *src,   \
+                                          int stride)                   \
+{                                                                       \
+    uint64_t half[16 * 2 + 17 * 2];                                     \
+    uint8_t * const halfH  = ((uint8_t*)half) + 256;                    \
+    uint8_t * const halfHV = ((uint8_t*)half);                          \
+    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16,        \
+                                                 stride, 17);           \
+    put ## RND ## pixels16_l2_ ## MMX(halfH, src, halfH, 16,            \
+                                      stride, 17);                      \
+    put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH,         \
+                                                 16, 16);               \
+    OPNAME ## pixels16_l2_ ## MMX(dst, halfH, halfHV, stride, 16, 16);  \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel16_mc31_ ## MMX(uint8_t *dst, uint8_t *src,   \
+                                          int stride)                   \
+{                                                                       \
+    uint64_t half[16 * 2 + 17 * 2];                                     \
+    uint8_t * const halfH  = ((uint8_t*)half) + 256;                    \
+    uint8_t * const halfHV = ((uint8_t*)half);                          \
+    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16,        \
+                                                 stride, 17);           \
+    put ## RND ## pixels16_l2_ ## MMX(halfH, src + 1, halfH, 16,        \
+                                      stride, 17);                      \
+    put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH,         \
+                                                 16, 16);               \
+    OPNAME ## pixels16_l2_ ## MMX(dst, halfH, halfHV, stride, 16, 16);  \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel16_mc13_ ## MMX(uint8_t *dst, uint8_t *src,   \
+                                          int stride)                   \
+{                                                                       \
+    uint64_t half[16 * 2 + 17 * 2];                                     \
+    uint8_t * const halfH  = ((uint8_t*)half) + 256;                    \
+    uint8_t * const halfHV = ((uint8_t*)half);                          \
+    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16,        \
+                                                 stride, 17);           \
+    put ## RND ## pixels16_l2_ ## MMX(halfH, src, halfH, 16,            \
+                                      stride, 17);                      \
+    put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH,         \
+                                                 16, 16);               \
+    OPNAME ## pixels16_l2_ ## MMX(dst, halfH + 16, halfHV, stride,      \
+                                  16, 16);                              \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel16_mc33_ ## MMX(uint8_t *dst, uint8_t *src,   \
+                                          int stride)                   \
+{                                                                       \
+    uint64_t half[16 * 2 + 17 * 2];                                     \
+    uint8_t * const halfH  = ((uint8_t*)half) + 256;                    \
+    uint8_t * const halfHV = ((uint8_t*)half);                          \
+    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16,        \
+                                                 stride, 17);           \
+    put ## RND ## pixels16_l2_ ## MMX(halfH, src + 1, halfH, 16,        \
+                                      stride, 17);                      \
+    put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH,         \
+                                                 16, 16);               \
+    OPNAME ## pixels16_l2_ ## MMX(dst, halfH + 16, halfHV, stride,      \
+                                  16, 16);                              \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel16_mc21_ ## MMX(uint8_t *dst, uint8_t *src,   \
+                                          int stride)                   \
+{                                                                       \
+    uint64_t half[16 * 2 + 17 * 2];                                     \
+    uint8_t * const halfH  = ((uint8_t*)half) + 256;                    \
+    uint8_t * const halfHV = ((uint8_t*)half);                          \
+    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16,        \
+                                                 stride, 17);           \
+    put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH,         \
+                                                 16, 16);               \
+    OPNAME ## pixels16_l2_ ## MMX(dst, halfH, halfHV, stride, 16, 16);  \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel16_mc23_ ## MMX(uint8_t *dst, uint8_t *src,   \
+                                          int stride)                   \
+{                                                                       \
+    uint64_t half[16 * 2 + 17 * 2];                                     \
+    uint8_t * const halfH  = ((uint8_t*)half) + 256;                    \
+    uint8_t * const halfHV = ((uint8_t*)half);                          \
+    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16,        \
+                                                 stride, 17);           \
+    put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH,         \
+                                                 16, 16);               \
+    OPNAME ## pixels16_l2_ ## MMX(dst, halfH + 16, halfHV, stride,      \
+                                  16, 16);                              \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel16_mc12_ ## MMX(uint8_t *dst, uint8_t *src,   \
+                                          int stride)                   \
+{                                                                       \
+    uint64_t half[17 * 2];                                              \
+    uint8_t * const halfH = ((uint8_t*)half);                           \
+    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16,        \
+                                                 stride, 17);           \
+    put ## RND ## pixels16_l2_ ## MMX(halfH, src, halfH, 16,            \
+                                      stride, 17);                      \
+    OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(dst, halfH, stride, 16);   \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel16_mc32_ ## MMX(uint8_t *dst, uint8_t *src,   \
+                                          int stride)                   \
+{                                                                       \
+    uint64_t half[17 * 2];                                              \
+    uint8_t * const halfH = ((uint8_t*)half);                           \
+    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16,        \
+                                                 stride, 17);           \
+    put ## RND ## pixels16_l2_ ## MMX(halfH, src + 1, halfH, 16,        \
+                                      stride, 17);                      \
+    OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(dst, halfH, stride, 16);   \
+}                                                                       \
+                                                                        \
+static void OPNAME ## qpel16_mc22_ ## MMX(uint8_t *dst, uint8_t *src,   \
+                                          int stride)                   \
+{                                                                       \
+    uint64_t half[17 * 2];                                              \
+    uint8_t * const halfH = ((uint8_t*)half);                           \
+    put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16,        \
+                                                 stride, 17);           \
+    OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(dst, halfH, stride, 16);   \
+}
+
+#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"   \
+    "mov"#size"        "#a", "#b"       \n\t"
+
+#define AVG_MMX2_OP(a, b, temp, size)           \
+    "mov"#size"        "#b", "#temp"    \n\t"   \
+    "pavgb          "#temp", "#a"       \n\t"   \
+    "mov"#size"        "#a", "#b"       \n\t"
+
+QPEL_BASE(put_,        ff_pw_16, _,        PUT_OP,       PUT_OP)
+QPEL_BASE(avg_,        ff_pw_16, _,        AVG_MMX2_OP,  AVG_3DNOW_OP)
+QPEL_BASE(put_no_rnd_, ff_pw_15, _no_rnd_, PUT_OP,       PUT_OP)
+QPEL_OP(put_,          ff_pw_16, _,        PUT_OP,       3dnow)
+QPEL_OP(avg_,          ff_pw_16, _,        AVG_3DNOW_OP, 3dnow)
+QPEL_OP(put_no_rnd_,   ff_pw_15, _no_rnd_, PUT_OP,       3dnow)
+QPEL_OP(put_,          ff_pw_16, _,        PUT_OP,       mmx2)
+QPEL_OP(avg_,          ff_pw_16, _,        AVG_MMX2_OP,  mmx2)
+QPEL_OP(put_no_rnd_,   ff_pw_15, _no_rnd_, PUT_OP,       mmx2)
 
 /***********************************/
 /* bilinear qpel: not compliant to any spec, only for -lavdopts fast */
 
-#define QPEL_2TAP_XY(OPNAME, SIZE, MMX, XY, HPEL)\
-static void OPNAME ## 2tap_qpel ## SIZE ## _mc ## XY ## _ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    OPNAME ## pixels ## SIZE ## HPEL(dst, src, stride, SIZE);\
-}
-#define QPEL_2TAP_L3(OPNAME, SIZE, MMX, XY, S0, S1, S2)\
-static void OPNAME ## 2tap_qpel ## SIZE ## _mc ## XY ## _ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    OPNAME ## 2tap_qpel ## SIZE ## _l3_ ## MMX(dst, src+S0, stride, SIZE, S1, S2);\
-}
-
-#define QPEL_2TAP(OPNAME, SIZE, MMX)\
-QPEL_2TAP_XY(OPNAME, SIZE, MMX, 20, _x2_ ## MMX)\
-QPEL_2TAP_XY(OPNAME, SIZE, MMX, 02, _y2_ ## MMX)\
-QPEL_2TAP_XY(OPNAME, SIZE, MMX, 22, _xy2_mmx)\
-static const qpel_mc_func OPNAME ## 2tap_qpel ## SIZE ## _mc00_ ## MMX =\
-                          OPNAME ## qpel ## SIZE ## _mc00_ ## MMX;\
-static const qpel_mc_func OPNAME ## 2tap_qpel ## SIZE ## _mc21_ ## MMX =\
-                          OPNAME ## 2tap_qpel ## SIZE ## _mc20_ ## MMX;\
-static const qpel_mc_func OPNAME ## 2tap_qpel ## SIZE ## _mc12_ ## MMX =\
-                          OPNAME ## 2tap_qpel ## SIZE ## _mc02_ ## MMX;\
-static void OPNAME ## 2tap_qpel ## SIZE ## _mc32_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    OPNAME ## pixels ## SIZE ## _y2_ ## MMX(dst, src+1, stride, SIZE);\
-}\
-static void OPNAME ## 2tap_qpel ## SIZE ## _mc23_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    OPNAME ## pixels ## SIZE ## _x2_ ## MMX(dst, src+stride, stride, SIZE);\
-}\
-QPEL_2TAP_L3(OPNAME, SIZE, MMX, 10, 0,         1,       0)\
-QPEL_2TAP_L3(OPNAME, SIZE, MMX, 30, 1,        -1,       0)\
-QPEL_2TAP_L3(OPNAME, SIZE, MMX, 01, 0,         stride,  0)\
-QPEL_2TAP_L3(OPNAME, SIZE, MMX, 03, stride,   -stride,  0)\
-QPEL_2TAP_L3(OPNAME, SIZE, MMX, 11, 0,         stride,  1)\
-QPEL_2TAP_L3(OPNAME, SIZE, MMX, 31, 1,         stride, -1)\
-QPEL_2TAP_L3(OPNAME, SIZE, MMX, 13, stride,   -stride,  1)\
-QPEL_2TAP_L3(OPNAME, SIZE, MMX, 33, stride+1, -stride, -1)\
+#define QPEL_2TAP_XY(OPNAME, SIZE, MMX, XY, HPEL)                              \
+static void OPNAME ## 2tap_qpel ## SIZE ## _mc ## XY ## _ ## MMX(uint8_t *dst, \
+                                                                 uint8_t *src, \
+                                                                 int stride)   \
+{                                                                              \
+    OPNAME ## pixels ## SIZE ## HPEL(dst, src, stride, SIZE);                  \
+}
+
+#define QPEL_2TAP_L3(OPNAME, SIZE, MMX, XY, S0, S1, S2)                        \
+static void OPNAME ## 2tap_qpel ## SIZE ## _mc ## XY ## _ ## MMX(uint8_t *dst, \
+                                                                 uint8_t *src, \
+                                                                 int stride)   \
+{                                                                              \
+    OPNAME ## 2tap_qpel ## SIZE ## _l3_ ## MMX(dst, src + S0, stride, SIZE,    \
+                                               S1, S2);                        \
+}
+
+#define QPEL_2TAP(OPNAME, SIZE, MMX)                                        \
+QPEL_2TAP_XY(OPNAME, SIZE, MMX, 20, _x2_ ## MMX)                            \
+QPEL_2TAP_XY(OPNAME, SIZE, MMX, 02, _y2_ ## MMX)                            \
+QPEL_2TAP_XY(OPNAME, SIZE, MMX, 22, _xy2_mmx)                               \
+static const qpel_mc_func OPNAME ## 2tap_qpel ## SIZE ## _mc00_ ## MMX =    \
+    OPNAME ## qpel ## SIZE ## _mc00_ ## MMX;                                \
+static const qpel_mc_func OPNAME ## 2tap_qpel ## SIZE ## _mc21_ ## MMX =    \
+    OPNAME ## 2tap_qpel ## SIZE ## _mc20_ ## MMX;                           \
+static const qpel_mc_func OPNAME ## 2tap_qpel ## SIZE ## _mc12_ ## MMX =    \
+    OPNAME ## 2tap_qpel ## SIZE ## _mc02_ ## MMX;                           \
+static void OPNAME ## 2tap_qpel ## SIZE ## _mc32_ ## MMX(uint8_t *dst,      \
+                                                         uint8_t *src,      \
+                                                         int stride)        \
+{                                                                           \
+    OPNAME ## pixels ## SIZE ## _y2_ ## MMX(dst, src + 1, stride, SIZE);    \
+}                                                                           \
+static void OPNAME ## 2tap_qpel ## SIZE ## _mc23_ ## MMX(uint8_t *dst,      \
+                                                         uint8_t *src,      \
+                                                         int stride)        \
+{                                                                           \
+    OPNAME ## pixels ## SIZE ## _x2_ ## MMX(dst, src + stride,              \
+                                            stride, SIZE);                  \
+}                                                                           \
+QPEL_2TAP_L3(OPNAME, SIZE, MMX, 10, 0,           1,       0)                \
+QPEL_2TAP_L3(OPNAME, SIZE, MMX, 30, 1,          -1,       0)                \
+QPEL_2TAP_L3(OPNAME, SIZE, MMX, 01, 0,           stride,  0)                \
+QPEL_2TAP_L3(OPNAME, SIZE, MMX, 03, stride,     -stride,  0)                \
+QPEL_2TAP_L3(OPNAME, SIZE, MMX, 11, 0,           stride,  1)                \
+QPEL_2TAP_L3(OPNAME, SIZE, MMX, 31, 1,           stride, -1)                \
+QPEL_2TAP_L3(OPNAME, SIZE, MMX, 13, stride,     -stride,  1)                \
+QPEL_2TAP_L3(OPNAME, SIZE, MMX, 33, stride + 1, -stride, -1)                \
 
 QPEL_2TAP(put_, 16, mmx2)
 QPEL_2TAP(avg_, 16, mmx2)
@@ -1573,265 +1794,276 @@ QPEL_2TAP(avg_,  8, 3dnow)
 
 
 #if HAVE_YASM
-typedef void emu_edge_core_func (uint8_t *buf, const uint8_t *src,
-                                 x86_reg linesize, x86_reg start_y,
-                                 x86_reg end_y, x86_reg block_h,
-                                 x86_reg start_x, x86_reg end_x,
-                                 x86_reg block_w);
+typedef void emu_edge_core_func(uint8_t *buf, const uint8_t *src,
+                                x86_reg linesize, x86_reg start_y,
+                                x86_reg end_y, x86_reg block_h,
+                                x86_reg start_x, x86_reg end_x,
+                                x86_reg block_w);
 extern emu_edge_core_func ff_emu_edge_core_mmx;
 extern emu_edge_core_func ff_emu_edge_core_sse;
 
-static av_always_inline
-void emulated_edge_mc(uint8_t *buf, const uint8_t *src, int linesize,
-                      int block_w, int block_h,
-                      int src_x, int src_y, int w, int h,
-                      emu_edge_core_func *core_fn)
+static av_always_inline void emulated_edge_mc(uint8_t *buf, const uint8_t *src,
+                                              int linesize,
+                                              int block_w, int block_h,
+                                              int src_x, int src_y,
+                                              int w, int h,
+                                              emu_edge_core_func *core_fn)
 {
-    int start_y, start_x, end_y, end_x, src_y_add=0;
-
-    if(src_y>= h){
-        src_y_add = h-1-src_y;
-        src_y=h-1;
-    }else if(src_y<=-block_h){
-        src_y_add = 1-block_h-src_y;
-        src_y=1-block_h;
+    int start_y, start_x, end_y, end_x, src_y_add = 0;
+
+    if (src_y >= h) {
+        src_y_add = h - 1 - src_y;
+        src_y     = h - 1;
+    } else if (src_y <= -block_h) {
+        src_y_add = 1 - block_h - src_y;
+        src_y     = 1 - block_h;
     }
-    if(src_x>= w){
-        src+= (w-1-src_x);
-        src_x=w-1;
-    }else if(src_x<=-block_w){
-        src+= (1-block_w-src_x);
-        src_x=1-block_w;
+    if (src_x >= w) {
+        src   += w - 1 - src_x;
+        src_x  = w - 1;
+    } else if (src_x <= -block_w) {
+        src   += 1 - block_w - src_x;
+        src_x  = 1 - block_w;
     }
 
-    start_y= FFMAX(0, -src_y);
-    start_x= FFMAX(0, -src_x);
-    end_y= FFMIN(block_h, h-src_y);
-    end_x= FFMIN(block_w, w-src_x);
+    start_y = FFMAX(0, -src_y);
+    start_x = FFMAX(0, -src_x);
+    end_y   = FFMIN(block_h, h-src_y);
+    end_x   = FFMIN(block_w, w-src_x);
     assert(start_x < end_x && block_w > 0);
     assert(start_y < end_y && block_h > 0);
 
     // fill in the to-be-copied part plus all above/below
-    src += (src_y_add+start_y)*linesize + start_x;
+    src += (src_y_add + start_y) * linesize + start_x;
     buf += start_x;
-    core_fn(buf, src, linesize, start_y, end_y, block_h, start_x, end_x, block_w);
+    core_fn(buf, src, linesize, start_y, end_y,
+            block_h, start_x, end_x, block_w);
 }
 
 #if ARCH_X86_32
-static av_noinline
-void emulated_edge_mc_mmx(uint8_t *buf, const uint8_t *src, int linesize,
-                          int block_w, int block_h,
-                          int src_x, int src_y, int w, int h)
+static av_noinline void emulated_edge_mc_mmx(uint8_t *buf, const uint8_t *src,
+                                             int linesize,
+                                             int block_w, int block_h,
+                                             int src_x, int src_y, int w, int h)
 {
     emulated_edge_mc(buf, src, linesize, block_w, block_h, src_x, src_y,
                      w, h, &ff_emu_edge_core_mmx);
 }
 #endif
-static av_noinline
-void emulated_edge_mc_sse(uint8_t *buf, const uint8_t *src, int linesize,
-                          int block_w, int block_h,
-                          int src_x, int src_y, int w, int h)
+
+static av_noinline void emulated_edge_mc_sse(uint8_t *buf, const uint8_t *src,
+                                             int linesize,
+                                             int block_w, int block_h,
+                                             int src_x, int src_y, int w, int h)
 {
     emulated_edge_mc(buf, src, linesize, block_w, block_h, src_x, src_y,
                      w, h, &ff_emu_edge_core_sse);
 }
 #endif /* HAVE_YASM */
 
-typedef void emulated_edge_mc_func (uint8_t *dst, const uint8_t *src,
-                                    int linesize, int block_w, int block_h,
-                                    int src_x, int src_y, int w, int h);
-
-static av_always_inline
-void gmc(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,
-         emulated_edge_mc_func *emu_edge_fn)
-{
-    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];
+typedef void emulated_edge_mc_func(uint8_t *dst, const uint8_t *src,
+                                   int linesize, int block_w, int block_h,
+                                   int src_x, int src_y, int w, int h);
+
+static av_always_inline void gmc(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,
+                                 emulated_edge_mc_func *emu_edge_fn)
+{
+    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)
+    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);
+        || (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 )
-    {
-        emu_edge_fn(edge_buf, src, stride, w+1, h+1, ix, iy, width, height);
+    src += ix + iy * stride;
+    if ((unsigned)ix >= width  - w ||
+        (unsigned)iy >= height - h) {
+        emu_edge_fn(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"
+    __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"
+    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"
+            __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
+                "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"
+                "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])
+                "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"(src[stride]), "m"(src[stride + 1]),
                   "m"(*r4), "m"(shift2)
             );
             src += stride;
         }
-        src += 4-h*stride;
+        src += 4 - h * stride;
     }
 }
 
 #if HAVE_YASM
 #if ARCH_X86_32
-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)
+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)
 {
     gmc(dst, src, stride, h, ox, oy, dxx, dxy, dyx, dyy, shift, r,
         width, height, &emulated_edge_mc_mmx);
 }
 #endif
-static void gmc_sse(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)
+static void gmc_sse(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)
 {
     gmc(dst, src, stride, h, ox, oy, dxx, dxy, dyx, dyy, shift, r,
         width, height, &emulated_edge_mc_sse);
 }
 #else
-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)
+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)
 {
     gmc(dst, src, stride, h, ox, oy, dxx, dxy, dyx, dyy, shift, r,
         width, height, &ff_emulated_edge_mc_8);
 }
 #endif
 
-#define PREFETCH(name, op) \
-static void name(void *mem, int stride, int h){\
-    const uint8_t *p= mem;\
-    do{\
-        __asm__ volatile(#op" %0" :: "m"(*p));\
-        p+= stride;\
-    }while(--h);\
+#define PREFETCH(name, op)                      \
+static 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 "h264_qpel_mmx.c"
 
-void ff_put_h264_chroma_mc8_mmx_rnd   (uint8_t *dst, uint8_t *src,
-                                       int stride, int h, int x, int y);
-void ff_avg_h264_chroma_mc8_mmx2_rnd  (uint8_t *dst, uint8_t *src,
-                                       int stride, int h, int x, int y);
-void ff_avg_h264_chroma_mc8_3dnow_rnd (uint8_t *dst, uint8_t *src,
-                                       int stride, int h, int x, int y);
-
-void ff_put_h264_chroma_mc4_mmx       (uint8_t *dst, uint8_t *src,
-                                       int stride, int h, int x, int y);
-void ff_avg_h264_chroma_mc4_mmx2      (uint8_t *dst, uint8_t *src,
-                                       int stride, int h, int x, int y);
-void ff_avg_h264_chroma_mc4_3dnow     (uint8_t *dst, uint8_t *src,
-                                       int stride, int h, int x, int y);
-
-void ff_put_h264_chroma_mc2_mmx2      (uint8_t *dst, uint8_t *src,
-                                       int stride, int h, int x, int y);
-void ff_avg_h264_chroma_mc2_mmx2      (uint8_t *dst, uint8_t *src,
-                                       int stride, int h, int x, int y);
-
-void ff_put_h264_chroma_mc8_ssse3_rnd (uint8_t *dst, uint8_t *src,
-                                       int stride, int h, int x, int y);
-void ff_put_h264_chroma_mc4_ssse3     (uint8_t *dst, uint8_t *src,
-                                       int stride, int h, int x, int y);
-
-void ff_avg_h264_chroma_mc8_ssse3_rnd (uint8_t *dst, uint8_t *src,
-                                       int stride, int h, int x, int y);
-void ff_avg_h264_chroma_mc4_ssse3     (uint8_t *dst, uint8_t *src,
-                                       int stride, int h, int x, int y);
-
-#define CHROMA_MC(OP, NUM, DEPTH, OPT) \
-void ff_ ## OP ## _h264_chroma_mc ## NUM ## _ ## DEPTH ## _ ## OPT \
-                                      (uint8_t *dst, uint8_t *src,\
+void ff_put_h264_chroma_mc8_mmx_rnd  (uint8_t *dst, uint8_t *src,
+                                      int stride, int h, int x, int y);
+void ff_avg_h264_chroma_mc8_mmx2_rnd (uint8_t *dst, uint8_t *src,
+                                      int stride, int h, int x, int y);
+void ff_avg_h264_chroma_mc8_3dnow_rnd(uint8_t *dst, uint8_t *src,
+                                      int stride, int h, int x, int y);
+
+void ff_put_h264_chroma_mc4_mmx      (uint8_t *dst, uint8_t *src,
+                                      int stride, int h, int x, int y);
+void ff_avg_h264_chroma_mc4_mmx2     (uint8_t *dst, uint8_t *src,
+                                      int stride, int h, int x, int y);
+void ff_avg_h264_chroma_mc4_3dnow    (uint8_t *dst, uint8_t *src,
+                                      int stride, int h, int x, int y);
+
+void ff_put_h264_chroma_mc2_mmx2     (uint8_t *dst, uint8_t *src,
+                                      int stride, int h, int x, int y);
+void ff_avg_h264_chroma_mc2_mmx2     (uint8_t *dst, uint8_t *src,
+                                      int stride, int h, int x, int y);
+
+void ff_put_h264_chroma_mc8_ssse3_rnd(uint8_t *dst, uint8_t *src,
+                                      int stride, int h, int x, int y);
+void ff_put_h264_chroma_mc4_ssse3    (uint8_t *dst, uint8_t *src,
+                                      int stride, int h, int x, int y);
+
+void ff_avg_h264_chroma_mc8_ssse3_rnd(uint8_t *dst, uint8_t *src,
+                                      int stride, int h, int x, int y);
+void ff_avg_h264_chroma_mc4_ssse3    (uint8_t *dst, uint8_t *src,
+                                      int stride, int h, int x, int y);
+
+#define CHROMA_MC(OP, NUM, DEPTH, OPT)                                  \
+void ff_ ## OP ## _h264_chroma_mc ## NUM ## _ ## DEPTH ## _ ## OPT      \
+                                      (uint8_t *dst, uint8_t *src,      \
                                        int stride, int h, int x, int y);
 
 CHROMA_MC(put, 2, 10, mmxext)
@@ -1843,25 +2075,37 @@ CHROMA_MC(avg, 8, 10, sse2)
 CHROMA_MC(put, 8, 10, avx)
 CHROMA_MC(avg, 8, 10, avx)
 
-/* CAVS specific */
-void ff_put_cavs_qpel8_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride) {
+/* CAVS-specific */
+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) {
+
+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) {
+
+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) {
+
+void ff_avg_cavs_qpel16_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride)
+{
     avg_pixels16_mmx(dst, src, stride, 16);
 }
 
-/* VC1 specific */
-void ff_put_vc1_mspel_mc00_mmx(uint8_t *dst, const uint8_t *src, int stride, int rnd) {
+/* VC-1-specific */
+void ff_put_vc1_mspel_mc00_mmx(uint8_t *dst, const uint8_t *src,
+                               int stride, int rnd)
+{
     put_pixels8_mmx(dst, src, stride, 8);
 }
-void ff_avg_vc1_mspel_mc00_mmx2(uint8_t *dst, const uint8_t *src, int stride, int rnd) {
+
+void ff_avg_vc1_mspel_mc00_mmx2(uint8_t *dst, const uint8_t *src,
+                                int stride, int rnd)
+{
     avg_pixels8_mmx2(dst, src, stride, 8);
 }
 
@@ -1943,90 +2187,102 @@ void ff_avg_dirac_pixels32_sse2(uint8_t *dst, const uint8_t *src[5], int stride,
 }
 #endif
 
-/* XXX: those functions should be suppressed ASAP when all IDCTs are
  converted */
+/* XXX: Those functions should be suppressed ASAP when all IDCTs are
* converted. */
 #if CONFIG_GPL
-static void ff_libmpeg2mmx_idct_put(uint8_t *dest, int line_size, DCTELEM *block)
+static void ff_libmpeg2mmx_idct_put(uint8_t *dest, int line_size,
+                                    DCTELEM *block)
 {
-    ff_mmx_idct (block);
+    ff_mmx_idct(block);
     ff_put_pixels_clamped_mmx(block, dest, line_size);
 }
-static void ff_libmpeg2mmx_idct_add(uint8_t *dest, int line_size, DCTELEM *block)
+
+static void ff_libmpeg2mmx_idct_add(uint8_t *dest, int line_size,
+                                    DCTELEM *block)
 {
-    ff_mmx_idct (block);
+    ff_mmx_idct(block);
     ff_add_pixels_clamped_mmx(block, dest, line_size);
 }
-static void ff_libmpeg2mmx2_idct_put(uint8_t *dest, int line_size, DCTELEM *block)
+
+static void ff_libmpeg2mmx2_idct_put(uint8_t *dest, int line_size,
+                                     DCTELEM *block)
 {
-    ff_mmxext_idct (block);
+    ff_mmxext_idct(block);
     ff_put_pixels_clamped_mmx(block, dest, line_size);
 }
-static void ff_libmpeg2mmx2_idct_add(uint8_t *dest, int line_size, DCTELEM *block)
+
+static void ff_libmpeg2mmx2_idct_add(uint8_t *dest, int line_size,
+                                     DCTELEM *block)
 {
-    ff_mmxext_idct (block);
+    ff_mmxext_idct(block);
     ff_add_pixels_clamped_mmx(block, dest, line_size);
 }
 #endif
+
 static void ff_idct_xvid_mmx_put(uint8_t *dest, int line_size, DCTELEM *block)
 {
-    ff_idct_xvid_mmx (block);
+    ff_idct_xvid_mmx(block);
     ff_put_pixels_clamped_mmx(block, dest, line_size);
 }
+
 static void ff_idct_xvid_mmx_add(uint8_t *dest, int line_size, DCTELEM *block)
 {
-    ff_idct_xvid_mmx (block);
+    ff_idct_xvid_mmx(block);
     ff_add_pixels_clamped_mmx(block, dest, line_size);
 }
+
 static void ff_idct_xvid_mmx2_put(uint8_t *dest, int line_size, DCTELEM *block)
 {
-    ff_idct_xvid_mmx2 (block);
+    ff_idct_xvid_mmx2(block);
     ff_put_pixels_clamped_mmx(block, dest, line_size);
 }
+
 static void ff_idct_xvid_mmx2_add(uint8_t *dest, int line_size, DCTELEM *block)
 {
-    ff_idct_xvid_mmx2 (block);
+    ff_idct_xvid_mmx2(block);
     ff_add_pixels_clamped_mmx(block, dest, line_size);
 }
 
 static void vorbis_inverse_coupling_3dnow(float *mag, float *ang, int blocksize)
 {
     int i;
-    __asm__ volatile("pxor %%mm7, %%mm7":);
-    for(i=0; i<blocksize; i+=2) {
-        __asm__ volatile(
-            "movq    %0,    %%mm0 \n\t"
-            "movq    %1,    %%mm1 \n\t"
-            "movq    %%mm0, %%mm2 \n\t"
-            "movq    %%mm1, %%mm3 \n\t"
-            "pfcmpge %%mm7, %%mm2 \n\t" // m <= 0.0
-            "pfcmpge %%mm7, %%mm3 \n\t" // a <= 0.0
-            "pslld   $31,   %%mm2 \n\t" // keep only the sign bit
-            "pxor    %%mm2, %%mm1 \n\t"
-            "movq    %%mm3, %%mm4 \n\t"
-            "pand    %%mm1, %%mm3 \n\t"
-            "pandn   %%mm1, %%mm4 \n\t"
-            "pfadd   %%mm0, %%mm3 \n\t" // a = m + ((a<0) & (a ^ sign(m)))
-            "pfsub   %%mm4, %%mm0 \n\t" // m = m + ((a>0) & (a ^ sign(m)))
-            "movq    %%mm3, %1    \n\t"
-            "movq    %%mm0, %0    \n\t"
-            :"+m"(mag[i]), "+m"(ang[i])
-            ::"memory"
+    __asm__ volatile ("pxor %%mm7, %%mm7":);
+    for (i = 0; i < blocksize; i += 2) {
+        __asm__ volatile (
+            "movq       %0, %%mm0   \n\t"
+            "movq       %1, %%mm1   \n\t"
+            "movq    %%mm0, %%mm2   \n\t"
+            "movq    %%mm1, %%mm3   \n\t"
+            "pfcmpge %%mm7, %%mm2   \n\t" // m <= 0.0
+            "pfcmpge %%mm7, %%mm3   \n\t" // a <= 0.0
+            "pslld     $31, %%mm2   \n\t" // keep only the sign bit
+            "pxor    %%mm2, %%mm1   \n\t"
+            "movq    %%mm3, %%mm4   \n\t"
+            "pand    %%mm1, %%mm3   \n\t"
+            "pandn   %%mm1, %%mm4   \n\t"
+            "pfadd   %%mm0, %%mm3   \n\t" // a = m + ((a < 0) & (a ^ sign(m)))
+            "pfsub   %%mm4, %%mm0   \n\t" // m = m + ((a > 0) & (a ^ sign(m)))
+            "movq    %%mm3, %1      \n\t"
+            "movq    %%mm0, %0      \n\t"
+            : "+m"(mag[i]), "+m"(ang[i])
+            :: "memory"
         );
     }
-    __asm__ volatile("femms");
+    __asm__ volatile ("femms");
 }
+
 static void vorbis_inverse_coupling_sse(float *mag, float *ang, int blocksize)
 {
     int i;
 
-    __asm__ volatile(
-            "movaps  %0,     %%xmm5 \n\t"
-        ::"m"(ff_pdw_80000000[0])
+    __asm__ volatile (
+        "movaps  %0, %%xmm5 \n\t"
+        :: "m"(ff_pdw_80000000[0])
     );
-    for(i=0; i<blocksize; i+=4) {
-        __asm__ volatile(
-            "movaps  %0,     %%xmm0 \n\t"
-            "movaps  %1,     %%xmm1 \n\t"
+    for (i = 0; i < blocksize; i += 4) {
+        __asm__ volatile (
+            "movaps      %0, %%xmm0 \n\t"
+            "movaps      %1, %%xmm1 \n\t"
             "xorps   %%xmm2, %%xmm2 \n\t"
             "xorps   %%xmm3, %%xmm3 \n\t"
             "cmpleps %%xmm0, %%xmm2 \n\t" // m <= 0.0
@@ -2036,12 +2292,12 @@ static void vorbis_inverse_coupling_sse(float *mag, float *ang, int blocksize)
             "movaps  %%xmm3, %%xmm4 \n\t"
             "andps   %%xmm1, %%xmm3 \n\t"
             "andnps  %%xmm1, %%xmm4 \n\t"
-            "addps   %%xmm0, %%xmm3 \n\t" // a = m + ((a<0) & (a ^ sign(m)))
-            "subps   %%xmm4, %%xmm0 \n\t" // m = m + ((a>0) & (a ^ sign(m)))
+            "addps   %%xmm0, %%xmm3 \n\t" // a = m + ((a < 0) & (a ^ sign(m)))
+            "subps   %%xmm4, %%xmm0 \n\t" // m = m + ((a > 0) & (a ^ sign(m)))
             "movaps  %%xmm3, %1     \n\t"
             "movaps  %%xmm0, %0     \n\t"
-            :"+m"(mag[i]), "+m"(ang[i])
-            ::"memory"
+            : "+m"(mag[i]), "+m"(ang[i])
+            :: "memory"
         );
     }
 }
@@ -2049,97 +2305,105 @@ static void vorbis_inverse_coupling_sse(float *mag, float *ang, int blocksize)
 #define IF1(x) x
 #define IF0(x)
 
-#define MIX5(mono,stereo)\
-    __asm__ volatile(\
-        "movss          0(%2), %%xmm5 \n"\
-        "movss          8(%2), %%xmm6 \n"\
-        "movss         24(%2), %%xmm7 \n"\
-        "shufps    $0, %%xmm5, %%xmm5 \n"\
-        "shufps    $0, %%xmm6, %%xmm6 \n"\
-        "shufps    $0, %%xmm7, %%xmm7 \n"\
-        "1: \n"\
-        "movaps       (%0,%1), %%xmm0 \n"\
-        "movaps  0x400(%0,%1), %%xmm1 \n"\
-        "movaps  0x800(%0,%1), %%xmm2 \n"\
-        "movaps  0xc00(%0,%1), %%xmm3 \n"\
-        "movaps 0x1000(%0,%1), %%xmm4 \n"\
-        "mulps         %%xmm5, %%xmm0 \n"\
-        "mulps         %%xmm6, %%xmm1 \n"\
-        "mulps         %%xmm5, %%xmm2 \n"\
-        "mulps         %%xmm7, %%xmm3 \n"\
-        "mulps         %%xmm7, %%xmm4 \n"\
- stereo("addps         %%xmm1, %%xmm0 \n")\
-        "addps         %%xmm1, %%xmm2 \n"\
-        "addps         %%xmm3, %%xmm0 \n"\
-        "addps         %%xmm4, %%xmm2 \n"\
-   mono("addps         %%xmm2, %%xmm0 \n")\
-        "movaps  %%xmm0,      (%0,%1) \n"\
- stereo("movaps  %%xmm2, 0x400(%0,%1) \n")\
-        "add $16, %0 \n"\
-        "jl 1b \n"\
-        :"+&r"(i)\
-        :"r"(samples[0]+len), "r"(matrix)\
-        :XMM_CLOBBERS("%xmm0", "%xmm1", "%xmm2", "%xmm3", \
-                      "%xmm4", "%xmm5", "%xmm6", "%xmm7",)\
-         "memory"\
+#define MIX5(mono, stereo)                                      \
+    __asm__ volatile (                                          \
+        "movss           0(%2), %%xmm5          \n"             \
+        "movss           8(%2), %%xmm6          \n"             \
+        "movss          24(%2), %%xmm7          \n"             \
+        "shufps     $0, %%xmm5, %%xmm5          \n"             \
+        "shufps     $0, %%xmm6, %%xmm6          \n"             \
+        "shufps     $0, %%xmm7, %%xmm7          \n"             \
+        "1:                                     \n"             \
+        "movaps       (%0, %1), %%xmm0          \n"             \
+        "movaps  0x400(%0, %1), %%xmm1          \n"             \
+        "movaps  0x800(%0, %1), %%xmm2          \n"             \
+        "movaps  0xc00(%0, %1), %%xmm3          \n"             \
+        "movaps 0x1000(%0, %1), %%xmm4          \n"             \
+        "mulps          %%xmm5, %%xmm0          \n"             \
+        "mulps          %%xmm6, %%xmm1          \n"             \
+        "mulps          %%xmm5, %%xmm2          \n"             \
+        "mulps          %%xmm7, %%xmm3          \n"             \
+        "mulps          %%xmm7, %%xmm4          \n"             \
+ stereo("addps          %%xmm1, %%xmm0          \n")            \
+        "addps          %%xmm1, %%xmm2          \n"             \
+        "addps          %%xmm3, %%xmm0          \n"             \
+        "addps          %%xmm4, %%xmm2          \n"             \
+   mono("addps          %%xmm2, %%xmm0          \n")            \
+        "movaps         %%xmm0, (%0, %1)        \n"             \
+ stereo("movaps         %%xmm2, 0x400(%0, %1)   \n")            \
+        "add               $16, %0              \n"             \
+        "jl                 1b                  \n"             \
+        : "+&r"(i)                                              \
+        : "r"(samples[0] + len), "r"(matrix)                    \
+        : XMM_CLOBBERS("%xmm0", "%xmm1", "%xmm2", "%xmm3",      \
+                      "%xmm4", "%xmm5", "%xmm6", "%xmm7",)      \
+         "memory"                                               \
     );
 
-#define MIX_MISC(stereo)\
-    __asm__ volatile(\
-        "1: \n"\
-        "movaps  (%3,%0), %%xmm0 \n"\
- stereo("movaps   %%xmm0, %%xmm1 \n")\
-        "mulps    %%xmm4, %%xmm0 \n"\
- stereo("mulps    %%xmm5, %%xmm1 \n")\
-        "lea 1024(%3,%0), %1 \n"\
-        "mov %5, %2 \n"\
-        "2: \n"\
-        "movaps   (%1),   %%xmm2 \n"\
- stereo("movaps   %%xmm2, %%xmm3 \n")\
-        "mulps   (%4,%2), %%xmm2 \n"\
- stereo("mulps 16(%4,%2), %%xmm3 \n")\
-        "addps    %%xmm2, %%xmm0 \n"\
- stereo("addps    %%xmm3, %%xmm1 \n")\
-        "add $1024, %1 \n"\
-        "add $32, %2 \n"\
-        "jl 2b \n"\
-        "movaps   %%xmm0,     (%3,%0) \n"\
- stereo("movaps   %%xmm1, 1024(%3,%0) \n")\
-        "add $16, %0 \n"\
-        "jl 1b \n"\
-        :"+&r"(i), "=&r"(j), "=&r"(k)\
-        :"r"(samples[0]+len), "r"(matrix_simd+in_ch), "g"((intptr_t)-32*(in_ch-1))\
-        :"memory"\
+#define MIX_MISC(stereo)                                        \
+    __asm__ volatile (                                          \
+        "1:                                 \n"                 \
+        "movaps     (%3, %0), %%xmm0        \n"                 \
+ stereo("movaps       %%xmm0, %%xmm1        \n")                \
+        "mulps        %%xmm4, %%xmm0        \n"                 \
+ stereo("mulps        %%xmm5, %%xmm1        \n")                \
+        "lea    1024(%3, %0), %1            \n"                 \
+        "mov              %5, %2            \n"                 \
+        "2:                                 \n"                 \
+        "movaps         (%1), %%xmm2        \n"                 \
+ stereo("movaps       %%xmm2, %%xmm3        \n")                \
+        "mulps      (%4, %2), %%xmm2        \n"                 \
+ stereo("mulps    16(%4, %2), %%xmm3        \n")                \
+        "addps        %%xmm2, %%xmm0        \n"                 \
+ stereo("addps        %%xmm3, %%xmm1        \n")                \
+        "add           $1024, %1            \n"                 \
+        "add             $32, %2            \n"                 \
+        "jl               2b                \n"                 \
+        "movaps       %%xmm0,     (%3, %0)  \n"                 \
+ stereo("movaps       %%xmm1, 1024(%3, %0)  \n")                \
+        "add             $16, %0            \n"                 \
+        "jl               1b                \n"                 \
+        : "+&r"(i), "=&r"(j), "=&r"(k)                          \
+        : "r"(samples[0] + len), "r"(matrix_simd + in_ch),      \
+          "g"((intptr_t) - 32 * (in_ch - 1))                    \
+        : "memory"                                              \
     );
 
-static void ac3_downmix_sse(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len)
+static void ac3_downmix_sse(float (*samples)[256], float (*matrix)[2],
+                            int out_ch, int in_ch, int len)
 {
     int (*matrix_cmp)[2] = (int(*)[2])matrix;
-    intptr_t i,j,k;
-
-    i = -len*sizeof(float);
-    if(in_ch == 5 && out_ch == 2 && !(matrix_cmp[0][1]|matrix_cmp[2][0]|matrix_cmp[3][1]|matrix_cmp[4][0]|(matrix_cmp[1][0]^matrix_cmp[1][1])|(matrix_cmp[0][0]^matrix_cmp[2][1]))) {
-        MIX5(IF0,IF1);
-    } else if(in_ch == 5 && out_ch == 1 && matrix_cmp[0][0]==matrix_cmp[2][0] && matrix_cmp[3][0]==matrix_cmp[4][0]) {
-        MIX5(IF1,IF0);
+    intptr_t i, j, k;
+
+    i = -len * sizeof(float);
+    if (in_ch == 5 && out_ch == 2 &&
+        !(matrix_cmp[0][1] | matrix_cmp[2][0]   |
+          matrix_cmp[3][1] | matrix_cmp[4][0]   |
+          (matrix_cmp[1][0] ^ matrix_cmp[1][1]) |
+          (matrix_cmp[0][0] ^ matrix_cmp[2][1]))) {
+        MIX5(IF0, IF1);
+    } else if (in_ch == 5 && out_ch == 1 &&
+               matrix_cmp[0][0] == matrix_cmp[2][0] &&
+               matrix_cmp[3][0] == matrix_cmp[4][0]) {
+        MIX5(IF1, IF0);
     } else {
         DECLARE_ALIGNED(16, float, matrix_simd)[AC3_MAX_CHANNELS][2][4];
-        j = 2*in_ch*sizeof(float);
-        __asm__ volatile(
-            "1: \n"
-            "sub $8, %0 \n"
-            "movss     (%2,%0), %%xmm4 \n"
-            "movss    4(%2,%0), %%xmm5 \n"
-            "shufps $0, %%xmm4, %%xmm4 \n"
-            "shufps $0, %%xmm5, %%xmm5 \n"
-            "movaps %%xmm4,   (%1,%0,4) \n"
-            "movaps %%xmm5, 16(%1,%0,4) \n"
-            "jg 1b \n"
-            :"+&r"(j)
-            :"r"(matrix_simd), "r"(matrix)
-            :"memory"
+        j = 2 * in_ch * sizeof(float);
+        __asm__ volatile (
+            "1:                                 \n"
+            "sub             $8, %0             \n"
+            "movss     (%2, %0), %%xmm4         \n"
+            "movss    4(%2, %0), %%xmm5         \n"
+            "shufps          $0, %%xmm4, %%xmm4 \n"
+            "shufps          $0, %%xmm5, %%xmm5 \n"
+            "movaps      %%xmm4,   (%1, %0, 4)  \n"
+            "movaps      %%xmm5, 16(%1, %0, 4)  \n"
+            "jg              1b                 \n"
+            : "+&r"(j)
+            : "r"(matrix_simd), "r"(matrix)
+            : "memory"
         );
-        if(out_ch == 2) {
+        if (out_ch == 2) {
             MIX_MISC(IF1);
         } else {
             MIX_MISC(IF0);
@@ -2147,216 +2411,232 @@ static void ac3_downmix_sse(float (*samples)[256], float (*matrix)[2], int out_c
     }
 }
 
-static void vector_fmul_3dnow(float *dst, const float *src0, const float *src1, int len){
-    x86_reg i = (len-4)*4;
-    __asm__ volatile(
-        "1: \n\t"
-        "movq    (%2,%0), %%mm0 \n\t"
-        "movq   8(%2,%0), %%mm1 \n\t"
-        "pfmul   (%3,%0), %%mm0 \n\t"
-        "pfmul  8(%3,%0), %%mm1 \n\t"
-        "movq   %%mm0,  (%1,%0) \n\t"
-        "movq   %%mm1, 8(%1,%0) \n\t"
-        "sub  $16, %0 \n\t"
-        "jge 1b \n\t"
-        "femms  \n\t"
-        :"+r"(i)
-        :"r"(dst), "r"(src0), "r"(src1)
-        :"memory"
+static void vector_fmul_3dnow(float *dst, const float *src0, const float *src1,
+                              int len)
+{
+    x86_reg i = (len - 4) * 4;
+    __asm__ volatile (
+        "1:                             \n\t"
+        "movq    (%2, %0), %%mm0        \n\t"
+        "movq   8(%2, %0), %%mm1        \n\t"
+        "pfmul   (%3, %0), %%mm0        \n\t"
+        "pfmul  8(%3, %0), %%mm1        \n\t"
+        "movq       %%mm0,  (%1, %0)    \n\t"
+        "movq       %%mm1, 8(%1, %0)    \n\t"
+        "sub          $16, %0           \n\t"
+        "jge           1b               \n\t"
+        "femms                          \n\t"
+        : "+r"(i)
+        : "r"(dst), "r"(src0), "r"(src1)
+        : "memory"
     );
 }
-static void vector_fmul_sse(float *dst, const float *src0, const float *src1, int len){
-    x86_reg i = (len-8)*4;
-    __asm__ volatile(
-        "1: \n\t"
-        "movaps    (%2,%0), %%xmm0 \n\t"
-        "movaps  16(%2,%0), %%xmm1 \n\t"
-        "mulps     (%3,%0), %%xmm0 \n\t"
-        "mulps   16(%3,%0), %%xmm1 \n\t"
-        "movaps  %%xmm0,   (%1,%0) \n\t"
-        "movaps  %%xmm1, 16(%1,%0) \n\t"
-        "sub  $32, %0 \n\t"
-        "jge 1b \n\t"
-        :"+r"(i)
-        :"r"(dst), "r"(src0), "r"(src1)
-        :"memory"
+
+static void vector_fmul_sse(float *dst, const float *src0, const float *src1,
+                            int len)
+{
+    x86_reg i = (len - 8) * 4;
+    __asm__ volatile (
+        "1:                             \n\t"
+        "movaps    (%2, %0), %%xmm0     \n\t"
+        "movaps  16(%2, %0), %%xmm1     \n\t"
+        "mulps     (%3, %0), %%xmm0     \n\t"
+        "mulps   16(%3, %0), %%xmm1     \n\t"
+        "movaps      %%xmm0,   (%1, %0) \n\t"
+        "movaps      %%xmm1, 16(%1, %0) \n\t"
+        "sub            $32, %0         \n\t"
+        "jge             1b             \n\t"
+        : "+r"(i)
+        : "r"(dst), "r"(src0), "r"(src1)
+        : "memory"
     );
 }
 
-static void vector_fmul_reverse_3dnow2(float *dst, const float *src0, const float *src1, int len){
-    x86_reg i = len*4-16;
-    __asm__ volatile(
-        "1: \n\t"
-        "pswapd   8(%1), %%mm0 \n\t"
-        "pswapd    (%1), %%mm1 \n\t"
-        "pfmul  (%3,%0), %%mm0 \n\t"
-        "pfmul 8(%3,%0), %%mm1 \n\t"
-        "movq  %%mm0,  (%2,%0) \n\t"
-        "movq  %%mm1, 8(%2,%0) \n\t"
-        "add   $16, %1 \n\t"
-        "sub   $16, %0 \n\t"
-        "jge   1b \n\t"
-        :"+r"(i), "+r"(src1)
-        :"r"(dst), "r"(src0)
+static void vector_fmul_reverse_3dnow2(float *dst, const float *src0,
+                                       const float *src1, int len)
+{
+    x86_reg i = len * 4 - 16;
+    __asm__ volatile (
+        "1:                             \n\t"
+        "pswapd     8(%1), %%mm0        \n\t"
+        "pswapd      (%1), %%mm1        \n\t"
+        "pfmul   (%3, %0), %%mm0        \n\t"
+        "pfmul  8(%3, %0), %%mm1        \n\t"
+        "movq       %%mm0,  (%2, %0)    \n\t"
+        "movq       %%mm1, 8(%2, %0)    \n\t"
+        "add          $16, %1           \n\t"
+        "sub          $16, %0           \n\t"
+        "jge           1b               \n\t"
+        : "+r"(i), "+r"(src1)
+        : "r"(dst), "r"(src0)
     );
-    __asm__ volatile("femms");
+    __asm__ volatile ("femms");
 }
-static void vector_fmul_reverse_sse(float *dst, const float *src0, const float *src1, int len){
-    x86_reg i = len*4-32;
-    __asm__ volatile(
-        "1: \n\t"
-        "movaps        16(%1), %%xmm0 \n\t"
-        "movaps          (%1), %%xmm1 \n\t"
-        "shufps $0x1b, %%xmm0, %%xmm0 \n\t"
-        "shufps $0x1b, %%xmm1, %%xmm1 \n\t"
-        "mulps        (%3,%0), %%xmm0 \n\t"
-        "mulps      16(%3,%0), %%xmm1 \n\t"
-        "movaps     %%xmm0,   (%2,%0) \n\t"
-        "movaps     %%xmm1, 16(%2,%0) \n\t"
-        "add    $32, %1 \n\t"
-        "sub    $32, %0 \n\t"
-        "jge    1b \n\t"
-        :"+r"(i), "+r"(src1)
-        :"r"(dst), "r"(src0)
+
+static void vector_fmul_reverse_sse(float *dst, const float *src0,
+                                    const float *src1, int len)
+{
+    x86_reg i = len * 4 - 32;
+    __asm__ volatile (
+        "1:                                 \n\t"
+        "movaps         16(%1), %%xmm0      \n\t"
+        "movaps           (%1), %%xmm1      \n\t"
+        "shufps  $0x1b, %%xmm0, %%xmm0      \n\t"
+        "shufps  $0x1b, %%xmm1, %%xmm1      \n\t"
+        "mulps        (%3, %0), %%xmm0      \n\t"
+        "mulps      16(%3, %0), %%xmm1      \n\t"
+        "movaps         %%xmm0,   (%2, %0)  \n\t"
+        "movaps         %%xmm1, 16(%2, %0)  \n\t"
+        "add               $32, %1          \n\t"
+        "sub               $32, %0          \n\t"
+        "jge                1b              \n\t"
+        : "+r"(i), "+r"(src1)
+        : "r"(dst), "r"(src0)
     );
 }
 
-static void vector_fmul_add_3dnow(float *dst, const float *src0, const float *src1,
-                                  const float *src2, int len){
-    x86_reg i = (len-4)*4;
-    __asm__ volatile(
-        "1: \n\t"
-        "movq    (%2,%0), %%mm0 \n\t"
-        "movq   8(%2,%0), %%mm1 \n\t"
-        "pfmul   (%3,%0), %%mm0 \n\t"
-        "pfmul  8(%3,%0), %%mm1 \n\t"
-        "pfadd   (%4,%0), %%mm0 \n\t"
-        "pfadd  8(%4,%0), %%mm1 \n\t"
-        "movq  %%mm0,   (%1,%0) \n\t"
-        "movq  %%mm1,  8(%1,%0) \n\t"
-        "sub  $16, %0 \n\t"
-        "jge  1b \n\t"
-        :"+r"(i)
-        :"r"(dst), "r"(src0), "r"(src1), "r"(src2)
-        :"memory"
+static void vector_fmul_add_3dnow(float *dst, const float *src0,
+                                  const float *src1, const float *src2, int len)
+{
+    x86_reg i = (len - 4) * 4;
+    __asm__ volatile (
+        "1:                             \n\t"
+        "movq   (%2, %0), %%mm0         \n\t"
+        "movq  8(%2, %0), %%mm1         \n\t"
+        "pfmul  (%3, %0), %%mm0         \n\t"
+        "pfmul 8(%3, %0), %%mm1         \n\t"
+        "pfadd  (%4, %0), %%mm0         \n\t"
+        "pfadd 8(%4, %0), %%mm1         \n\t"
+        "movq      %%mm0,  (%1, %0)     \n\t"
+        "movq      %%mm1, 8(%1, %0)     \n\t"
+        "sub         $16, %0            \n\t"
+        "jge          1b                \n\t"
+        : "+r"(i)
+        : "r"(dst), "r"(src0), "r"(src1), "r"(src2)
+        : "memory"
     );
-    __asm__ volatile("femms");
+    __asm__ volatile ("femms");
 }
-static void vector_fmul_add_sse(float *dst, const float *src0, const float *src1,
-                                const float *src2, int len){
-    x86_reg i = (len-8)*4;
-    __asm__ volatile(
-        "1: \n\t"
-        "movaps   (%2,%0), %%xmm0 \n\t"
-        "movaps 16(%2,%0), %%xmm1 \n\t"
-        "mulps    (%3,%0), %%xmm0 \n\t"
-        "mulps  16(%3,%0), %%xmm1 \n\t"
-        "addps    (%4,%0), %%xmm0 \n\t"
-        "addps  16(%4,%0), %%xmm1 \n\t"
-        "movaps %%xmm0,   (%1,%0) \n\t"
-        "movaps %%xmm1, 16(%1,%0) \n\t"
-        "sub  $32, %0 \n\t"
-        "jge  1b \n\t"
-        :"+r"(i)
-        :"r"(dst), "r"(src0), "r"(src1), "r"(src2)
-        :"memory"
+
+static void vector_fmul_add_sse(float *dst, const float *src0,
+                                const float *src1, const float *src2, int len)
+{
+    x86_reg i = (len - 8) * 4;
+    __asm__ volatile (
+        "1:                             \n\t"
+        "movaps   (%2, %0), %%xmm0      \n\t"
+        "movaps 16(%2, %0), %%xmm1      \n\t"
+        "mulps    (%3, %0), %%xmm0      \n\t"
+        "mulps  16(%3, %0), %%xmm1      \n\t"
+        "addps    (%4, %0), %%xmm0      \n\t"
+        "addps  16(%4, %0), %%xmm1      \n\t"
+        "movaps     %%xmm0,   (%1, %0)  \n\t"
+        "movaps     %%xmm1, 16(%1, %0)  \n\t"
+        "sub           $32, %0          \n\t"
+        "jge            1b              \n\t"
+        : "+r"(i)
+        : "r"(dst), "r"(src0), "r"(src1), "r"(src2)
+        : "memory"
     );
 }
 
 #if HAVE_6REGS
-static void vector_fmul_window_3dnow2(float *dst, const float *src0, const float *src1,
-                                      const float *win, int len){
-    x86_reg i = -len*4;
-    x86_reg j = len*4-8;
-    __asm__ volatile(
-        "1: \n"
-        "pswapd  (%5,%1), %%mm1 \n"
-        "movq    (%5,%0), %%mm0 \n"
-        "pswapd  (%4,%1), %%mm5 \n"
-        "movq    (%3,%0), %%mm4 \n"
-        "movq      %%mm0, %%mm2 \n"
-        "movq      %%mm1, %%mm3 \n"
-        "pfmul     %%mm4, %%mm2 \n" // src0[len+i]*win[len+i]
-        "pfmul     %%mm5, %%mm3 \n" // src1[    j]*win[len+j]
-        "pfmul     %%mm4, %%mm1 \n" // src0[len+i]*win[len+j]
-        "pfmul     %%mm5, %%mm0 \n" // src1[    j]*win[len+i]
-        "pfadd     %%mm3, %%mm2 \n"
-        "pfsub     %%mm0, %%mm1 \n"
-        "pswapd    %%mm2, %%mm2 \n"
-        "movq      %%mm1, (%2,%0) \n"
-        "movq      %%mm2, (%2,%1) \n"
-        "sub $8, %1 \n"
-        "add $8, %0 \n"
-        "jl 1b \n"
-        "femms \n"
-        :"+r"(i), "+r"(j)
-        :"r"(dst+len), "r"(src0+len), "r"(src1), "r"(win+len)
+static void vector_fmul_window_3dnow2(float *dst, const float *src0,
+                                      const float *src1, const float *win,
+                                      int len)
+{
+    x86_reg i = -len * 4;
+    x86_reg j =  len * 4 - 8;
+    __asm__ volatile (
+        "1:                             \n"
+        "pswapd (%5, %1), %%mm1         \n"
+        "movq   (%5, %0), %%mm0         \n"
+        "pswapd (%4, %1), %%mm5         \n"
+        "movq   (%3, %0), %%mm4         \n"
+        "movq      %%mm0, %%mm2         \n"
+        "movq      %%mm1, %%mm3         \n"
+        "pfmul     %%mm4, %%mm2         \n" // src0[len + i] * win[len + i]
+        "pfmul     %%mm5, %%mm3         \n" // src1[j]       * win[len + j]
+        "pfmul     %%mm4, %%mm1         \n" // src0[len + i] * win[len + j]
+        "pfmul     %%mm5, %%mm0         \n" // src1[j]       * win[len + i]
+        "pfadd     %%mm3, %%mm2         \n"
+        "pfsub     %%mm0, %%mm1         \n"
+        "pswapd    %%mm2, %%mm2         \n"
+        "movq      %%mm1, (%2, %0)      \n"
+        "movq      %%mm2, (%2, %1)      \n"
+        "sub          $8, %1            \n"
+        "add          $8, %0            \n"
+        "jl           1b                \n"
+        "femms                          \n"
+        : "+r"(i), "+r"(j)
+        : "r"(dst + len), "r"(src0 + len), "r"(src1), "r"(win + len)
     );
 }
 
-static void vector_fmul_window_sse(float *dst, const float *src0, const float *src1,
-                                   const float *win, int len){
-    x86_reg i = -len*4;
-    x86_reg j = len*4-16;
-    __asm__ volatile(
-        "1: \n"
-        "movaps       (%5,%1), %%xmm1 \n"
-        "movaps       (%5,%0), %%xmm0 \n"
-        "movaps       (%4,%1), %%xmm5 \n"
-        "movaps       (%3,%0), %%xmm4 \n"
-        "shufps $0x1b, %%xmm1, %%xmm1 \n"
-        "shufps $0x1b, %%xmm5, %%xmm5 \n"
-        "movaps        %%xmm0, %%xmm2 \n"
-        "movaps        %%xmm1, %%xmm3 \n"
-        "mulps         %%xmm4, %%xmm2 \n" // src0[len+i]*win[len+i]
-        "mulps         %%xmm5, %%xmm3 \n" // src1[    j]*win[len+j]
-        "mulps         %%xmm4, %%xmm1 \n" // src0[len+i]*win[len+j]
-        "mulps         %%xmm5, %%xmm0 \n" // src1[    j]*win[len+i]
-        "addps         %%xmm3, %%xmm2 \n"
-        "subps         %%xmm0, %%xmm1 \n"
-        "shufps $0x1b, %%xmm2, %%xmm2 \n"
-        "movaps        %%xmm1, (%2,%0) \n"
-        "movaps        %%xmm2, (%2,%1) \n"
-        "sub $16, %1 \n"
-        "add $16, %0 \n"
-        "jl 1b \n"
-        :"+r"(i), "+r"(j)
-        :"r"(dst+len), "r"(src0+len), "r"(src1), "r"(win+len)
+static void vector_fmul_window_sse(float *dst, const float *src0,
+                                   const float *src1, const float *win, int len)
+{
+    x86_reg i = -len * 4;
+    x86_reg j =  len * 4 - 16;
+    __asm__ volatile (
+        "1:                             \n"
+        "movaps      (%5, %1), %%xmm1   \n"
+        "movaps      (%5, %0), %%xmm0   \n"
+        "movaps      (%4, %1), %%xmm5   \n"
+        "movaps      (%3, %0), %%xmm4   \n"
+        "shufps $0x1b, %%xmm1, %%xmm1   \n"
+        "shufps $0x1b, %%xmm5, %%xmm5   \n"
+        "movaps        %%xmm0, %%xmm2   \n"
+        "movaps        %%xmm1, %%xmm3   \n"
+        "mulps         %%xmm4, %%xmm2   \n" // src0[len + i] * win[len + i]
+        "mulps         %%xmm5, %%xmm3   \n" // src1[j]       * win[len + j]
+        "mulps         %%xmm4, %%xmm1   \n" // src0[len + i] * win[len + j]
+        "mulps         %%xmm5, %%xmm0   \n" // src1[j]       * win[len + i]
+        "addps         %%xmm3, %%xmm2   \n"
+        "subps         %%xmm0, %%xmm1   \n"
+        "shufps $0x1b, %%xmm2, %%xmm2   \n"
+        "movaps        %%xmm1, (%2, %0) \n"
+        "movaps        %%xmm2, (%2, %1) \n"
+        "sub              $16, %1       \n"
+        "add              $16, %0       \n"
+        "jl                1b           \n"
+        : "+r"(i), "+r"(j)
+        : "r"(dst + len), "r"(src0 + len), "r"(src1), "r"(win + len)
     );
 }
 #endif /* HAVE_6REGS */
 
-static void vector_clipf_sse(float *dst, const float *src, float min, float max,
-                             int len)
+static void vector_clipf_sse(float *dst, const float *src,
+                             float min, float max, int len)
 {
-    x86_reg i = (len-16)*4;
-    __asm__ volatile(
-        "movss  %3, %%xmm4 \n"
-        "movss  %4, %%xmm5 \n"
-        "shufps $0, %%xmm4, %%xmm4 \n"
-        "shufps $0, %%xmm5, %%xmm5 \n"
-        "1: \n\t"
-        "movaps    (%2,%0), %%xmm0 \n\t" // 3/1 on intel
-        "movaps  16(%2,%0), %%xmm1 \n\t"
-        "movaps  32(%2,%0), %%xmm2 \n\t"
-        "movaps  48(%2,%0), %%xmm3 \n\t"
-        "maxps      %%xmm4, %%xmm0 \n\t"
-        "maxps      %%xmm4, %%xmm1 \n\t"
-        "maxps      %%xmm4, %%xmm2 \n\t"
-        "maxps      %%xmm4, %%xmm3 \n\t"
-        "minps      %%xmm5, %%xmm0 \n\t"
-        "minps      %%xmm5, %%xmm1 \n\t"
-        "minps      %%xmm5, %%xmm2 \n\t"
-        "minps      %%xmm5, %%xmm3 \n\t"
-        "movaps  %%xmm0,   (%1,%0) \n\t"
-        "movaps  %%xmm1, 16(%1,%0) \n\t"
-        "movaps  %%xmm2, 32(%1,%0) \n\t"
-        "movaps  %%xmm3, 48(%1,%0) \n\t"
-        "sub  $64, %0 \n\t"
-        "jge 1b \n\t"
-        :"+&r"(i)
-        :"r"(dst), "r"(src), "m"(min), "m"(max)
-        :"memory"
+    x86_reg i = (len - 16) * 4;
+    __asm__ volatile (
+        "movss          %3, %%xmm4      \n\t"
+        "movss          %4, %%xmm5      \n\t"
+        "shufps $0, %%xmm4, %%xmm4      \n\t"
+        "shufps $0, %%xmm5, %%xmm5      \n\t"
+        "1:                             \n\t"
+        "movaps   (%2, %0), %%xmm0      \n\t" // 3/1 on intel
+        "movaps 16(%2, %0), %%xmm1      \n\t"
+        "movaps 32(%2, %0), %%xmm2      \n\t"
+        "movaps 48(%2, %0), %%xmm3      \n\t"
+        "maxps      %%xmm4, %%xmm0      \n\t"
+        "maxps      %%xmm4, %%xmm1      \n\t"
+        "maxps      %%xmm4, %%xmm2      \n\t"
+        "maxps      %%xmm4, %%xmm3      \n\t"
+        "minps      %%xmm5, %%xmm0      \n\t"
+        "minps      %%xmm5, %%xmm1      \n\t"
+        "minps      %%xmm5, %%xmm2      \n\t"
+        "minps      %%xmm5, %%xmm3      \n\t"
+        "movaps     %%xmm0,   (%1, %0)  \n\t"
+        "movaps     %%xmm1, 16(%1, %0)  \n\t"
+        "movaps     %%xmm2, 32(%1, %0)  \n\t"
+        "movaps     %%xmm3, 48(%1, %0)  \n\t"
+        "sub           $64, %0          \n\t"
+        "jge            1b              \n\t"
+        : "+&r"(i)
+        : "r"(dst), "r"(src), "m"(min), "m"(max)
+        : "memory"
     );
 }
 
@@ -2364,7 +2644,8 @@ void ff_vp3_idct_mmx(int16_t *input_data);
 void ff_vp3_idct_put_mmx(uint8_t *dest, int line_size, DCTELEM *block);
 void ff_vp3_idct_add_mmx(uint8_t *dest, int line_size, DCTELEM *block);
 
-void ff_vp3_idct_dc_add_mmx2(uint8_t *dest, int line_size, const DCTELEM *block);
+void ff_vp3_idct_dc_add_mmx2(uint8_t *dest, int line_size,
+                             const DCTELEM *block);
 
 void ff_vp3_v_loop_filter_mmx2(uint8_t *src, int stride, int *bounding_values);
 void ff_vp3_h_loop_filter_mmx2(uint8_t *src, int stride, int *bounding_values);
@@ -2373,11 +2654,19 @@ void ff_vp3_idct_sse2(int16_t *input_data);
 void ff_vp3_idct_put_sse2(uint8_t *dest, int line_size, DCTELEM *block);
 void ff_vp3_idct_add_sse2(uint8_t *dest, int line_size, DCTELEM *block);
 
-int32_t ff_scalarproduct_int16_mmx2(const int16_t *v1, const int16_t *v2, int order, int shift);
-int32_t ff_scalarproduct_int16_sse2(const int16_t *v1, const int16_t *v2, int order, int shift);
-int32_t ff_scalarproduct_and_madd_int16_mmx2(int16_t *v1, const int16_t *v2, const int16_t *v3, int order, int mul);
-int32_t ff_scalarproduct_and_madd_int16_sse2(int16_t *v1, const int16_t *v2, const int16_t *v3, int order, int mul);
-int32_t ff_scalarproduct_and_madd_int16_ssse3(int16_t *v1, const int16_t *v2, const int16_t *v3, int order, int mul);
+int32_t ff_scalarproduct_int16_mmx2(const int16_t *v1, const int16_t *v2,
+                                    int order, int shift);
+int32_t ff_scalarproduct_int16_sse2(const int16_t *v1, const int16_t *v2,
+                                    int order, int shift);
+int32_t ff_scalarproduct_and_madd_int16_mmx2(int16_t *v1, const int16_t *v2,
+                                             const int16_t *v3,
+                                             int order, int mul);
+int32_t ff_scalarproduct_and_madd_int16_sse2(int16_t *v1, const int16_t *v2,
+                                             const int16_t *v3,
+                                             int order, int mul);
+int32_t ff_scalarproduct_and_madd_int16_ssse3(int16_t *v1, const int16_t *v2,
+                                              const int16_t *v3,
+                                              int order, int mul);
 
 void ff_apply_window_int16_mmxext    (int16_t *output, const int16_t *input,
                                       const int16_t *window, unsigned int len);
@@ -2395,27 +2684,32 @@ void ff_apply_window_int16_ssse3_atom(int16_t *output, const int16_t *input,
 void ff_bswap32_buf_ssse3(uint32_t *dst, const uint32_t *src, int w);
 void ff_bswap32_buf_sse2(uint32_t *dst, const uint32_t *src, int w);
 
-void ff_add_hfyu_median_prediction_mmx2(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top);
-int  ff_add_hfyu_left_prediction_ssse3(uint8_t *dst, const uint8_t *src, int w, int left);
-int  ff_add_hfyu_left_prediction_sse4(uint8_t *dst, const uint8_t *src, int w, int left);
+void ff_add_hfyu_median_prediction_mmx2(uint8_t *dst, const uint8_t *top,
+                                        const uint8_t *diff, int w,
+                                        int *left, int *left_top);
+int  ff_add_hfyu_left_prediction_ssse3(uint8_t *dst, const uint8_t *src,
+                                       int w, int left);
+int  ff_add_hfyu_left_prediction_sse4(uint8_t *dst, const uint8_t *src,
+                                      int w, int left);
 
 float ff_scalarproduct_float_sse(const float *v1, const float *v2, int order);
 
-void ff_vector_clip_int32_mmx     (int32_t *dst, const int32_t *src, int32_t min,
-                                   int32_t max, unsigned int len);
-void ff_vector_clip_int32_sse2    (int32_t *dst, const int32_t *src, int32_t min,
-                                   int32_t max, unsigned int len);
-void ff_vector_clip_int32_int_sse2(int32_t *dst, const int32_t *src, int32_t min,
-                                   int32_t max, unsigned int len);
-void ff_vector_clip_int32_sse4    (int32_t *dst, const int32_t *src, int32_t min,
-                                   int32_t max, unsigned int len);
+void ff_vector_clip_int32_mmx     (int32_t *dst, const int32_t *src,
+                                   int32_t min, int32_t max, unsigned int len);
+void ff_vector_clip_int32_sse2    (int32_t *dst, const int32_t *src,
+                                   int32_t min, int32_t max, unsigned int len);
+void ff_vector_clip_int32_int_sse2(int32_t *dst, const int32_t *src,
+                                   int32_t min, int32_t max, unsigned int len);
+void ff_vector_clip_int32_sse4    (int32_t *dst, const int32_t *src,
+                                   int32_t min, int32_t max, unsigned int len);
 
 extern void ff_butterflies_float_interleave_sse(float *dst, const float *src0,
                                                 const float *src1, int len);
 extern void ff_butterflies_float_interleave_avx(float *dst, const float *src0,
                                                 const float *src1, int len);
 
-#define SET_QPEL_FUNCS(PFX, IDX, SIZE, CPU, PREFIX) \
+#define SET_QPEL_FUNCS(PFX, IDX, SIZE, CPU, PREFIX)                          \
+    do {                                                                     \
     c->PFX ## _pixels_tab[IDX][ 0] = PREFIX ## PFX ## SIZE ## _mc00_ ## CPU; \
     c->PFX ## _pixels_tab[IDX][ 1] = PREFIX ## PFX ## SIZE ## _mc10_ ## CPU; \
     c->PFX ## _pixels_tab[IDX][ 2] = PREFIX ## PFX ## SIZE ## _mc20_ ## CPU; \
@@ -2431,25 +2725,32 @@ extern void ff_butterflies_float_interleave_avx(float *dst, const float *src0,
     c->PFX ## _pixels_tab[IDX][12] = PREFIX ## PFX ## SIZE ## _mc03_ ## CPU; \
     c->PFX ## _pixels_tab[IDX][13] = PREFIX ## PFX ## SIZE ## _mc13_ ## CPU; \
     c->PFX ## _pixels_tab[IDX][14] = PREFIX ## PFX ## SIZE ## _mc23_ ## CPU; \
-    c->PFX ## _pixels_tab[IDX][15] = PREFIX ## PFX ## SIZE ## _mc33_ ## CPU
-
-#define SET_HPEL_FUNCS(PFX, IDX, SIZE, CPU) \
-    c->PFX ## _pixels_tab[IDX][0] = PFX ## _pixels ## SIZE ## _ ## CPU;    \
-    c->PFX ## _pixels_tab[IDX][1] = PFX ## _pixels ## SIZE ## _x2_ ## CPU; \
-    c->PFX ## _pixels_tab[IDX][2] = PFX ## _pixels ## SIZE ## _y2_ ## CPU; \
-    c->PFX ## _pixels_tab[IDX][3] = PFX ## _pixels ## SIZE ## _xy2_ ## CPU
-
-#define H264_QPEL_FUNCS(x, y, CPU) \
-    c->put_h264_qpel_pixels_tab[0][x+y*4] = put_h264_qpel16_mc##x##y##_##CPU; \
-    c->put_h264_qpel_pixels_tab[1][x+y*4] = put_h264_qpel8_mc##x##y##_##CPU;  \
-    c->avg_h264_qpel_pixels_tab[0][x+y*4] = avg_h264_qpel16_mc##x##y##_##CPU; \
-    c->avg_h264_qpel_pixels_tab[1][x+y*4] = avg_h264_qpel8_mc##x##y##_##CPU
-
-#define H264_QPEL_FUNCS_10(x, y, CPU) \
-    c->put_h264_qpel_pixels_tab[0][x+y*4] = ff_put_h264_qpel16_mc##x##y##_10_##CPU; \
-    c->put_h264_qpel_pixels_tab[1][x+y*4] = ff_put_h264_qpel8_mc##x##y##_10_##CPU;  \
-    c->avg_h264_qpel_pixels_tab[0][x+y*4] = ff_avg_h264_qpel16_mc##x##y##_10_##CPU; \
-    c->avg_h264_qpel_pixels_tab[1][x+y*4] = ff_avg_h264_qpel8_mc##x##y##_10_##CPU;
+    c->PFX ## _pixels_tab[IDX][15] = PREFIX ## PFX ## SIZE ## _mc33_ ## CPU; \
+    } while (0)
+
+#define SET_HPEL_FUNCS(PFX, IDX, SIZE, CPU)                                     \
+    do {                                                                        \
+        c->PFX ## _pixels_tab[IDX][0] = PFX ## _pixels ## SIZE ## _     ## CPU; \
+        c->PFX ## _pixels_tab[IDX][1] = PFX ## _pixels ## SIZE ## _x2_  ## CPU; \
+        c->PFX ## _pixels_tab[IDX][2] = PFX ## _pixels ## SIZE ## _y2_  ## CPU; \
+        c->PFX ## _pixels_tab[IDX][3] = PFX ## _pixels ## SIZE ## _xy2_ ## CPU; \
+    } while (0)
+
+#define H264_QPEL_FUNCS(x, y, CPU)                                                            \
+    do {                                                                                      \
+        c->put_h264_qpel_pixels_tab[0][x + y * 4] = put_h264_qpel16_mc ## x ## y ## _ ## CPU; \
+        c->put_h264_qpel_pixels_tab[1][x + y * 4] = put_h264_qpel8_mc  ## x ## y ## _ ## CPU; \
+        c->avg_h264_qpel_pixels_tab[0][x + y * 4] = avg_h264_qpel16_mc ## x ## y ## _ ## CPU; \
+        c->avg_h264_qpel_pixels_tab[1][x + y * 4] = avg_h264_qpel8_mc  ## x ## y ## _ ## CPU; \
+    } while (0)
+
+#define H264_QPEL_FUNCS_10(x, y, CPU)                                                               \
+    do {                                                                                            \
+        c->put_h264_qpel_pixels_tab[0][x + y * 4] = ff_put_h264_qpel16_mc ## x ## y ## _10_ ## CPU; \
+        c->put_h264_qpel_pixels_tab[1][x + y * 4] = ff_put_h264_qpel8_mc  ## x ## y ## _10_ ## CPU; \
+        c->avg_h264_qpel_pixels_tab[0][x + y * 4] = ff_avg_h264_qpel16_mc ## x ## y ## _10_ ## CPU; \
+        c->avg_h264_qpel_pixels_tab[1][x + y * 4] = ff_avg_h264_qpel8_mc  ## x ## y ## _10_ ## CPU; \
+    } while (0)
 
 static void dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx, int mm_flags)
 {
@@ -2464,18 +2765,18 @@ static void dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx, int mm_flags)
         c->clear_blocks = clear_blocks_mmx;
         c->draw_edges   = draw_edges_mmx;
 
-        SET_HPEL_FUNCS(put, 0, 16, mmx);
+        SET_HPEL_FUNCS(put,        0, 16, mmx);
         SET_HPEL_FUNCS(put_no_rnd, 0, 16, mmx);
-        SET_HPEL_FUNCS(avg, 0, 16, mmx);
+        SET_HPEL_FUNCS(avg,        0, 16, mmx);
         SET_HPEL_FUNCS(avg_no_rnd, 0, 16, mmx);
-        SET_HPEL_FUNCS(put, 1, 8, mmx);
-        SET_HPEL_FUNCS(put_no_rnd, 1, 8, mmx);
-        SET_HPEL_FUNCS(avg, 1, 8, mmx);
-        SET_HPEL_FUNCS(avg_no_rnd, 1, 8, mmx);
+        SET_HPEL_FUNCS(put,        1,  8, mmx);
+        SET_HPEL_FUNCS(put_no_rnd, 1,  8, mmx);
+        SET_HPEL_FUNCS(avg,        1,  8, mmx);
+        SET_HPEL_FUNCS(avg_no_rnd, 1,  8, mmx);
     }
 
 #if ARCH_X86_32 || !HAVE_YASM
-    c->gmc= gmc_mmx;
+    c->gmc = gmc_mmx;
 #endif
 #if ARCH_X86_32 && HAVE_YASM
     if (!high_bit_depth)
@@ -2543,47 +2844,48 @@ static void dsputil_init_mmx2(DSPContext *c, AVCodecContext *avctx,
             c->vp3_h_loop_filter = ff_vp3_h_loop_filter_mmx2;
         }
     }
-    if (CONFIG_VP3_DECODER && HAVE_YASM) {
+    if (CONFIG_VP3_DECODER && HAVE_YASM)
         c->vp3_idct_dc_add = ff_vp3_idct_dc_add_mmx2;
-    }
 
-    if (CONFIG_VP3_DECODER
-        && (avctx->codec_id == CODEC_ID_VP3 || avctx->codec_id == CODEC_ID_THEORA)) {
+    if (CONFIG_VP3_DECODER && (avctx->codec_id == CODEC_ID_VP3 ||
+                               avctx->codec_id == CODEC_ID_THEORA)) {
         c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_exact_mmx2;
         c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_exact_mmx2;
     }
 
-    SET_QPEL_FUNCS(put_qpel, 0, 16, mmx2, );
-    SET_QPEL_FUNCS(put_qpel, 1, 8, mmx2, );
-    SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, mmx2, );
-    SET_QPEL_FUNCS(put_no_rnd_qpel, 1, 8, mmx2, );
-    SET_QPEL_FUNCS(avg_qpel, 0, 16, mmx2, );
-    SET_QPEL_FUNCS(avg_qpel, 1, 8, mmx2, );
+    if (CONFIG_H264QPEL) {
+        SET_QPEL_FUNCS(put_qpel,        0, 16, mmx2, );
+        SET_QPEL_FUNCS(put_qpel,        1,  8, mmx2, );
+        SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, mmx2, );
+        SET_QPEL_FUNCS(put_no_rnd_qpel, 1,  8, mmx2, );
+        SET_QPEL_FUNCS(avg_qpel,        0, 16, mmx2, );
+        SET_QPEL_FUNCS(avg_qpel,        1,  8, mmx2, );
 
-    if (!high_bit_depth) {
-        SET_QPEL_FUNCS(put_h264_qpel, 0, 16, mmx2, );
-        SET_QPEL_FUNCS(put_h264_qpel, 1, 8, mmx2, );
-        SET_QPEL_FUNCS(put_h264_qpel, 2, 4, mmx2, );
-        SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, mmx2, );
-        SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, mmx2, );
-        SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, mmx2, );
-    } else if (bit_depth == 10) {
+        if (!high_bit_depth) {
+            SET_QPEL_FUNCS(put_h264_qpel, 0, 16, mmx2, );
+            SET_QPEL_FUNCS(put_h264_qpel, 1,  8, mmx2, );
+            SET_QPEL_FUNCS(put_h264_qpel, 2,  4, mmx2, );
+            SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, mmx2, );
+            SET_QPEL_FUNCS(avg_h264_qpel, 1,  8, mmx2, );
+            SET_QPEL_FUNCS(avg_h264_qpel, 2,  4, mmx2, );
+        } else if (bit_depth == 10) {
 #if HAVE_YASM
 #if !ARCH_X86_64
-        SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 10_mmxext, ff_);
-        SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_mmxext, ff_);
-        SET_QPEL_FUNCS(put_h264_qpel, 1, 8,  10_mmxext, ff_);
-        SET_QPEL_FUNCS(avg_h264_qpel, 1, 8,  10_mmxext, ff_);
+            SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 10_mmxext, ff_);
+            SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_mmxext, ff_);
+            SET_QPEL_FUNCS(put_h264_qpel, 1,  8, 10_mmxext, ff_);
+            SET_QPEL_FUNCS(avg_h264_qpel, 1,  8, 10_mmxext, ff_);
 #endif
-        SET_QPEL_FUNCS(put_h264_qpel, 2, 4,  10_mmxext, ff_);
-        SET_QPEL_FUNCS(avg_h264_qpel, 2, 4,  10_mmxext, ff_);
+            SET_QPEL_FUNCS(put_h264_qpel, 2, 4,  10_mmxext, ff_);
+            SET_QPEL_FUNCS(avg_h264_qpel, 2, 4,  10_mmxext, ff_);
 #endif
-    }
+        }
 
-    SET_QPEL_FUNCS(put_2tap_qpel, 0, 16, mmx2, );
-    SET_QPEL_FUNCS(put_2tap_qpel, 1, 8, mmx2, );
-    SET_QPEL_FUNCS(avg_2tap_qpel, 0, 16, mmx2, );
-    SET_QPEL_FUNCS(avg_2tap_qpel, 1, 8, mmx2, );
+        SET_QPEL_FUNCS(put_2tap_qpel, 0, 16, mmx2, );
+        SET_QPEL_FUNCS(put_2tap_qpel, 1,  8, mmx2, );
+        SET_QPEL_FUNCS(avg_2tap_qpel, 0, 16, mmx2, );
+        SET_QPEL_FUNCS(avg_2tap_qpel, 1,  8, mmx2, );
+    }
 
 #if HAVE_YASM
     if (!high_bit_depth && CONFIG_H264CHROMA) {
@@ -2599,7 +2901,7 @@ static void dsputil_init_mmx2(DSPContext *c, AVCodecContext *avctx,
         c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_10_mmxext;
     }
 
-    c->add_hfyu_median_prediction = ff_add_hfyu_median_prediction_mmx2;
+    c->add_hfyu_median_prediction   = ff_add_hfyu_median_prediction_mmx2;
 
     c->scalarproduct_int16          = ff_scalarproduct_int16_mmx2;
     c->scalarproduct_and_madd_int16 = ff_scalarproduct_and_madd_int16_mmx2;
@@ -2645,32 +2947,34 @@ static void dsputil_init_3dnow(DSPContext *c, AVCodecContext *avctx,
         }
     }
 
-    if (CONFIG_VP3_DECODER
-        && (avctx->codec_id == CODEC_ID_VP3 || avctx->codec_id == CODEC_ID_THEORA)) {
+    if (CONFIG_VP3_DECODER && (avctx->codec_id == CODEC_ID_VP3 ||
+                               avctx->codec_id == CODEC_ID_THEORA)) {
         c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_exact_3dnow;
         c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_exact_3dnow;
     }
 
-    SET_QPEL_FUNCS(put_qpel, 0, 16, 3dnow, );
-    SET_QPEL_FUNCS(put_qpel, 1, 8, 3dnow, );
-    SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, 3dnow, );
-    SET_QPEL_FUNCS(put_no_rnd_qpel, 1, 8, 3dnow, );
-    SET_QPEL_FUNCS(avg_qpel, 0, 16, 3dnow, );
-    SET_QPEL_FUNCS(avg_qpel, 1, 8, 3dnow, );
+    if (CONFIG_H264QPEL) {
+        SET_QPEL_FUNCS(put_qpel,        0, 16, 3dnow, );
+        SET_QPEL_FUNCS(put_qpel,        1,  8, 3dnow, );
+        SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, 3dnow, );
+        SET_QPEL_FUNCS(put_no_rnd_qpel, 1,  8, 3dnow, );
+        SET_QPEL_FUNCS(avg_qpel,        0, 16, 3dnow, );
+        SET_QPEL_FUNCS(avg_qpel,        1,  8, 3dnow, );
 
-    if (!high_bit_depth) {
-        SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 3dnow, );
-        SET_QPEL_FUNCS(put_h264_qpel, 1, 8, 3dnow, );
-        SET_QPEL_FUNCS(put_h264_qpel, 2, 4, 3dnow, );
-        SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 3dnow, );
-        SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, 3dnow, );
-        SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, 3dnow, );
-    }
+        if (!high_bit_depth) {
+            SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 3dnow, );
+            SET_QPEL_FUNCS(put_h264_qpel, 1,  8, 3dnow, );
+            SET_QPEL_FUNCS(put_h264_qpel, 2,  4, 3dnow, );
+            SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 3dnow, );
+            SET_QPEL_FUNCS(avg_h264_qpel, 1,  8, 3dnow, );
+            SET_QPEL_FUNCS(avg_h264_qpel, 2,  4, 3dnow, );
+        }
 
-    SET_QPEL_FUNCS(put_2tap_qpel, 0, 16, 3dnow, );
-    SET_QPEL_FUNCS(put_2tap_qpel, 1, 8, 3dnow, );
-    SET_QPEL_FUNCS(avg_2tap_qpel, 0, 16, 3dnow, );
-    SET_QPEL_FUNCS(avg_2tap_qpel, 1, 8, 3dnow, );
+        SET_QPEL_FUNCS(put_2tap_qpel, 0, 16, 3dnow, );
+        SET_QPEL_FUNCS(put_2tap_qpel, 1,  8, 3dnow, );
+        SET_QPEL_FUNCS(avg_2tap_qpel, 0, 16, 3dnow, );
+        SET_QPEL_FUNCS(avg_2tap_qpel, 1,  8, 3dnow, );
+    }
 
 #if HAVE_YASM
     if (!high_bit_depth && CONFIG_H264CHROMA) {
@@ -2702,7 +3006,7 @@ static void dsputil_init_sse(DSPContext *c, AVCodecContext *avctx, int mm_flags)
     const int high_bit_depth = avctx->bits_per_raw_sample > 8;
 
     if (!high_bit_depth) {
-        if (!(CONFIG_MPEG_XVMC_DECODER && avctx->xvmc_acceleration > 1)){
+        if (!(CONFIG_MPEG_XVMC_DECODER && avctx->xvmc_acceleration > 1)) {
             /* XvMCCreateBlocks() may not allocate 16-byte aligned blocks */
             c->clear_block  = clear_block_sse;
             c->clear_blocks = clear_blocks_sse;
@@ -2745,11 +3049,12 @@ static void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx,
             c->put_pixels_tab[0][0]        = put_pixels16_sse2;
             c->put_no_rnd_pixels_tab[0][0] = put_pixels16_sse2;
             c->avg_pixels_tab[0][0]        = avg_pixels16_sse2;
-            H264_QPEL_FUNCS(0, 0, sse2);
+            if (CONFIG_H264QPEL)
+                H264_QPEL_FUNCS(0, 0, sse2);
         }
     }
 
-    if (!high_bit_depth) {
+    if (!high_bit_depth && CONFIG_H264QPEL) {
         H264_QPEL_FUNCS(0, 1, sse2);
         H264_QPEL_FUNCS(0, 2, sse2);
         H264_QPEL_FUNCS(0, 3, sse2);
@@ -2766,14 +3071,15 @@ static void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx,
 
 #if HAVE_YASM
     if (bit_depth == 10) {
-        SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_sse2, ff_);
-        SET_QPEL_FUNCS(put_h264_qpel, 1, 8,  10_sse2, ff_);
-        SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 10_sse2, ff_);
-        SET_QPEL_FUNCS(avg_h264_qpel, 1, 8,  10_sse2, ff_);
-        H264_QPEL_FUNCS_10(1, 0, sse2_cache64);
-        H264_QPEL_FUNCS_10(2, 0, sse2_cache64);
-        H264_QPEL_FUNCS_10(3, 0, sse2_cache64);
-
+        if (CONFIG_H264QPEL) {
+            SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_sse2, ff_);
+            SET_QPEL_FUNCS(put_h264_qpel, 1,  8, 10_sse2, ff_);
+            SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 10_sse2, ff_);
+            SET_QPEL_FUNCS(avg_h264_qpel, 1,  8, 10_sse2, ff_);
+            H264_QPEL_FUNCS_10(1, 0, sse2_cache64);
+            H264_QPEL_FUNCS_10(2, 0, sse2_cache64);
+            H264_QPEL_FUNCS_10(3, 0, sse2_cache64);
+        }
         if (CONFIG_H264CHROMA) {
             c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_10_sse2;
             c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_10_sse2;
@@ -2789,7 +3095,7 @@ static void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx,
     }
     if (avctx->flags & CODEC_FLAG_BITEXACT) {
         c->apply_window_int16 = ff_apply_window_int16_sse2_ba;
-    } else  if (!(mm_flags & AV_CPU_FLAG_SSE2SLOW)) {
+    } else if (!(mm_flags & AV_CPU_FLAG_SSE2SLOW)) {
         c->apply_window_int16 = ff_apply_window_int16_sse2;
     }
     c->bswap_buf = ff_bswap32_buf_sse2;
@@ -2803,7 +3109,7 @@ static void dsputil_init_ssse3(DSPContext *c, AVCodecContext *avctx,
     const int high_bit_depth = avctx->bits_per_raw_sample > 8;
     const int bit_depth      = avctx->bits_per_raw_sample;
 
-    if (!high_bit_depth) {
+    if (!high_bit_depth && CONFIG_H264QPEL) {
         H264_QPEL_FUNCS(1, 0, ssse3);
         H264_QPEL_FUNCS(1, 1, ssse3);
         H264_QPEL_FUNCS(1, 2, ssse3);
@@ -2818,7 +3124,7 @@ static void dsputil_init_ssse3(DSPContext *c, AVCodecContext *avctx,
         H264_QPEL_FUNCS(3, 3, ssse3);
     }
 #if HAVE_YASM
-    else if (bit_depth == 10) {
+    else if (bit_depth == 10 && CONFIG_H264QPEL) {
         H264_QPEL_FUNCS_10(1, 0, ssse3_cache64);
         H264_QPEL_FUNCS_10(2, 0, ssse3_cache64);
         H264_QPEL_FUNCS_10(3, 0, ssse3_cache64);
@@ -2833,14 +3139,12 @@ static void dsputil_init_ssse3(DSPContext *c, AVCodecContext *avctx,
     if (mm_flags & AV_CPU_FLAG_SSE4) // not really sse4, just slow on Conroe
         c->add_hfyu_left_prediction = ff_add_hfyu_left_prediction_sse4;
 
-    if (mm_flags & AV_CPU_FLAG_ATOM) {
+    if (mm_flags & AV_CPU_FLAG_ATOM)
         c->apply_window_int16 = ff_apply_window_int16_ssse3_atom;
-    } else {
+    else
         c->apply_window_int16 = ff_apply_window_int16_ssse3;
-    }
-    if (!(mm_flags & (AV_CPU_FLAG_SSE42|AV_CPU_FLAG_3DNOW))) { // cachesplit
+    if (!(mm_flags & (AV_CPU_FLAG_SSE42|AV_CPU_FLAG_3DNOW))) // cachesplit
         c->scalarproduct_and_madd_int16 = ff_scalarproduct_and_madd_int16_ssse3;
-    }
     c->bswap_buf = ff_bswap32_buf_ssse3;
 #endif
 #endif
@@ -2862,9 +3166,11 @@ static void dsputil_init_avx(DSPContext *c, AVCodecContext *avctx, int mm_flags)
     if (bit_depth == 10) {
         // AVX implies !cache64.
         // TODO: Port cache(32|64) detection from x264.
-        H264_QPEL_FUNCS_10(1, 0, sse2);
-        H264_QPEL_FUNCS_10(2, 0, sse2);
-        H264_QPEL_FUNCS_10(3, 0, sse2);
+        if (CONFIG_H264QPEL) {
+            H264_QPEL_FUNCS_10(1, 0, sse2);
+            H264_QPEL_FUNCS_10(2, 0, sse2);
+            H264_QPEL_FUNCS_10(3, 0, sse2);
+        }
 
         if (CONFIG_H264CHROMA) {
             c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_10_avx;
@@ -2875,13 +3181,13 @@ static void dsputil_init_avx(DSPContext *c, AVCodecContext *avctx, int mm_flags)
 #endif
 }
 
-void ff_dsputil_init_mmx(DSPContextc, AVCodecContext *avctx)
+void ff_dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx)
 {
     int mm_flags = av_get_cpu_flags();
 
     if (avctx->dsp_mask) {
         if (avctx->dsp_mask & AV_CPU_FLAG_FORCE)
-            mm_flags |= (avctx->dsp_mask & 0xffff);
+            mm_flags |=   avctx->dsp_mask & 0xffff;
         else
             mm_flags &= ~(avctx->dsp_mask & 0xffff);
     }
@@ -2902,56 +3208,57 @@ void ff_dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
 #endif
 
     if (mm_flags & AV_CPU_FLAG_MMX) {
-        const int idct_algo= avctx->idct_algo;
+        const int idct_algo = avctx->idct_algo;
 
         if (avctx->lowres == 0 && avctx->bits_per_raw_sample <= 8) {
-            if(idct_algo==FF_IDCT_AUTO || idct_algo==FF_IDCT_SIMPLEMMX){
-                c->idct_put= ff_simple_idct_put_mmx;
-                c->idct_add= ff_simple_idct_add_mmx;
-                c->idct    = ff_simple_idct_mmx;
-                c->idct_permutation_type= FF_SIMPLE_IDCT_PERM;
+            if (idct_algo == FF_IDCT_AUTO || idct_algo == FF_IDCT_SIMPLEMMX) {
+                c->idct_put              = ff_simple_idct_put_mmx;
+                c->idct_add              = ff_simple_idct_add_mmx;
+                c->idct                  = ff_simple_idct_mmx;
+                c->idct_permutation_type = FF_SIMPLE_IDCT_PERM;
 #if CONFIG_GPL
-            }else if(idct_algo==FF_IDCT_LIBMPEG2MMX){
-                if(mm_flags & AV_CPU_FLAG_MMX2){
-                    c->idct_put= ff_libmpeg2mmx2_idct_put;
-                    c->idct_add= ff_libmpeg2mmx2_idct_add;
-                    c->idct    = ff_mmxext_idct;
-                }else{
-                    c->idct_put= ff_libmpeg2mmx_idct_put;
-                    c->idct_add= ff_libmpeg2mmx_idct_add;
-                    c->idct    = ff_mmx_idct;
+            } else if (idct_algo == FF_IDCT_LIBMPEG2MMX) {
+                if (mm_flags & AV_CPU_FLAG_MMX2) {
+                    c->idct_put = ff_libmpeg2mmx2_idct_put;
+                    c->idct_add = ff_libmpeg2mmx2_idct_add;
+                    c->idct     = ff_mmxext_idct;
+                } else {
+                    c->idct_put = ff_libmpeg2mmx_idct_put;
+                    c->idct_add = ff_libmpeg2mmx_idct_add;
+                    c->idct     = ff_mmx_idct;
                 }
-                c->idct_permutation_type= FF_LIBMPEG2_IDCT_PERM;
+                c->idct_permutation_type = FF_LIBMPEG2_IDCT_PERM;
 #endif
-            }else if((CONFIG_VP3_DECODER || CONFIG_VP5_DECODER || CONFIG_VP6_DECODER) &&
-                     idct_algo==FF_IDCT_VP3 && HAVE_YASM){
-                if(mm_flags & AV_CPU_FLAG_SSE2){
-                    c->idct_put= ff_vp3_idct_put_sse2;
-                    c->idct_add= ff_vp3_idct_add_sse2;
-                    c->idct    = ff_vp3_idct_sse2;
-                    c->idct_permutation_type= FF_TRANSPOSE_IDCT_PERM;
-                }else{
-                    c->idct_put= ff_vp3_idct_put_mmx;
-                    c->idct_add= ff_vp3_idct_add_mmx;
-                    c->idct    = ff_vp3_idct_mmx;
-                    c->idct_permutation_type= FF_PARTTRANS_IDCT_PERM;
+            } else if ((CONFIG_VP3_DECODER || CONFIG_VP5_DECODER ||
+                        CONFIG_VP6_DECODER) &&
+                       idct_algo == FF_IDCT_VP3 && HAVE_YASM) {
+                if (mm_flags & AV_CPU_FLAG_SSE2) {
+                    c->idct_put              = ff_vp3_idct_put_sse2;
+                    c->idct_add              = ff_vp3_idct_add_sse2;
+                    c->idct                  = ff_vp3_idct_sse2;
+                    c->idct_permutation_type = FF_TRANSPOSE_IDCT_PERM;
+                } else {
+                    c->idct_put              = ff_vp3_idct_put_mmx;
+                    c->idct_add              = ff_vp3_idct_add_mmx;
+                    c->idct                  = ff_vp3_idct_mmx;
+                    c->idct_permutation_type = FF_PARTTRANS_IDCT_PERM;
                 }
-            }else if(idct_algo==FF_IDCT_CAVS){
-                    c->idct_permutation_type= FF_TRANSPOSE_IDCT_PERM;
-            }else if(idct_algo==FF_IDCT_XVIDMMX){
-                if(mm_flags & AV_CPU_FLAG_SSE2){
-                    c->idct_put= ff_idct_xvid_sse2_put;
-                    c->idct_add= ff_idct_xvid_sse2_add;
-                    c->idct    = ff_idct_xvid_sse2;
-                    c->idct_permutation_type= FF_SSE2_IDCT_PERM;
-                }else if(mm_flags & AV_CPU_FLAG_MMX2){
-                    c->idct_put= ff_idct_xvid_mmx2_put;
-                    c->idct_add= ff_idct_xvid_mmx2_add;
-                    c->idct    = ff_idct_xvid_mmx2;
-                }else{
-                    c->idct_put= ff_idct_xvid_mmx_put;
-                    c->idct_add= ff_idct_xvid_mmx_add;
-                    c->idct    = ff_idct_xvid_mmx;
+            } else if (idct_algo == FF_IDCT_CAVS) {
+                    c->idct_permutation_type = FF_TRANSPOSE_IDCT_PERM;
+            } else if (idct_algo == FF_IDCT_XVIDMMX) {
+                if (mm_flags & AV_CPU_FLAG_SSE2) {
+                    c->idct_put              = ff_idct_xvid_sse2_put;
+                    c->idct_add              = ff_idct_xvid_sse2_add;
+                    c->idct                  = ff_idct_xvid_sse2;
+                    c->idct_permutation_type = FF_SSE2_IDCT_PERM;
+                } else if (mm_flags & AV_CPU_FLAG_MMX2) {
+                    c->idct_put              = ff_idct_xvid_mmx2_put;
+                    c->idct_add              = ff_idct_xvid_mmx2_add;
+                    c->idct                  = ff_idct_xvid_mmx2;
+                } else {
+                    c->idct_put              = ff_idct_xvid_mmx_put;
+                    c->idct_add              = ff_idct_xvid_mmx_add;
+                    c->idct                  = ff_idct_xvid_mmx;
                 }
             }
         }
@@ -2962,13 +3269,13 @@ void ff_dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
     if (mm_flags & AV_CPU_FLAG_MMX2)
         dsputil_init_mmx2(c, avctx, mm_flags);
 
-    if (HAVE_AMD3DNOW && (mm_flags & AV_CPU_FLAG_3DNOW))
+    if (mm_flags & AV_CPU_FLAG_3DNOW && HAVE_AMD3DNOW)
         dsputil_init_3dnow(c, avctx, mm_flags);
 
-    if (HAVE_AMD3DNOWEXT && (mm_flags & AV_CPU_FLAG_3DNOWEXT))
+    if (mm_flags & AV_CPU_FLAG_3DNOWEXT && HAVE_AMD3DNOWEXT)
         dsputil_init_3dnow2(c, avctx, mm_flags);
 
-    if (HAVE_SSE && (mm_flags & AV_CPU_FLAG_SSE))
+    if (mm_flags & AV_CPU_FLAG_SSE && HAVE_SSE)
         dsputil_init_sse(c, avctx, mm_flags);
 
     if (mm_flags & AV_CPU_FLAG_SSE2)