[FFmpeg-cvslog] x86: h264: Convert 8-bit QPEL inline assembly to YASM

Daniel Kang git at videolan.org
Mon Nov 26 02:22:09 CET 2012


ffmpeg | branch: master | Daniel Kang <daniel.d.kang at gmail.com> | Sat Oct 13 10:04:50 2012 -0500| [610e00b3594bf0f2a75713f20e9c4edf0d03a818] | committer: Diego Biurrun

x86: h264: Convert 8-bit QPEL inline assembly to YASM

Signed-off-by: Diego Biurrun <diego at biurrun.de>

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=610e00b3594bf0f2a75713f20e9c4edf0d03a818
---

 libavcodec/x86/Makefile               |    3 +-
 libavcodec/x86/dsputil.asm            |  231 +++++++
 libavcodec/x86/dsputil_avg_template.c |  130 ----
 libavcodec/x86/dsputil_mmx.c          |  107 +---
 libavcodec/x86/h264_qpel.c            | 1141 +++++----------------------------
 libavcodec/x86/h264_qpel_8bit.asm     |  862 +++++++++++++++++++++++++
 6 files changed, 1276 insertions(+), 1198 deletions(-)

diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile
index 053290b..1e70d53 100644
--- a/libavcodec/x86/Makefile
+++ b/libavcodec/x86/Makefile
@@ -51,7 +51,8 @@ YASM-OBJS-$(CONFIG_H264DSP)            += x86/h264_deblock.o            \
                                           x86/h264_weight_10bit.o
 YASM-OBJS-$(CONFIG_H264PRED)           += x86/h264_intrapred.o          \
                                           x86/h264_intrapred_10bit.o
-YASM-OBJS-$(CONFIG_H264QPEL)           += x86/h264_qpel_10bit.o
+YASM-OBJS-$(CONFIG_H264QPEL)           += x86/h264_qpel_8bit.o          \
+                                          x86/h264_qpel_10bit.o
 YASM-OBJS-$(CONFIG_MPEGAUDIODSP)       += x86/imdct36.o
 YASM-OBJS-$(CONFIG_PNG_DECODER)        += x86/pngdsp.o
 YASM-OBJS-$(CONFIG_PRORES_DECODER)     += x86/proresdsp.o
diff --git a/libavcodec/x86/dsputil.asm b/libavcodec/x86/dsputil.asm
index 726d4c6..dd0c795 100644
--- a/libavcodec/x86/dsputil.asm
+++ b/libavcodec/x86/dsputil.asm
@@ -1354,3 +1354,234 @@ BSWAP32_BUF
 
 INIT_XMM ssse3
 BSWAP32_BUF
+
+%macro op_avgh 3
+    movh   %3, %2
+    pavgb  %1, %3
+    movh   %2, %1
+%endmacro
+
+%macro op_avg 2
+    pavgb  %1, %2
+    mova   %2, %1
+%endmacro
+
+%macro op_puth 2-3
+    movh   %2, %1
+%endmacro
+
+%macro op_put 2
+    mova   %2, %1
+%endmacro
+
+; void pixels4_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h)
+%macro PIXELS4_L2 1
+%define OP op_%1h
+cglobal %1_pixels4_l2, 6,6
+    movsxdifnidn r3, r3d
+    movsxdifnidn r4, r4d
+    test        r5d, 1
+    je        .loop
+    movd         m0, [r1]
+    movd         m1, [r2]
+    add          r1, r4
+    add          r2, 4
+    pavgb        m0, m1
+    OP           m0, [r0], m3
+    add          r0, r3
+    dec         r5d
+.loop:
+    mova         m0, [r1]
+    mova         m1, [r1+r4]
+    lea          r1, [r1+2*r4]
+    pavgb        m0, [r2]
+    pavgb        m1, [r2+4]
+    OP           m0, [r0], m3
+    OP           m1, [r0+r3], m3
+    lea          r0, [r0+2*r3]
+    mova         m0, [r1]
+    mova         m1, [r1+r4]
+    lea          r1, [r1+2*r4]
+    pavgb        m0, [r2+8]
+    pavgb        m1, [r2+12]
+    OP           m0, [r0], m3
+    OP           m1, [r0+r3], m3
+    lea          r0, [r0+2*r3]
+    add          r2, 16
+    sub         r5d, 4
+    jne       .loop
+    REP_RET
+%endmacro
+
+INIT_MMX mmxext
+PIXELS4_L2 put
+PIXELS4_L2 avg
+
+; void pixels8_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h)
+%macro PIXELS8_L2 1
+%define OP op_%1
+cglobal %1_pixels8_l2, 6,6
+    movsxdifnidn r3, r3d
+    movsxdifnidn r4, r4d
+    test        r5d, 1
+    je        .loop
+    mova         m0, [r1]
+    mova         m1, [r2]
+    add          r1, r4
+    add          r2, 8
+    pavgb        m0, m1
+    OP           m0, [r0]
+    add          r0, r3
+    dec         r5d
+.loop:
+    mova         m0, [r1]
+    mova         m1, [r1+r4]
+    lea          r1, [r1+2*r4]
+    pavgb        m0, [r2]
+    pavgb        m1, [r2+8]
+    OP           m0, [r0]
+    OP           m1, [r0+r3]
+    lea          r0, [r0+2*r3]
+    mova         m0, [r1]
+    mova         m1, [r1+r4]
+    lea          r1, [r1+2*r4]
+    pavgb        m0, [r2+16]
+    pavgb        m1, [r2+24]
+    OP           m0, [r0]
+    OP           m1, [r0+r3]
+    lea          r0, [r0+2*r3]
+    add          r2, 32
+    sub         r5d, 4
+    jne       .loop
+    REP_RET
+%endmacro
+
+INIT_MMX mmxext
+PIXELS8_L2 put
+PIXELS8_L2 avg
+
+; void pixels16_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h)
+%macro PIXELS16_L2 1
+%define OP op_%1
+cglobal %1_pixels16_l2, 6,6
+    movsxdifnidn r3, r3d
+    movsxdifnidn r4, r4d
+    test        r5d, 1
+    je        .loop
+    mova         m0, [r1]
+    mova         m1, [r1+8]
+    pavgb        m0, [r2]
+    pavgb        m1, [r2+8]
+    add          r1, r4
+    add          r2, 16
+    OP           m0, [r0]
+    OP           m1, [r0+8]
+    add          r0, r3
+    dec         r5d
+.loop:
+    mova         m0, [r1]
+    mova         m1, [r1+8]
+    add          r1, r4
+    pavgb        m0, [r2]
+    pavgb        m1, [r2+8]
+    OP           m0, [r0]
+    OP           m1, [r0+8]
+    add          r0, r3
+    mova         m0, [r1]
+    mova         m1, [r1+8]
+    add          r1, r4
+    pavgb        m0, [r2+16]
+    pavgb        m1, [r2+24]
+    OP           m0, [r0]
+    OP           m1, [r0+8]
+    add          r0, r3
+    add          r2, 32
+    sub         r5d, 2
+    jne       .loop
+    REP_RET
+%endmacro
+
+INIT_MMX mmxext
+PIXELS16_L2 put
+PIXELS16_L2 avg
+
+INIT_MMX mmxext
+; void pixels(uint8_t *block, const uint8_t *pixels, int line_size, int h)
+%macro PIXELS48 2
+%if %2 == 4
+%define OP movh
+%else
+%define OP mova
+%endif
+cglobal %1_pixels%2, 4,5
+    movsxdifnidn r2, r2d
+    lea          r4, [r2*3]
+.loop:
+    OP           m0, [r1]
+    OP           m1, [r1+r2]
+    OP           m2, [r1+r2*2]
+    OP           m3, [r1+r4]
+    lea          r1, [r1+r2*4]
+%ifidn %1, avg
+    pavgb        m0, [r0]
+    pavgb        m1, [r0+r2]
+    pavgb        m2, [r0+r2*2]
+    pavgb        m3, [r0+r4]
+%endif
+    OP         [r0], m0
+    OP      [r0+r2], m1
+    OP    [r0+r2*2], m2
+    OP      [r0+r4], m3
+    sub         r3d, 4
+    lea          r0, [r0+r2*4]
+    jne       .loop
+    RET
+%endmacro
+
+PIXELS48 put, 4
+PIXELS48 avg, 4
+PIXELS48 put, 8
+PIXELS48 avg, 8
+
+INIT_XMM sse2
+; void put_pixels16_sse2(uint8_t *block, const uint8_t *pixels, int line_size, int h)
+cglobal put_pixels16, 4,5,4
+    movsxdifnidn r2, r2d
+    lea          r4, [r2*3]
+.loop:
+    movu         m0, [r1]
+    movu         m1, [r1+r2]
+    movu         m2, [r1+r2*2]
+    movu         m3, [r1+r4]
+    lea          r1, [r1+r2*4]
+    mova       [r0], m0
+    mova    [r0+r2], m1
+    mova  [r0+r2*2], m2
+    mova    [r0+r4], m3
+    sub         r3d, 4
+    lea          r0, [r0+r2*4]
+    jnz       .loop
+    REP_RET
+
+; void avg_pixels16_sse2(uint8_t *block, const uint8_t *pixels, int line_size, int h)
+cglobal avg_pixels16, 4,5,4
+    movsxdifnidn r2, r2d
+    lea          r4, [r2*3]
+.loop:
+    movu         m0, [r1]
+    movu         m1, [r1+r2]
+    movu         m2, [r1+r2*2]
+    movu         m3, [r1+r4]
+    lea          r1, [r1+r2*4]
+    pavgb        m0, [r0]
+    pavgb        m1, [r0+r2]
+    pavgb        m2, [r0+r2*2]
+    pavgb        m3, [r0+r4]
+    mova       [r0], m0
+    mova    [r0+r2], m1
+    mova  [r0+r2*2], m2
+    mova    [r0+r4], m3
+    sub         r3d, 4
+    lea          r0, [r0+r2*4]
+    jnz       .loop
+    REP_RET
diff --git a/libavcodec/x86/dsputil_avg_template.c b/libavcodec/x86/dsputil_avg_template.c
index cffcd4e..1a4343d 100644
--- a/libavcodec/x86/dsputil_avg_template.c
+++ b/libavcodec/x86/dsputil_avg_template.c
@@ -56,57 +56,6 @@ static void DEF(put_pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line_
 }
 
 #ifndef SKIP_FOR_3DNOW
-static void DEF(put_pixels4_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h)
-{
-    __asm__ volatile(
-        "testl $1, %0                   \n\t"
-            " jz 1f                     \n\t"
-        "movd   (%1), %%mm0             \n\t"
-        "movd   (%2), %%mm1             \n\t"
-        "add    %4, %1                  \n\t"
-        "add    $4, %2                  \n\t"
-        PAVGB" %%mm1, %%mm0             \n\t"
-        "movd   %%mm0, (%3)             \n\t"
-        "add    %5, %3                  \n\t"
-        "decl   %0                      \n\t"
-        "1:                             \n\t"
-        "movd   (%1), %%mm0             \n\t"
-        "add    %4, %1                  \n\t"
-        "movd   (%1), %%mm1             \n\t"
-        "movd   (%2), %%mm2             \n\t"
-        "movd   4(%2), %%mm3            \n\t"
-        "add    %4, %1                  \n\t"
-        PAVGB" %%mm2, %%mm0             \n\t"
-        PAVGB" %%mm3, %%mm1             \n\t"
-        "movd   %%mm0, (%3)             \n\t"
-        "add    %5, %3                  \n\t"
-        "movd   %%mm1, (%3)             \n\t"
-        "add    %5, %3                  \n\t"
-        "movd   (%1), %%mm0             \n\t"
-        "add    %4, %1                  \n\t"
-        "movd   (%1), %%mm1             \n\t"
-        "movd   8(%2), %%mm2            \n\t"
-        "movd   12(%2), %%mm3           \n\t"
-        "add    %4, %1                  \n\t"
-        PAVGB" %%mm2, %%mm0             \n\t"
-        PAVGB" %%mm3, %%mm1             \n\t"
-        "movd   %%mm0, (%3)             \n\t"
-        "add    %5, %3                  \n\t"
-        "movd   %%mm1, (%3)             \n\t"
-        "add    %5, %3                  \n\t"
-        "add    $16, %2                 \n\t"
-        "subl   $4, %0                  \n\t"
-        "jnz    1b                      \n\t"
-#if !HAVE_EBX_AVAILABLE //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cannot be used
-        :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst)
-#else
-        :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)
-#endif
-        :"S"((x86_reg)src1Stride), "D"((x86_reg)dstStride)
-        :"memory");
-}
-
-
 static void DEF(put_pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h)
 {
     __asm__ volatile(
@@ -227,58 +176,6 @@ static void DEF(put_no_rnd_pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src
         :"memory");*/
 }
 
-static void DEF(avg_pixels4_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h)
-{
-    __asm__ volatile(
-        "testl $1, %0                   \n\t"
-            " jz 1f                     \n\t"
-        "movd   (%1), %%mm0             \n\t"
-        "movd   (%2), %%mm1             \n\t"
-        "add    %4, %1                  \n\t"
-        "add    $4, %2                  \n\t"
-        PAVGB" %%mm1, %%mm0             \n\t"
-        PAVGB" (%3), %%mm0              \n\t"
-        "movd   %%mm0, (%3)             \n\t"
-        "add    %5, %3                  \n\t"
-        "decl   %0                      \n\t"
-        "1:                             \n\t"
-        "movd   (%1), %%mm0             \n\t"
-        "add    %4, %1                  \n\t"
-        "movd   (%1), %%mm1             \n\t"
-        "add    %4, %1                  \n\t"
-        PAVGB" (%2), %%mm0              \n\t"
-        PAVGB" 4(%2), %%mm1             \n\t"
-        PAVGB" (%3), %%mm0              \n\t"
-        "movd   %%mm0, (%3)             \n\t"
-        "add    %5, %3                  \n\t"
-        PAVGB" (%3), %%mm1              \n\t"
-        "movd   %%mm1, (%3)             \n\t"
-        "add    %5, %3                  \n\t"
-        "movd   (%1), %%mm0             \n\t"
-        "add    %4, %1                  \n\t"
-        "movd   (%1), %%mm1             \n\t"
-        "add    %4, %1                  \n\t"
-        PAVGB" 8(%2), %%mm0             \n\t"
-        PAVGB" 12(%2), %%mm1            \n\t"
-        PAVGB" (%3), %%mm0              \n\t"
-        "movd   %%mm0, (%3)             \n\t"
-        "add    %5, %3                  \n\t"
-        PAVGB" (%3), %%mm1              \n\t"
-        "movd   %%mm1, (%3)             \n\t"
-        "add    %5, %3                  \n\t"
-        "add    $16, %2                 \n\t"
-        "subl   $4, %0                  \n\t"
-        "jnz    1b                      \n\t"
-#if !HAVE_EBX_AVAILABLE //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cannot be used
-        :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst)
-#else
-        :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)
-#endif
-        :"S"((x86_reg)src1Stride), "D"((x86_reg)dstStride)
-        :"memory");
-}
-
-
 static void DEF(avg_pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h)
 {
     __asm__ volatile(
@@ -876,33 +773,6 @@ static void DEF(avg_pixels8_xy2)(uint8_t *block, const uint8_t *pixels, int line
         :"%"REG_a,  "memory");
 }
 
-#ifndef SKIP_FOR_3DNOW
-static void DEF(avg_pixels4)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
-{
-    do {
-        __asm__ volatile(
-            "movd (%1), %%mm0               \n\t"
-            "movd (%1, %2), %%mm1           \n\t"
-            "movd (%1, %2, 2), %%mm2        \n\t"
-            "movd (%1, %3), %%mm3           \n\t"
-            PAVGB" (%0), %%mm0              \n\t"
-            PAVGB" (%0, %2), %%mm1          \n\t"
-            PAVGB" (%0, %2, 2), %%mm2       \n\t"
-            PAVGB" (%0, %3), %%mm3          \n\t"
-            "movd %%mm0, (%1)               \n\t"
-            "movd %%mm1, (%1, %2)           \n\t"
-            "movd %%mm2, (%1, %2, 2)        \n\t"
-            "movd %%mm3, (%1, %3)           \n\t"
-            ::"S"(pixels), "D"(block),
-             "r" ((x86_reg)line_size), "r"((x86_reg)3L*line_size)
-            :"memory");
-        block += 4*line_size;
-        pixels += 4*line_size;
-        h -= 4;
-    } while(h > 0);
-}
-#endif /* SKIP_FOR_3DNOW */
-
 //FIXME the following could be optimized too ...
 static void DEF(put_no_rnd_pixels16_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h){
     DEF(put_no_rnd_pixels8_x2)(block  , pixels  , line_size, h);
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c
index ed6cff3..d403a14 100644
--- a/libavcodec/x86/dsputil_mmx.c
+++ b/libavcodec/x86/dsputil_mmx.c
@@ -366,33 +366,6 @@ void ff_add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels,
     } while (--i);
 }
 
-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"
-        );
-}
-
 static void put_pixels8_mmx(uint8_t *block, const uint8_t *pixels,
                             int line_size, int h)
 {
@@ -455,56 +428,6 @@ static void put_pixels16_mmx(uint8_t *block, const uint8_t *pixels,
         );
 }
 
-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"
-        );
-}
-
-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"
-        );
-}
-
 #define CLEAR_BLOCKS(name, n)                           \
 static void name(DCTELEM *blocks)                       \
 {                                                       \
@@ -2381,27 +2304,23 @@ static void dsputil_init_mmxext(DSPContext *c, AVCodecContext *avctx,
     }
 #endif /* HAVE_INLINE_ASM */
 
+#if HAVE_MMXEXT_EXTERNAL
     if (CONFIG_H264QPEL) {
-#if HAVE_INLINE_ASM
         SET_QPEL_FUNCS(put_qpel,        0, 16, mmxext, );
         SET_QPEL_FUNCS(put_qpel,        1,  8, mmxext, );
         SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, mmxext, );
         SET_QPEL_FUNCS(put_no_rnd_qpel, 1,  8, mmxext, );
         SET_QPEL_FUNCS(avg_qpel,        0, 16, mmxext, );
         SET_QPEL_FUNCS(avg_qpel,        1,  8, mmxext, );
-#endif /* HAVE_INLINE_ASM */
 
         if (!high_bit_depth) {
-#if HAVE_INLINE_ASM
             SET_QPEL_FUNCS(put_h264_qpel, 0, 16, mmxext, );
             SET_QPEL_FUNCS(put_h264_qpel, 1,  8, mmxext, );
             SET_QPEL_FUNCS(put_h264_qpel, 2,  4, mmxext, );
             SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, mmxext, );
             SET_QPEL_FUNCS(avg_h264_qpel, 1,  8, mmxext, );
             SET_QPEL_FUNCS(avg_h264_qpel, 2,  4, mmxext, );
-#endif /* HAVE_INLINE_ASM */
         } 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_);
@@ -2410,18 +2329,14 @@ static void dsputil_init_mmxext(DSPContext *c, AVCodecContext *avctx,
 #endif
             SET_QPEL_FUNCS(put_h264_qpel, 2, 4,  10_mmxext, ff_);
             SET_QPEL_FUNCS(avg_h264_qpel, 2, 4,  10_mmxext, ff_);
-#endif /* HAVE_YASM */
         }
 
-#if HAVE_INLINE_ASM
         SET_QPEL_FUNCS(put_2tap_qpel, 0, 16, mmxext, );
         SET_QPEL_FUNCS(put_2tap_qpel, 1,  8, mmxext, );
         SET_QPEL_FUNCS(avg_2tap_qpel, 0, 16, mmxext, );
         SET_QPEL_FUNCS(avg_2tap_qpel, 1,  8, mmxext, );
-#endif /* HAVE_INLINE_ASM */
     }
 
-#if HAVE_YASM
     if (!high_bit_depth && CONFIG_H264CHROMA) {
         c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_rnd_mmxext;
         c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_mmxext;
@@ -2447,7 +2362,7 @@ static void dsputil_init_mmxext(DSPContext *c, AVCodecContext *avctx,
     } else {
         c->apply_window_int16 = ff_apply_window_int16_round_mmxext;
     }
-#endif /* HAVE_YASM */
+#endif /* HAVE_MMXEXT_EXTERNAL */
 }
 
 static void dsputil_init_3dnow(DSPContext *c, AVCodecContext *avctx,
@@ -2546,17 +2461,16 @@ static void dsputil_init_sse(DSPContext *c, AVCodecContext *avctx, int mm_flags)
 static void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx,
                               int mm_flags)
 {
+#if HAVE_SSE2_EXTERNAL
     const int bit_depth      = avctx->bits_per_raw_sample;
-
-#if HAVE_INLINE_ASM
     const int high_bit_depth = bit_depth > 8;
 
     if (!(mm_flags & AV_CPU_FLAG_SSE2SLOW)) {
         // these functions are slower than mmx on AMD, but faster on Intel
         if (!high_bit_depth) {
-            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;
+            c->put_pixels_tab[0][0]        = ff_put_pixels16_sse2;
+            c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_sse2;
+            c->avg_pixels_tab[0][0]        = ff_avg_pixels16_sse2;
             if (CONFIG_H264QPEL)
                 H264_QPEL_FUNCS(0, 0, sse2);
         }
@@ -2583,9 +2497,7 @@ static void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx,
         c->idct                  = ff_idct_xvid_sse2;
         c->idct_permutation_type = FF_SSE2_IDCT_PERM;
     }
-#endif /* HAVE_INLINE_ASM */
 
-#if HAVE_YASM
     if (bit_depth == 10) {
         if (CONFIG_H264QPEL) {
             SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_sse2, ff_);
@@ -2615,16 +2527,16 @@ static void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx,
         c->apply_window_int16 = ff_apply_window_int16_round_sse2;
     }
     c->bswap_buf = ff_bswap32_buf_sse2;
-#endif /* HAVE_YASM */
+#endif /* HAVE_SSE2_EXTERNAL */
 }
 
 static void dsputil_init_ssse3(DSPContext *c, AVCodecContext *avctx,
                                int mm_flags)
 {
+#if HAVE_SSSE3_EXTERNAL
     const int high_bit_depth = avctx->bits_per_raw_sample > 8;
     const int bit_depth      = avctx->bits_per_raw_sample;
 
-#if HAVE_SSSE3_INLINE
     if (!high_bit_depth && CONFIG_H264QPEL) {
         H264_QPEL_FUNCS(1, 0, ssse3);
         H264_QPEL_FUNCS(1, 1, ssse3);
@@ -2639,9 +2551,6 @@ static void dsputil_init_ssse3(DSPContext *c, AVCodecContext *avctx,
         H264_QPEL_FUNCS(3, 2, ssse3);
         H264_QPEL_FUNCS(3, 3, ssse3);
     }
-#endif /* HAVE_SSSE3_INLINE */
-
-#if HAVE_SSSE3_EXTERNAL
     if (bit_depth == 10 && CONFIG_H264QPEL) {
         H264_QPEL_FUNCS_10(1, 0, ssse3_cache64);
         H264_QPEL_FUNCS_10(2, 0, ssse3_cache64);
diff --git a/libavcodec/x86/h264_qpel.c b/libavcodec/x86/h264_qpel.c
index 075506f..bc56d09 100644
--- a/libavcodec/x86/h264_qpel.c
+++ b/libavcodec/x86/h264_qpel.c
@@ -19,1019 +19,229 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
+#include "libavutil/cpu.h"
+#include "libavutil/x86/asm.h"
+#include "libavcodec/dsputil.h"
+#include "libavcodec/mpegvideo.h"
 #include "dsputil_mmx.h"
 
-#if HAVE_INLINE_ASM
-
-/***********************************/
-/* motion compensation */
-
-#define QPEL_H264V_MM(A,B,C,D,E,F,OP,T,Z,d,q)\
-        "mov"#q" "#C", "#T"         \n\t"\
-        "mov"#d" (%0), "#F"         \n\t"\
-        "paddw "#D", "#T"           \n\t"\
-        "psllw $2, "#T"             \n\t"\
-        "psubw "#B", "#T"           \n\t"\
-        "psubw "#E", "#T"           \n\t"\
-        "punpcklbw "#Z", "#F"       \n\t"\
-        "pmullw "MANGLE(ff_pw_5)", "#T"\n\t"\
-        "paddw "MANGLE(ff_pw_16)", "#A"\n\t"\
-        "add %2, %0                 \n\t"\
-        "paddw "#F", "#A"           \n\t"\
-        "paddw "#A", "#T"           \n\t"\
-        "psraw $5, "#T"             \n\t"\
-        "packuswb "#T", "#T"        \n\t"\
-        OP(T, (%1), A, d)\
-        "add %3, %1                 \n\t"
-
-#define QPEL_H264HV_MM(A,B,C,D,E,F,OF,T,Z,d,q)\
-        "mov"#q" "#C", "#T"         \n\t"\
-        "mov"#d" (%0), "#F"         \n\t"\
-        "paddw "#D", "#T"           \n\t"\
-        "psllw $2, "#T"             \n\t"\
-        "paddw "MANGLE(ff_pw_16)", "#A"\n\t"\
-        "psubw "#B", "#T"           \n\t"\
-        "psubw "#E", "#T"           \n\t"\
-        "punpcklbw "#Z", "#F"       \n\t"\
-        "pmullw "MANGLE(ff_pw_5)", "#T"\n\t"\
-        "paddw "#F", "#A"           \n\t"\
-        "add %2, %0                 \n\t"\
-        "paddw "#A", "#T"           \n\t"\
-        "mov"#q" "#T", "#OF"(%1)    \n\t"
-
-#define QPEL_H264V(A,B,C,D,E,F,OP) QPEL_H264V_MM(A,B,C,D,E,F,OP,%%mm6,%%mm7,d,q)
-#define QPEL_H264HV(A,B,C,D,E,F,OF) QPEL_H264HV_MM(A,B,C,D,E,F,OF,%%mm6,%%mm7,d,q)
-#define QPEL_H264V_XMM(A,B,C,D,E,F,OP) QPEL_H264V_MM(A,B,C,D,E,F,OP,%%xmm6,%%xmm7,q,dqa)
-#define QPEL_H264HV_XMM(A,B,C,D,E,F,OF) QPEL_H264HV_MM(A,B,C,D,E,F,OF,%%xmm6,%%xmm7,q,dqa)
-
+#if HAVE_YASM
+void ff_put_pixels4_mmxext(uint8_t *block, const uint8_t *pixels, int line_size, int h);
+void ff_avg_pixels4_mmxext(uint8_t *block, const uint8_t *pixels, int line_size, int h);
+void ff_put_pixels8_mmxext(uint8_t *block, const uint8_t *pixels, int line_size, int h);
+void ff_avg_pixels8_mmxext(uint8_t *block, const uint8_t *pixels, int line_size, int h);
+static void ff_put_pixels16_mmxext(uint8_t *block, const uint8_t *pixels,
+                                   int line_size, int h)
+{
+    ff_put_pixels8_mmxext(block,     pixels,     line_size, h);
+    ff_put_pixels8_mmxext(block + 8, pixels + 8, line_size, h);
+}
+static void ff_avg_pixels16_mmxext(uint8_t *block, const uint8_t *pixels,
+                                   int line_size, int h)
+{
+    ff_avg_pixels8_mmxext(block,     pixels,     line_size, h);
+    ff_avg_pixels8_mmxext(block + 8, pixels + 8, line_size, h);
+}
+void ff_put_pixels4_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2,
+                              int dstStride, int src1Stride, int h);
+void ff_avg_pixels4_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2,
+                              int dstStride, int src1Stride, int h);
+void ff_put_pixels8_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2,
+                              int dstStride, int src1Stride, int h);
+void ff_avg_pixels8_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2,
+                              int dstStride, int src1Stride, int h);
+void ff_put_pixels16_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2,
+                               int dstStride, int src1Stride, int h);
+void ff_avg_pixels16_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2,
+                               int dstStride, int src1Stride, int h);
+void ff_put_pixels16_sse2(uint8_t *block, const uint8_t *pixels,
+                          int line_size, int h);
+void ff_avg_pixels16_sse2(uint8_t *block, const uint8_t *pixels,
+                          int line_size, int h);
+#define ff_put_pixels8_l2_sse2  ff_put_pixels8_l2_mmxext
+#define ff_avg_pixels8_l2_sse2  ff_avg_pixels8_l2_mmxext
+#define ff_put_pixels16_l2_sse2 ff_put_pixels16_l2_mmxext
+#define ff_avg_pixels16_l2_sse2 ff_avg_pixels16_l2_mmxext
+
+#define DEF_QPEL(OPNAME)\
+void ff_ ## OPNAME ## _h264_qpel4_h_lowpass_mmxext(uint8_t *dst, uint8_t *src, int dstStride, int srcStride);\
+void ff_ ## OPNAME ## _h264_qpel8_h_lowpass_mmxext(uint8_t *dst, uint8_t *src, int dstStride, int srcStride);\
+void ff_ ## OPNAME ## _h264_qpel8_h_lowpass_ssse3(uint8_t *dst, uint8_t *src, int dstStride, int srcStride);\
+void ff_ ## OPNAME ## _h264_qpel4_h_lowpass_l2_mmxext(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride);\
+void ff_ ## OPNAME ## _h264_qpel8_h_lowpass_l2_mmxext(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride);\
+void ff_ ## OPNAME ## _h264_qpel8_h_lowpass_l2_ssse3(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride);\
+void ff_ ## OPNAME ## _h264_qpel4_v_lowpass_mmxext(uint8_t *dst, uint8_t *src, int dstStride, int srcStride);\
+void ff_ ## OPNAME ## _h264_qpel8or16_v_lowpass_op_mmxext(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h);\
+void ff_ ## OPNAME ## _h264_qpel8or16_v_lowpass_sse2(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h);\
+void ff_ ## OPNAME ## _h264_qpel4_hv_lowpass_v_mmxext(uint8_t *src, int16_t *tmp, int srcStride);\
+void ff_ ## OPNAME ## _h264_qpel4_hv_lowpass_h_mmxext(int16_t *tmp, uint8_t *dst, int dstStride);\
+void ff_ ## OPNAME ## _h264_qpel8or16_hv1_lowpass_op_mmxext(uint8_t *src, int16_t *tmp, int srcStride, int size);\
+void ff_ ## OPNAME ## _h264_qpel8or16_hv1_lowpass_op_sse2(uint8_t *src, int16_t *tmp, int srcStride, int size);\
+void ff_ ## OPNAME ## _h264_qpel8or16_hv2_lowpass_op_mmxext(uint8_t *dst, int16_t *tmp, int dstStride, int unused, int h);\
+void ff_ ## OPNAME ## _h264_qpel8or16_hv2_lowpass_ssse3(uint8_t *dst, int16_t *tmp, int dstStride, int tmpStride, int size);\
+void ff_ ## OPNAME ## _pixels4_l2_shift5_mmxext(uint8_t *dst, int16_t *src16, uint8_t *src8, int dstStride, int src8Stride, int h);\
+void ff_ ## OPNAME ## _pixels8_l2_shift5_mmxext(uint8_t *dst, int16_t *src16, uint8_t *src8, int dstStride, int src8Stride, int h);
+
+DEF_QPEL(avg)
+DEF_QPEL(put)
 
 #define QPEL_H264(OPNAME, OP, MMX)\
-static av_noinline void OPNAME ## h264_qpel4_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
-    int h=4;\
-\
-    __asm__ volatile(\
-        "pxor %%mm7, %%mm7          \n\t"\
-        "movq "MANGLE(ff_pw_5) ", %%mm4\n\t"\
-        "movq "MANGLE(ff_pw_16)", %%mm5\n\t"\
-        "1:                         \n\t"\
-        "movd  -1(%0), %%mm1        \n\t"\
-        "movd    (%0), %%mm2        \n\t"\
-        "movd   1(%0), %%mm3        \n\t"\
-        "movd   2(%0), %%mm0        \n\t"\
-        "punpcklbw %%mm7, %%mm1     \n\t"\
-        "punpcklbw %%mm7, %%mm2     \n\t"\
-        "punpcklbw %%mm7, %%mm3     \n\t"\
-        "punpcklbw %%mm7, %%mm0     \n\t"\
-        "paddw %%mm0, %%mm1         \n\t"\
-        "paddw %%mm3, %%mm2         \n\t"\
-        "movd  -2(%0), %%mm0        \n\t"\
-        "movd   3(%0), %%mm3        \n\t"\
-        "punpcklbw %%mm7, %%mm0     \n\t"\
-        "punpcklbw %%mm7, %%mm3     \n\t"\
-        "paddw %%mm3, %%mm0         \n\t"\
-        "psllw $2, %%mm2            \n\t"\
-        "psubw %%mm1, %%mm2         \n\t"\
-        "pmullw %%mm4, %%mm2        \n\t"\
-        "paddw %%mm5, %%mm0         \n\t"\
-        "paddw %%mm2, %%mm0         \n\t"\
-        "psraw $5, %%mm0            \n\t"\
-        "packuswb %%mm0, %%mm0      \n\t"\
-        OP(%%mm0, (%1),%%mm6, d)\
-        "add %3, %0                 \n\t"\
-        "add %4, %1                 \n\t"\
-        "decl %2                    \n\t"\
-        " jnz 1b                    \n\t"\
-        : "+a"(src), "+c"(dst), "+g"(h)\
-        : "d"((x86_reg)srcStride), "S"((x86_reg)dstStride)\
-        : "memory"\
-    );\
-}\
-static av_noinline void OPNAME ## h264_qpel4_h_lowpass_l2_ ## MMX(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride){\
-    int h=4;\
-    __asm__ volatile(\
-        "pxor %%mm7, %%mm7          \n\t"\
-        "movq %0, %%mm4             \n\t"\
-        "movq %1, %%mm5             \n\t"\
-        :: "m"(ff_pw_5), "m"(ff_pw_16)\
-    );\
-    do{\
-    __asm__ volatile(\
-        "movd  -1(%0), %%mm1        \n\t"\
-        "movd    (%0), %%mm2        \n\t"\
-        "movd   1(%0), %%mm3        \n\t"\
-        "movd   2(%0), %%mm0        \n\t"\
-        "punpcklbw %%mm7, %%mm1     \n\t"\
-        "punpcklbw %%mm7, %%mm2     \n\t"\
-        "punpcklbw %%mm7, %%mm3     \n\t"\
-        "punpcklbw %%mm7, %%mm0     \n\t"\
-        "paddw %%mm0, %%mm1         \n\t"\
-        "paddw %%mm3, %%mm2         \n\t"\
-        "movd  -2(%0), %%mm0        \n\t"\
-        "movd   3(%0), %%mm3        \n\t"\
-        "punpcklbw %%mm7, %%mm0     \n\t"\
-        "punpcklbw %%mm7, %%mm3     \n\t"\
-        "paddw %%mm3, %%mm0         \n\t"\
-        "psllw $2, %%mm2            \n\t"\
-        "psubw %%mm1, %%mm2         \n\t"\
-        "pmullw %%mm4, %%mm2        \n\t"\
-        "paddw %%mm5, %%mm0         \n\t"\
-        "paddw %%mm2, %%mm0         \n\t"\
-        "movd   (%2), %%mm3         \n\t"\
-        "psraw $5, %%mm0            \n\t"\
-        "packuswb %%mm0, %%mm0      \n\t"\
-        PAVGB" %%mm3, %%mm0         \n\t"\
-        OP(%%mm0, (%1),%%mm6, d)\
-        "add %4, %0                 \n\t"\
-        "add %4, %1                 \n\t"\
-        "add %3, %2                 \n\t"\
-        : "+a"(src), "+c"(dst), "+d"(src2)\
-        : "D"((x86_reg)src2Stride), "S"((x86_reg)dstStride)\
-        : "memory"\
-    );\
-    }while(--h);\
-}\
-static av_noinline void OPNAME ## h264_qpel4_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
-    src -= 2*srcStride;\
-    __asm__ volatile(\
-        "pxor %%mm7, %%mm7          \n\t"\
-        "movd (%0), %%mm0           \n\t"\
-        "add %2, %0                 \n\t"\
-        "movd (%0), %%mm1           \n\t"\
-        "add %2, %0                 \n\t"\
-        "movd (%0), %%mm2           \n\t"\
-        "add %2, %0                 \n\t"\
-        "movd (%0), %%mm3           \n\t"\
-        "add %2, %0                 \n\t"\
-        "movd (%0), %%mm4           \n\t"\
-        "add %2, %0                 \n\t"\
-        "punpcklbw %%mm7, %%mm0     \n\t"\
-        "punpcklbw %%mm7, %%mm1     \n\t"\
-        "punpcklbw %%mm7, %%mm2     \n\t"\
-        "punpcklbw %%mm7, %%mm3     \n\t"\
-        "punpcklbw %%mm7, %%mm4     \n\t"\
-        QPEL_H264V(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP)\
-        QPEL_H264V(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP)\
-        QPEL_H264V(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP)\
-        QPEL_H264V(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP)\
-         \
-        : "+a"(src), "+c"(dst)\
-        : "S"((x86_reg)srcStride), "D"((x86_reg)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
-        : "memory"\
-    );\
-}\
-static av_noinline void OPNAME ## h264_qpel4_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
-    int h=4;\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel4_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
     int w=3;\
     src -= 2*srcStride+2;\
     while(w--){\
-        __asm__ volatile(\
-            "pxor %%mm7, %%mm7      \n\t"\
-            "movd (%0), %%mm0       \n\t"\
-            "add %2, %0             \n\t"\
-            "movd (%0), %%mm1       \n\t"\
-            "add %2, %0             \n\t"\
-            "movd (%0), %%mm2       \n\t"\
-            "add %2, %0             \n\t"\
-            "movd (%0), %%mm3       \n\t"\
-            "add %2, %0             \n\t"\
-            "movd (%0), %%mm4       \n\t"\
-            "add %2, %0             \n\t"\
-            "punpcklbw %%mm7, %%mm0 \n\t"\
-            "punpcklbw %%mm7, %%mm1 \n\t"\
-            "punpcklbw %%mm7, %%mm2 \n\t"\
-            "punpcklbw %%mm7, %%mm3 \n\t"\
-            "punpcklbw %%mm7, %%mm4 \n\t"\
-            QPEL_H264HV(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, 0*8*3)\
-            QPEL_H264HV(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, 1*8*3)\
-            QPEL_H264HV(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, 2*8*3)\
-            QPEL_H264HV(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, 3*8*3)\
-             \
-            : "+a"(src)\
-            : "c"(tmp), "S"((x86_reg)srcStride)\
-            : "memory"\
-        );\
+        ff_ ## OPNAME ## h264_qpel4_hv_lowpass_v_mmxext(src, tmp, srcStride);\
         tmp += 4;\
-        src += 4 - 9*srcStride;\
+        src += 4;\
     }\
     tmp -= 3*4;\
-    __asm__ volatile(\
-        "1:                         \n\t"\
-        "movq     (%0), %%mm0       \n\t"\
-        "paddw  10(%0), %%mm0       \n\t"\
-        "movq    2(%0), %%mm1       \n\t"\
-        "paddw   8(%0), %%mm1       \n\t"\
-        "movq    4(%0), %%mm2       \n\t"\
-        "paddw   6(%0), %%mm2       \n\t"\
-        "psubw %%mm1, %%mm0         \n\t"/*a-b   (abccba)*/\
-        "psraw $2, %%mm0            \n\t"/*(a-b)/4 */\
-        "psubw %%mm1, %%mm0         \n\t"/*(a-b)/4-b */\
-        "paddsw %%mm2, %%mm0        \n\t"\
-        "psraw $2, %%mm0            \n\t"/*((a-b)/4-b+c)/4 */\
-        "paddw %%mm2, %%mm0         \n\t"/*(a-5*b+20*c)/16 */\
-        "psraw $6, %%mm0            \n\t"\
-        "packuswb %%mm0, %%mm0      \n\t"\
-        OP(%%mm0, (%1),%%mm7, d)\
-        "add $24, %0                \n\t"\
-        "add %3, %1                 \n\t"\
-        "decl %2                    \n\t"\
-        " jnz 1b                    \n\t"\
-        : "+a"(tmp), "+c"(dst), "+g"(h)\
-        : "S"((x86_reg)dstStride)\
-        : "memory"\
-    );\
-}\
-\
-static av_noinline void OPNAME ## h264_qpel8_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
-    int h=8;\
-    __asm__ volatile(\
-        "pxor %%mm7, %%mm7          \n\t"\
-        "movq "MANGLE(ff_pw_5)", %%mm6\n\t"\
-        "1:                         \n\t"\
-        "movq    (%0), %%mm0        \n\t"\
-        "movq   1(%0), %%mm2        \n\t"\
-        "movq %%mm0, %%mm1          \n\t"\
-        "movq %%mm2, %%mm3          \n\t"\
-        "punpcklbw %%mm7, %%mm0     \n\t"\
-        "punpckhbw %%mm7, %%mm1     \n\t"\
-        "punpcklbw %%mm7, %%mm2     \n\t"\
-        "punpckhbw %%mm7, %%mm3     \n\t"\
-        "paddw %%mm2, %%mm0         \n\t"\
-        "paddw %%mm3, %%mm1         \n\t"\
-        "psllw $2, %%mm0            \n\t"\
-        "psllw $2, %%mm1            \n\t"\
-        "movq   -1(%0), %%mm2       \n\t"\
-        "movq    2(%0), %%mm4       \n\t"\
-        "movq %%mm2, %%mm3          \n\t"\
-        "movq %%mm4, %%mm5          \n\t"\
-        "punpcklbw %%mm7, %%mm2     \n\t"\
-        "punpckhbw %%mm7, %%mm3     \n\t"\
-        "punpcklbw %%mm7, %%mm4     \n\t"\
-        "punpckhbw %%mm7, %%mm5     \n\t"\
-        "paddw %%mm4, %%mm2         \n\t"\
-        "paddw %%mm3, %%mm5         \n\t"\
-        "psubw %%mm2, %%mm0         \n\t"\
-        "psubw %%mm5, %%mm1         \n\t"\
-        "pmullw %%mm6, %%mm0        \n\t"\
-        "pmullw %%mm6, %%mm1        \n\t"\
-        "movd   -2(%0), %%mm2       \n\t"\
-        "movd    7(%0), %%mm5       \n\t"\
-        "punpcklbw %%mm7, %%mm2     \n\t"\
-        "punpcklbw %%mm7, %%mm5     \n\t"\
-        "paddw %%mm3, %%mm2         \n\t"\
-        "paddw %%mm5, %%mm4         \n\t"\
-        "movq "MANGLE(ff_pw_16)", %%mm5\n\t"\
-        "paddw %%mm5, %%mm2         \n\t"\
-        "paddw %%mm5, %%mm4         \n\t"\
-        "paddw %%mm2, %%mm0         \n\t"\
-        "paddw %%mm4, %%mm1         \n\t"\
-        "psraw $5, %%mm0            \n\t"\
-        "psraw $5, %%mm1            \n\t"\
-        "packuswb %%mm1, %%mm0      \n\t"\
-        OP(%%mm0, (%1),%%mm5, q)\
-        "add %3, %0                 \n\t"\
-        "add %4, %1                 \n\t"\
-        "decl %2                    \n\t"\
-        " jnz 1b                    \n\t"\
-        : "+a"(src), "+c"(dst), "+g"(h)\
-        : "d"((x86_reg)srcStride), "S"((x86_reg)dstStride)\
-        : "memory"\
-    );\
+    ff_ ## OPNAME ## h264_qpel4_hv_lowpass_h_mmxext(tmp, dst, dstStride);\
 }\
 \
-static av_noinline void OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride){\
-    int h=8;\
-    __asm__ volatile(\
-        "pxor %%mm7, %%mm7          \n\t"\
-        "movq "MANGLE(ff_pw_5)", %%mm6\n\t"\
-        "1:                         \n\t"\
-        "movq    (%0), %%mm0        \n\t"\
-        "movq   1(%0), %%mm2        \n\t"\
-        "movq %%mm0, %%mm1          \n\t"\
-        "movq %%mm2, %%mm3          \n\t"\
-        "punpcklbw %%mm7, %%mm0     \n\t"\
-        "punpckhbw %%mm7, %%mm1     \n\t"\
-        "punpcklbw %%mm7, %%mm2     \n\t"\
-        "punpckhbw %%mm7, %%mm3     \n\t"\
-        "paddw %%mm2, %%mm0         \n\t"\
-        "paddw %%mm3, %%mm1         \n\t"\
-        "psllw $2, %%mm0            \n\t"\
-        "psllw $2, %%mm1            \n\t"\
-        "movq   -1(%0), %%mm2       \n\t"\
-        "movq    2(%0), %%mm4       \n\t"\
-        "movq %%mm2, %%mm3          \n\t"\
-        "movq %%mm4, %%mm5          \n\t"\
-        "punpcklbw %%mm7, %%mm2     \n\t"\
-        "punpckhbw %%mm7, %%mm3     \n\t"\
-        "punpcklbw %%mm7, %%mm4     \n\t"\
-        "punpckhbw %%mm7, %%mm5     \n\t"\
-        "paddw %%mm4, %%mm2         \n\t"\
-        "paddw %%mm3, %%mm5         \n\t"\
-        "psubw %%mm2, %%mm0         \n\t"\
-        "psubw %%mm5, %%mm1         \n\t"\
-        "pmullw %%mm6, %%mm0        \n\t"\
-        "pmullw %%mm6, %%mm1        \n\t"\
-        "movd   -2(%0), %%mm2       \n\t"\
-        "movd    7(%0), %%mm5       \n\t"\
-        "punpcklbw %%mm7, %%mm2     \n\t"\
-        "punpcklbw %%mm7, %%mm5     \n\t"\
-        "paddw %%mm3, %%mm2         \n\t"\
-        "paddw %%mm5, %%mm4         \n\t"\
-        "movq "MANGLE(ff_pw_16)", %%mm5\n\t"\
-        "paddw %%mm5, %%mm2         \n\t"\
-        "paddw %%mm5, %%mm4         \n\t"\
-        "paddw %%mm2, %%mm0         \n\t"\
-        "paddw %%mm4, %%mm1         \n\t"\
-        "psraw $5, %%mm0            \n\t"\
-        "psraw $5, %%mm1            \n\t"\
-        "movq (%2), %%mm4           \n\t"\
-        "packuswb %%mm1, %%mm0      \n\t"\
-        PAVGB" %%mm4, %%mm0         \n\t"\
-        OP(%%mm0, (%1),%%mm5, q)\
-        "add %5, %0                 \n\t"\
-        "add %5, %1                 \n\t"\
-        "add %4, %2                 \n\t"\
-        "decl %3                    \n\t"\
-        "jg 1b                      \n\t"\
-        : "+a"(src), "+c"(dst), "+d"(src2), "+g"(h)\
-        : "D"((x86_reg)src2Stride), "S"((x86_reg)dstStride)\
-        : "memory"\
-    );\
-}\
-\
-static av_noinline void OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
-    int w= 2;\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
     src -= 2*srcStride;\
-    \
-    while(w--){\
-        __asm__ volatile(\
-            "pxor %%mm7, %%mm7          \n\t"\
-            "movd (%0), %%mm0           \n\t"\
-            "add %2, %0                 \n\t"\
-            "movd (%0), %%mm1           \n\t"\
-            "add %2, %0                 \n\t"\
-            "movd (%0), %%mm2           \n\t"\
-            "add %2, %0                 \n\t"\
-            "movd (%0), %%mm3           \n\t"\
-            "add %2, %0                 \n\t"\
-            "movd (%0), %%mm4           \n\t"\
-            "add %2, %0                 \n\t"\
-            "punpcklbw %%mm7, %%mm0     \n\t"\
-            "punpcklbw %%mm7, %%mm1     \n\t"\
-            "punpcklbw %%mm7, %%mm2     \n\t"\
-            "punpcklbw %%mm7, %%mm3     \n\t"\
-            "punpcklbw %%mm7, %%mm4     \n\t"\
-            QPEL_H264V(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP)\
-            QPEL_H264V(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP)\
-            QPEL_H264V(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP)\
-            QPEL_H264V(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP)\
-            QPEL_H264V(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, OP)\
-            QPEL_H264V(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, OP)\
-            QPEL_H264V(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP)\
-            QPEL_H264V(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP)\
-            "cmpl $16, %4               \n\t"\
-            "jne 2f                     \n\t"\
-            QPEL_H264V(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP)\
-            QPEL_H264V(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP)\
-            QPEL_H264V(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, OP)\
-            QPEL_H264V(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, OP)\
-            QPEL_H264V(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP)\
-            QPEL_H264V(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP)\
-            QPEL_H264V(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP)\
-            QPEL_H264V(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP)\
-            "2:                         \n\t"\
-            \
-            : "+a"(src), "+c"(dst)\
-            : "S"((x86_reg)srcStride), "D"((x86_reg)dstStride), "rm"(h)\
-            : "memory"\
-        );\
-        src += 4-(h+5)*srcStride;\
-        dst += 4-h*dstStride;\
-    }\
+    ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_op_mmxext(dst, src, dstStride, srcStride, h);\
+    src += 4;\
+    dst += 4;\
+    ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_op_mmxext(dst, src, dstStride, srcStride, h);\
 }\
-static av_always_inline void OPNAME ## h264_qpel8or16_hv1_lowpass_ ## MMX(int16_t *tmp, uint8_t *src, int tmpStride, int srcStride, int size){\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel8or16_hv1_lowpass_ ## MMX(int16_t *tmp, uint8_t *src, int tmpStride, int srcStride, int size){\
     int w = (size+8)>>2;\
     src -= 2*srcStride+2;\
     while(w--){\
-        __asm__ volatile(\
-            "pxor %%mm7, %%mm7      \n\t"\
-            "movd (%0), %%mm0       \n\t"\
-            "add %2, %0             \n\t"\
-            "movd (%0), %%mm1       \n\t"\
-            "add %2, %0             \n\t"\
-            "movd (%0), %%mm2       \n\t"\
-            "add %2, %0             \n\t"\
-            "movd (%0), %%mm3       \n\t"\
-            "add %2, %0             \n\t"\
-            "movd (%0), %%mm4       \n\t"\
-            "add %2, %0             \n\t"\
-            "punpcklbw %%mm7, %%mm0 \n\t"\
-            "punpcklbw %%mm7, %%mm1 \n\t"\
-            "punpcklbw %%mm7, %%mm2 \n\t"\
-            "punpcklbw %%mm7, %%mm3 \n\t"\
-            "punpcklbw %%mm7, %%mm4 \n\t"\
-            QPEL_H264HV(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, 0*48)\
-            QPEL_H264HV(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, 1*48)\
-            QPEL_H264HV(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, 2*48)\
-            QPEL_H264HV(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, 3*48)\
-            QPEL_H264HV(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, 4*48)\
-            QPEL_H264HV(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, 5*48)\
-            QPEL_H264HV(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, 6*48)\
-            QPEL_H264HV(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, 7*48)\
-            "cmpl $16, %3           \n\t"\
-            "jne 2f                 \n\t"\
-            QPEL_H264HV(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1,  8*48)\
-            QPEL_H264HV(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2,  9*48)\
-            QPEL_H264HV(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, 10*48)\
-            QPEL_H264HV(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, 11*48)\
-            QPEL_H264HV(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, 12*48)\
-            QPEL_H264HV(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, 13*48)\
-            QPEL_H264HV(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, 14*48)\
-            QPEL_H264HV(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, 15*48)\
-            "2:                     \n\t"\
-            : "+a"(src)\
-            : "c"(tmp), "S"((x86_reg)srcStride), "rm"(size)\
-            : "memory"\
-            );\
+        ff_ ## OPNAME ## h264_qpel8or16_hv1_lowpass_op_mmxext(src, tmp, srcStride, size);\
         tmp += 4;\
-        src += 4 - (size+5)*srcStride;\
+        src += 4;\
     }\
 }\
-static av_always_inline void OPNAME ## h264_qpel8or16_hv2_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, int dstStride, int tmpStride, int size){\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel8or16_hv2_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, int dstStride, int tmpStride, int size){\
     int w = size>>4;\
     do{\
-    int h = size;\
-    __asm__ volatile(\
-        "1:                         \n\t"\
-        "movq     (%0), %%mm0       \n\t"\
-        "movq    8(%0), %%mm3       \n\t"\
-        "movq    2(%0), %%mm1       \n\t"\
-        "movq   10(%0), %%mm4       \n\t"\
-        "paddw   %%mm4, %%mm0       \n\t"\
-        "paddw   %%mm3, %%mm1       \n\t"\
-        "paddw  18(%0), %%mm3       \n\t"\
-        "paddw  16(%0), %%mm4       \n\t"\
-        "movq    4(%0), %%mm2       \n\t"\
-        "movq   12(%0), %%mm5       \n\t"\
-        "paddw   6(%0), %%mm2       \n\t"\
-        "paddw  14(%0), %%mm5       \n\t"\
-        "psubw %%mm1, %%mm0         \n\t"\
-        "psubw %%mm4, %%mm3         \n\t"\
-        "psraw $2, %%mm0            \n\t"\
-        "psraw $2, %%mm3            \n\t"\
-        "psubw %%mm1, %%mm0         \n\t"\
-        "psubw %%mm4, %%mm3         \n\t"\
-        "paddsw %%mm2, %%mm0        \n\t"\
-        "paddsw %%mm5, %%mm3        \n\t"\
-        "psraw $2, %%mm0            \n\t"\
-        "psraw $2, %%mm3            \n\t"\
-        "paddw %%mm2, %%mm0         \n\t"\
-        "paddw %%mm5, %%mm3         \n\t"\
-        "psraw $6, %%mm0            \n\t"\
-        "psraw $6, %%mm3            \n\t"\
-        "packuswb %%mm3, %%mm0      \n\t"\
-        OP(%%mm0, (%1),%%mm7, q)\
-        "add $48, %0                \n\t"\
-        "add %3, %1                 \n\t"\
-        "decl %2                    \n\t"\
-        " jnz 1b                    \n\t"\
-        : "+a"(tmp), "+c"(dst), "+g"(h)\
-        : "S"((x86_reg)dstStride)\
-        : "memory"\
-    );\
-    tmp += 8 - size*24;\
-    dst += 8 - size*dstStride;\
+    ff_ ## OPNAME ## h264_qpel8or16_hv2_lowpass_op_mmxext(dst, tmp, dstStride, 0, size);\
+    tmp += 8;\
+    dst += 8;\
     }while(w--);\
 }\
 \
-static void OPNAME ## h264_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
-    OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 8);\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
+    ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 8);\
 }\
-static av_noinline void OPNAME ## h264_qpel16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
-    OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 16);\
-    OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
+    ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 16);\
+    ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
 }\
 \
-static void OPNAME ## h264_qpel16_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
-    OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
-    OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel16_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
+    ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
+    ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
     src += 8*srcStride;\
     dst += 8*dstStride;\
-    OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
-    OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
+    ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
+    ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
 }\
 \
-static av_noinline void OPNAME ## h264_qpel16_h_lowpass_l2_ ## MMX(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride){\
-    OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
-    OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel16_h_lowpass_l2_ ## MMX(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride){\
+    ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
+    ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
     src += 8*dstStride;\
     dst += 8*dstStride;\
     src2 += 8*src2Stride;\
-    OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
-    OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
+    ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
+    ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
 }\
 \
-static av_noinline void OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride, int size){\
-          put_h264_qpel8or16_hv1_lowpass_ ## MMX(tmp, src, tmpStride, srcStride, size);\
-    OPNAME ## h264_qpel8or16_hv2_lowpass_ ## MMX(dst, tmp, dstStride, tmpStride, size);\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride, int size){\
+    ff_put_h264_qpel8or16_hv1_lowpass_ ## MMX(tmp, src, tmpStride, srcStride, size);\
+    ff_ ## OPNAME ## h264_qpel8or16_hv2_lowpass_ ## MMX(dst, tmp, dstStride, tmpStride, size);\
 }\
-static void OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
-    OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst  , tmp  , src  , dstStride, tmpStride, srcStride, 8);\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
+    ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst  , tmp  , src  , dstStride, tmpStride, srcStride, 8);\
 }\
 \
-static void OPNAME ## h264_qpel16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
-    OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst  , tmp  , src  , dstStride, tmpStride, srcStride, 16);\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
+    ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst  , tmp  , src  , dstStride, tmpStride, srcStride, 16);\
 }\
 \
-static av_noinline void OPNAME ## pixels4_l2_shift5_ ## MMX(uint8_t *dst, int16_t *src16, uint8_t *src8, int dstStride, int src8Stride, int h)\
-{\
-    __asm__ volatile(\
-        "movq      (%1), %%mm0          \n\t"\
-        "movq    24(%1), %%mm1          \n\t"\
-        "psraw      $5,  %%mm0          \n\t"\
-        "psraw      $5,  %%mm1          \n\t"\
-        "packuswb %%mm0, %%mm0          \n\t"\
-        "packuswb %%mm1, %%mm1          \n\t"\
-        PAVGB"     (%0), %%mm0          \n\t"\
-        PAVGB"  (%0,%3), %%mm1          \n\t"\
-        OP(%%mm0, (%2),    %%mm4, d)\
-        OP(%%mm1, (%2,%4), %%mm5, d)\
-        "lea  (%0,%3,2), %0             \n\t"\
-        "lea  (%2,%4,2), %2             \n\t"\
-        "movq    48(%1), %%mm0          \n\t"\
-        "movq    72(%1), %%mm1          \n\t"\
-        "psraw      $5,  %%mm0          \n\t"\
-        "psraw      $5,  %%mm1          \n\t"\
-        "packuswb %%mm0, %%mm0          \n\t"\
-        "packuswb %%mm1, %%mm1          \n\t"\
-        PAVGB"     (%0), %%mm0          \n\t"\
-        PAVGB"  (%0,%3), %%mm1          \n\t"\
-        OP(%%mm0, (%2),    %%mm4, d)\
-        OP(%%mm1, (%2,%4), %%mm5, d)\
-        :"+a"(src8), "+c"(src16), "+d"(dst)\
-        :"S"((x86_reg)src8Stride), "D"((x86_reg)dstStride)\
-        :"memory");\
-}\
-static av_noinline void OPNAME ## pixels8_l2_shift5_ ## MMX(uint8_t *dst, int16_t *src16, uint8_t *src8, int dstStride, int src8Stride, int h)\
+static av_always_inline void ff_ ## OPNAME ## pixels16_l2_shift5_ ## MMX(uint8_t *dst, int16_t *src16, uint8_t *src8, int dstStride, int src8Stride, int h)\
 {\
-    do{\
-    __asm__ volatile(\
-        "movq      (%1), %%mm0          \n\t"\
-        "movq     8(%1), %%mm1          \n\t"\
-        "movq    48(%1), %%mm2          \n\t"\
-        "movq  8+48(%1), %%mm3          \n\t"\
-        "psraw      $5,  %%mm0          \n\t"\
-        "psraw      $5,  %%mm1          \n\t"\
-        "psraw      $5,  %%mm2          \n\t"\
-        "psraw      $5,  %%mm3          \n\t"\
-        "packuswb %%mm1, %%mm0          \n\t"\
-        "packuswb %%mm3, %%mm2          \n\t"\
-        PAVGB"     (%0), %%mm0          \n\t"\
-        PAVGB"  (%0,%3), %%mm2          \n\t"\
-        OP(%%mm0, (%2), %%mm5, q)\
-        OP(%%mm2, (%2,%4), %%mm5, q)\
-        ::"a"(src8), "c"(src16), "d"(dst),\
-          "r"((x86_reg)src8Stride), "r"((x86_reg)dstStride)\
-        :"memory");\
-        src8 += 2L*src8Stride;\
-        src16 += 48;\
-        dst += 2L*dstStride;\
-    }while(h-=2);\
-}\
-static void OPNAME ## pixels16_l2_shift5_ ## MMX(uint8_t *dst, int16_t *src16, uint8_t *src8, int dstStride, int src8Stride, int h)\
-{\
-    OPNAME ## pixels8_l2_shift5_ ## MMX(dst  , src16  , src8  , dstStride, src8Stride, h);\
-    OPNAME ## pixels8_l2_shift5_ ## MMX(dst+8, src16+8, src8+8, dstStride, src8Stride, h);\
+    ff_ ## OPNAME ## pixels8_l2_shift5_ ## MMX(dst  , src16  , src8  , dstStride, src8Stride, h);\
+    ff_ ## OPNAME ## pixels8_l2_shift5_ ## MMX(dst+8, src16+8, src8+8, dstStride, src8Stride, h);\
 }\
 
 
 #if ARCH_X86_64
 #define QPEL_H264_H16_XMM(OPNAME, OP, MMX)\
-static av_noinline void OPNAME ## h264_qpel16_h_lowpass_l2_ ## MMX(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride){\
-    int h=16;\
-    __asm__ volatile(\
-        "pxor %%xmm15, %%xmm15      \n\t"\
-        "movdqa %6, %%xmm14         \n\t"\
-        "movdqa %7, %%xmm13         \n\t"\
-        "1:                         \n\t"\
-        "lddqu    6(%0), %%xmm1     \n\t"\
-        "lddqu   -2(%0), %%xmm7     \n\t"\
-        "movdqa  %%xmm1, %%xmm0     \n\t"\
-        "punpckhbw %%xmm15, %%xmm1  \n\t"\
-        "punpcklbw %%xmm15, %%xmm0  \n\t"\
-        "punpcklbw %%xmm15, %%xmm7  \n\t"\
-        "movdqa  %%xmm1, %%xmm2     \n\t"\
-        "movdqa  %%xmm0, %%xmm6     \n\t"\
-        "movdqa  %%xmm1, %%xmm3     \n\t"\
-        "movdqa  %%xmm0, %%xmm8     \n\t"\
-        "movdqa  %%xmm1, %%xmm4     \n\t"\
-        "movdqa  %%xmm0, %%xmm9     \n\t"\
-        "movdqa  %%xmm0, %%xmm12    \n\t"\
-        "movdqa  %%xmm1, %%xmm11    \n\t"\
-        "palignr $10,%%xmm0, %%xmm11\n\t"\
-        "palignr $10,%%xmm7, %%xmm12\n\t"\
-        "palignr $2, %%xmm0, %%xmm4 \n\t"\
-        "palignr $2, %%xmm7, %%xmm9 \n\t"\
-        "palignr $4, %%xmm0, %%xmm3 \n\t"\
-        "palignr $4, %%xmm7, %%xmm8 \n\t"\
-        "palignr $6, %%xmm0, %%xmm2 \n\t"\
-        "palignr $6, %%xmm7, %%xmm6 \n\t"\
-        "paddw   %%xmm0 ,%%xmm11    \n\t"\
-        "palignr $8, %%xmm0, %%xmm1 \n\t"\
-        "palignr $8, %%xmm7, %%xmm0 \n\t"\
-        "paddw   %%xmm12,%%xmm7     \n\t"\
-        "paddw   %%xmm3, %%xmm2     \n\t"\
-        "paddw   %%xmm8, %%xmm6     \n\t"\
-        "paddw   %%xmm4, %%xmm1     \n\t"\
-        "paddw   %%xmm9, %%xmm0     \n\t"\
-        "psllw   $2,     %%xmm2     \n\t"\
-        "psllw   $2,     %%xmm6     \n\t"\
-        "psubw   %%xmm1, %%xmm2     \n\t"\
-        "psubw   %%xmm0, %%xmm6     \n\t"\
-        "paddw   %%xmm13,%%xmm11    \n\t"\
-        "paddw   %%xmm13,%%xmm7     \n\t"\
-        "pmullw  %%xmm14,%%xmm2     \n\t"\
-        "pmullw  %%xmm14,%%xmm6     \n\t"\
-        "lddqu   (%2),   %%xmm3     \n\t"\
-        "paddw   %%xmm11,%%xmm2     \n\t"\
-        "paddw   %%xmm7, %%xmm6     \n\t"\
-        "psraw   $5,     %%xmm2     \n\t"\
-        "psraw   $5,     %%xmm6     \n\t"\
-        "packuswb %%xmm2,%%xmm6     \n\t"\
-        "pavgb   %%xmm3, %%xmm6     \n\t"\
-        OP(%%xmm6, (%1), %%xmm4, dqa)\
-        "add %5, %0                 \n\t"\
-        "add %5, %1                 \n\t"\
-        "add %4, %2                 \n\t"\
-        "decl %3                    \n\t"\
-        "jg 1b                      \n\t"\
-        : "+a"(src), "+c"(dst), "+d"(src2), "+g"(h)\
-        : "D"((x86_reg)src2Stride), "S"((x86_reg)dstStride),\
-          "m"(ff_pw_5), "m"(ff_pw_16)\
-        : XMM_CLOBBERS("%xmm0" , "%xmm1" , "%xmm2" , "%xmm3" , \
-                       "%xmm4" , "%xmm5" , "%xmm6" , "%xmm7" , \
-                       "%xmm8" , "%xmm9" , "%xmm10", "%xmm11", \
-                       "%xmm12", "%xmm13", "%xmm14", "%xmm15",)\
-          "memory"\
-    );\
-}
+
+void ff_avg_h264_qpel16_h_lowpass_l2_ssse3(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride);
+void ff_put_h264_qpel16_h_lowpass_l2_ssse3(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride);
+
 #else // ARCH_X86_64
 #define QPEL_H264_H16_XMM(OPNAME, OP, MMX)\
-static av_noinline void OPNAME ## h264_qpel16_h_lowpass_l2_ ## MMX(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride){\
-    OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
-    OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel16_h_lowpass_l2_ ## MMX(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride){\
+    ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
+    ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
     src += 8*dstStride;\
     dst += 8*dstStride;\
     src2 += 8*src2Stride;\
-    OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
-    OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
+    ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
+    ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
 }
 #endif // ARCH_X86_64
 
 #define QPEL_H264_H_XMM(OPNAME, OP, MMX)\
-static av_noinline void OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride){\
-    int h=8;\
-    __asm__ volatile(\
-        "pxor %%xmm7, %%xmm7        \n\t"\
-        "movdqa "MANGLE(ff_pw_5)", %%xmm6\n\t"\
-        "1:                         \n\t"\
-        "lddqu   -2(%0), %%xmm1     \n\t"\
-        "movdqa  %%xmm1, %%xmm0     \n\t"\
-        "punpckhbw %%xmm7, %%xmm1   \n\t"\
-        "punpcklbw %%xmm7, %%xmm0   \n\t"\
-        "movdqa  %%xmm1, %%xmm2     \n\t"\
-        "movdqa  %%xmm1, %%xmm3     \n\t"\
-        "movdqa  %%xmm1, %%xmm4     \n\t"\
-        "movdqa  %%xmm1, %%xmm5     \n\t"\
-        "palignr $2, %%xmm0, %%xmm4 \n\t"\
-        "palignr $4, %%xmm0, %%xmm3 \n\t"\
-        "palignr $6, %%xmm0, %%xmm2 \n\t"\
-        "palignr $8, %%xmm0, %%xmm1 \n\t"\
-        "palignr $10,%%xmm0, %%xmm5 \n\t"\
-        "paddw   %%xmm5, %%xmm0     \n\t"\
-        "paddw   %%xmm3, %%xmm2     \n\t"\
-        "paddw   %%xmm4, %%xmm1     \n\t"\
-        "psllw   $2,     %%xmm2     \n\t"\
-        "movq    (%2),   %%xmm3     \n\t"\
-        "psubw   %%xmm1, %%xmm2     \n\t"\
-        "paddw "MANGLE(ff_pw_16)", %%xmm0\n\t"\
-        "pmullw  %%xmm6, %%xmm2     \n\t"\
-        "paddw   %%xmm0, %%xmm2     \n\t"\
-        "psraw   $5,     %%xmm2     \n\t"\
-        "packuswb %%xmm2, %%xmm2    \n\t"\
-        "pavgb   %%xmm3, %%xmm2     \n\t"\
-        OP(%%xmm2, (%1), %%xmm4, q)\
-        "add %5, %0                 \n\t"\
-        "add %5, %1                 \n\t"\
-        "add %4, %2                 \n\t"\
-        "decl %3                    \n\t"\
-        "jg 1b                      \n\t"\
-        : "+a"(src), "+c"(dst), "+d"(src2), "+g"(h)\
-        : "D"((x86_reg)src2Stride), "S"((x86_reg)dstStride)\
-        : XMM_CLOBBERS("%xmm0", "%xmm1", "%xmm2", "%xmm3", \
-                       "%xmm4", "%xmm5", "%xmm6", "%xmm7",)\
-          "memory"\
-    );\
-}\
 QPEL_H264_H16_XMM(OPNAME, OP, MMX)\
-\
-static av_noinline void OPNAME ## h264_qpel8_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
-    int h=8;\
-    __asm__ volatile(\
-        "pxor %%xmm7, %%xmm7        \n\t"\
-        "movdqa "MANGLE(ff_pw_5)", %%xmm6\n\t"\
-        "1:                         \n\t"\
-        "lddqu   -2(%0), %%xmm1     \n\t"\
-        "movdqa  %%xmm1, %%xmm0     \n\t"\
-        "punpckhbw %%xmm7, %%xmm1   \n\t"\
-        "punpcklbw %%xmm7, %%xmm0   \n\t"\
-        "movdqa  %%xmm1, %%xmm2     \n\t"\
-        "movdqa  %%xmm1, %%xmm3     \n\t"\
-        "movdqa  %%xmm1, %%xmm4     \n\t"\
-        "movdqa  %%xmm1, %%xmm5     \n\t"\
-        "palignr $2, %%xmm0, %%xmm4 \n\t"\
-        "palignr $4, %%xmm0, %%xmm3 \n\t"\
-        "palignr $6, %%xmm0, %%xmm2 \n\t"\
-        "palignr $8, %%xmm0, %%xmm1 \n\t"\
-        "palignr $10,%%xmm0, %%xmm5 \n\t"\
-        "paddw   %%xmm5, %%xmm0     \n\t"\
-        "paddw   %%xmm3, %%xmm2     \n\t"\
-        "paddw   %%xmm4, %%xmm1     \n\t"\
-        "psllw   $2,     %%xmm2     \n\t"\
-        "psubw   %%xmm1, %%xmm2     \n\t"\
-        "paddw   "MANGLE(ff_pw_16)", %%xmm0\n\t"\
-        "pmullw  %%xmm6, %%xmm2     \n\t"\
-        "paddw   %%xmm0, %%xmm2     \n\t"\
-        "psraw   $5,     %%xmm2     \n\t"\
-        "packuswb %%xmm2, %%xmm2    \n\t"\
-        OP(%%xmm2, (%1), %%xmm4, q)\
-        "add %3, %0                 \n\t"\
-        "add %4, %1                 \n\t"\
-        "decl %2                    \n\t"\
-        " jnz 1b                    \n\t"\
-        : "+a"(src), "+c"(dst), "+g"(h)\
-        : "D"((x86_reg)srcStride), "S"((x86_reg)dstStride)\
-        : XMM_CLOBBERS("%xmm0", "%xmm1", "%xmm2", "%xmm3", \
-                       "%xmm4", "%xmm5", "%xmm6", "%xmm7",)\
-          "memory"\
-    );\
-}\
-static void OPNAME ## h264_qpel16_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
-    OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
-    OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel16_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
+    ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
+    ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
     src += 8*srcStride;\
     dst += 8*dstStride;\
-    OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
-    OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
+    ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
+    ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
 }\
 
 #define QPEL_H264_V_XMM(OPNAME, OP, MMX)\
-static av_noinline void OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
-    src -= 2*srcStride;\
-    \
-    __asm__ volatile(\
-        "pxor %%xmm7, %%xmm7        \n\t"\
-        "movq (%0), %%xmm0          \n\t"\
-        "add %2, %0                 \n\t"\
-        "movq (%0), %%xmm1          \n\t"\
-        "add %2, %0                 \n\t"\
-        "movq (%0), %%xmm2          \n\t"\
-        "add %2, %0                 \n\t"\
-        "movq (%0), %%xmm3          \n\t"\
-        "add %2, %0                 \n\t"\
-        "movq (%0), %%xmm4          \n\t"\
-        "add %2, %0                 \n\t"\
-        "punpcklbw %%xmm7, %%xmm0   \n\t"\
-        "punpcklbw %%xmm7, %%xmm1   \n\t"\
-        "punpcklbw %%xmm7, %%xmm2   \n\t"\
-        "punpcklbw %%xmm7, %%xmm3   \n\t"\
-        "punpcklbw %%xmm7, %%xmm4   \n\t"\
-        QPEL_H264V_XMM(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, OP)\
-        QPEL_H264V_XMM(%%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, OP)\
-        QPEL_H264V_XMM(%%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, OP)\
-        QPEL_H264V_XMM(%%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, OP)\
-        QPEL_H264V_XMM(%%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, %%xmm3, OP)\
-        QPEL_H264V_XMM(%%xmm5, %%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, OP)\
-        QPEL_H264V_XMM(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, OP)\
-        QPEL_H264V_XMM(%%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, OP)\
-        "cmpl $16, %4               \n\t"\
-        "jne 2f                     \n\t"\
-        QPEL_H264V_XMM(%%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, OP)\
-        QPEL_H264V_XMM(%%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, OP)\
-        QPEL_H264V_XMM(%%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, %%xmm3, OP)\
-        QPEL_H264V_XMM(%%xmm5, %%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, OP)\
-        QPEL_H264V_XMM(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, OP)\
-        QPEL_H264V_XMM(%%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, OP)\
-        QPEL_H264V_XMM(%%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, OP)\
-        QPEL_H264V_XMM(%%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, OP)\
-        "2:                          \n\t"\
-        \
-        : "+a"(src), "+c"(dst)\
-        : "S"((x86_reg)srcStride), "D"((x86_reg)dstStride), "rm"(h)\
-        : XMM_CLOBBERS("%xmm0", "%xmm1", "%xmm2", "%xmm3", \
-                       "%xmm4", "%xmm5", "%xmm6", "%xmm7",)\
-          "memory"\
-    );\
-}\
-static void OPNAME ## h264_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
-    OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 8);\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
+    ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 8);\
 }\
-static av_noinline void OPNAME ## h264_qpel16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
-    OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 16);\
-    OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
+    ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 16);\
+    ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
 }
 
-static av_always_inline void put_h264_qpel8or16_hv1_lowpass_sse2(int16_t *tmp, uint8_t *src, int tmpStride, int srcStride, int size){
+static av_always_inline void ff_put_h264_qpel8or16_hv1_lowpass_sse2(int16_t *tmp, uint8_t *src, int tmpStride, int srcStride, int size){
     int w = (size+8)>>3;
     src -= 2*srcStride+2;
     while(w--){
-        __asm__ volatile(
-            "pxor %%xmm7, %%xmm7        \n\t"
-            "movq (%0), %%xmm0          \n\t"
-            "add %2, %0                 \n\t"
-            "movq (%0), %%xmm1          \n\t"
-            "add %2, %0                 \n\t"
-            "movq (%0), %%xmm2          \n\t"
-            "add %2, %0                 \n\t"
-            "movq (%0), %%xmm3          \n\t"
-            "add %2, %0                 \n\t"
-            "movq (%0), %%xmm4          \n\t"
-            "add %2, %0                 \n\t"
-            "punpcklbw %%xmm7, %%xmm0   \n\t"
-            "punpcklbw %%xmm7, %%xmm1   \n\t"
-            "punpcklbw %%xmm7, %%xmm2   \n\t"
-            "punpcklbw %%xmm7, %%xmm3   \n\t"
-            "punpcklbw %%xmm7, %%xmm4   \n\t"
-            QPEL_H264HV_XMM(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, 0*48)
-            QPEL_H264HV_XMM(%%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, 1*48)
-            QPEL_H264HV_XMM(%%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, 2*48)
-            QPEL_H264HV_XMM(%%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, 3*48)
-            QPEL_H264HV_XMM(%%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, %%xmm3, 4*48)
-            QPEL_H264HV_XMM(%%xmm5, %%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, 5*48)
-            QPEL_H264HV_XMM(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, 6*48)
-            QPEL_H264HV_XMM(%%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, 7*48)
-            "cmpl $16, %3               \n\t"
-            "jne 2f                     \n\t"
-            QPEL_H264HV_XMM(%%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1,  8*48)
-            QPEL_H264HV_XMM(%%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2,  9*48)
-            QPEL_H264HV_XMM(%%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, %%xmm3, 10*48)
-            QPEL_H264HV_XMM(%%xmm5, %%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, 11*48)
-            QPEL_H264HV_XMM(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, 12*48)
-            QPEL_H264HV_XMM(%%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, 13*48)
-            QPEL_H264HV_XMM(%%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, 14*48)
-            QPEL_H264HV_XMM(%%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, 15*48)
-            "2:                         \n\t"
-            : "+a"(src)
-            : "c"(tmp), "S"((x86_reg)srcStride), "rm"(size)
-            : XMM_CLOBBERS("%xmm0", "%xmm1", "%xmm2", "%xmm3",
-                           "%xmm4", "%xmm5", "%xmm6", "%xmm7",)
-              "memory"
-        );
+        ff_put_h264_qpel8or16_hv1_lowpass_op_sse2(src, tmp, srcStride, size);
         tmp += 8;
-        src += 8 - (size+5)*srcStride;
+        src += 8;
     }
 }
 
-#define QPEL_H264_HV2_XMM(OPNAME, OP, MMX)\
-static av_always_inline void OPNAME ## h264_qpel8or16_hv2_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, int dstStride, int tmpStride, int size){\
-    int h = size;\
-    if(size == 16){\
-        __asm__ volatile(\
-            "1:                         \n\t"\
-            "movdqa 32(%0), %%xmm4      \n\t"\
-            "movdqa 16(%0), %%xmm5      \n\t"\
-            "movdqa   (%0), %%xmm7      \n\t"\
-            "movdqa %%xmm4, %%xmm3      \n\t"\
-            "movdqa %%xmm4, %%xmm2      \n\t"\
-            "movdqa %%xmm4, %%xmm1      \n\t"\
-            "movdqa %%xmm4, %%xmm0      \n\t"\
-            "palignr $10, %%xmm5, %%xmm0 \n\t"\
-            "palignr  $8, %%xmm5, %%xmm1 \n\t"\
-            "palignr  $6, %%xmm5, %%xmm2 \n\t"\
-            "palignr  $4, %%xmm5, %%xmm3 \n\t"\
-            "palignr  $2, %%xmm5, %%xmm4 \n\t"\
-            "paddw  %%xmm5, %%xmm0      \n\t"\
-            "paddw  %%xmm4, %%xmm1      \n\t"\
-            "paddw  %%xmm3, %%xmm2      \n\t"\
-            "movdqa %%xmm5, %%xmm6      \n\t"\
-            "movdqa %%xmm5, %%xmm4      \n\t"\
-            "movdqa %%xmm5, %%xmm3      \n\t"\
-            "palignr  $8, %%xmm7, %%xmm4 \n\t"\
-            "palignr  $2, %%xmm7, %%xmm6 \n\t"\
-            "palignr $10, %%xmm7, %%xmm3 \n\t"\
-            "paddw  %%xmm6, %%xmm4      \n\t"\
-            "movdqa %%xmm5, %%xmm6      \n\t"\
-            "palignr  $6, %%xmm7, %%xmm5 \n\t"\
-            "palignr  $4, %%xmm7, %%xmm6 \n\t"\
-            "paddw  %%xmm7, %%xmm3      \n\t"\
-            "paddw  %%xmm6, %%xmm5      \n\t"\
-            \
-            "psubw  %%xmm1, %%xmm0      \n\t"\
-            "psubw  %%xmm4, %%xmm3      \n\t"\
-            "psraw      $2, %%xmm0      \n\t"\
-            "psraw      $2, %%xmm3      \n\t"\
-            "psubw  %%xmm1, %%xmm0      \n\t"\
-            "psubw  %%xmm4, %%xmm3      \n\t"\
-            "paddw  %%xmm2, %%xmm0      \n\t"\
-            "paddw  %%xmm5, %%xmm3      \n\t"\
-            "psraw      $2, %%xmm0      \n\t"\
-            "psraw      $2, %%xmm3      \n\t"\
-            "paddw  %%xmm2, %%xmm0      \n\t"\
-            "paddw  %%xmm5, %%xmm3      \n\t"\
-            "psraw      $6, %%xmm0      \n\t"\
-            "psraw      $6, %%xmm3      \n\t"\
-            "packuswb %%xmm0, %%xmm3    \n\t"\
-            OP(%%xmm3, (%1), %%xmm7, dqa)\
-            "add $48, %0                \n\t"\
-            "add %3, %1                 \n\t"\
-            "decl %2                    \n\t"\
-            " jnz 1b                    \n\t"\
-            : "+a"(tmp), "+c"(dst), "+g"(h)\
-            : "S"((x86_reg)dstStride)\
-            : XMM_CLOBBERS("%xmm0", "%xmm1", "%xmm2", "%xmm3", \
-                           "%xmm4", "%xmm5", "%xmm6", "%xmm7",)\
-              "memory"\
-        );\
-    }else{\
-        __asm__ volatile(\
-            "1:                         \n\t"\
-            "movdqa 16(%0), %%xmm1      \n\t"\
-            "movdqa   (%0), %%xmm0      \n\t"\
-            "movdqa %%xmm1, %%xmm2      \n\t"\
-            "movdqa %%xmm1, %%xmm3      \n\t"\
-            "movdqa %%xmm1, %%xmm4      \n\t"\
-            "movdqa %%xmm1, %%xmm5      \n\t"\
-            "palignr $10, %%xmm0, %%xmm5 \n\t"\
-            "palignr  $8, %%xmm0, %%xmm4 \n\t"\
-            "palignr  $6, %%xmm0, %%xmm3 \n\t"\
-            "palignr  $4, %%xmm0, %%xmm2 \n\t"\
-            "palignr  $2, %%xmm0, %%xmm1 \n\t"\
-            "paddw  %%xmm5, %%xmm0      \n\t"\
-            "paddw  %%xmm4, %%xmm1      \n\t"\
-            "paddw  %%xmm3, %%xmm2      \n\t"\
-            "psubw  %%xmm1, %%xmm0      \n\t"\
-            "psraw      $2, %%xmm0      \n\t"\
-            "psubw  %%xmm1, %%xmm0      \n\t"\
-            "paddw  %%xmm2, %%xmm0      \n\t"\
-            "psraw      $2, %%xmm0      \n\t"\
-            "paddw  %%xmm2, %%xmm0      \n\t"\
-            "psraw      $6, %%xmm0      \n\t"\
-            "packuswb %%xmm0, %%xmm0    \n\t"\
-            OP(%%xmm0, (%1), %%xmm7, q)\
-            "add $48, %0                \n\t"\
-            "add %3, %1                 \n\t"\
-            "decl %2                    \n\t"\
-            " jnz 1b                    \n\t"\
-            : "+a"(tmp), "+c"(dst), "+g"(h)\
-            : "S"((x86_reg)dstStride)\
-            : XMM_CLOBBERS("%xmm0", "%xmm1", "%xmm2", "%xmm3", \
-                           "%xmm4", "%xmm5", "%xmm6", "%xmm7",)\
-              "memory"\
-        );\
-    }\
-}
-
 #define QPEL_H264_HV_XMM(OPNAME, OP, MMX)\
-static av_noinline void OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride, int size){\
-          put_h264_qpel8or16_hv1_lowpass_sse2(tmp, src, tmpStride, srcStride, size);\
-    OPNAME ## h264_qpel8or16_hv2_lowpass_ ## MMX(dst, tmp, dstStride, tmpStride, size);\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride, int size){\
+    ff_put_h264_qpel8or16_hv1_lowpass_sse2(tmp, src, tmpStride, srcStride, size);\
+    ff_ ## OPNAME ## h264_qpel8or16_hv2_lowpass_ ## MMX(dst, tmp, dstStride, tmpStride, size);\
 }\
-static void OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
-    OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst, tmp, src, dstStride, tmpStride, srcStride, 8);\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
+    ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst, tmp, src, dstStride, tmpStride, srcStride, 8);\
 }\
-static void OPNAME ## h264_qpel16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
-    OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst, tmp, src, dstStride, tmpStride, srcStride, 16);\
+static av_always_inline void ff_ ## OPNAME ## h264_qpel16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
+    ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst, tmp, src, dstStride, tmpStride, srcStride, 16);\
 }\
 
-#define put_pixels8_l2_sse2 put_pixels8_l2_mmxext
-#define avg_pixels8_l2_sse2 avg_pixels8_l2_mmxext
-#define put_pixels16_l2_sse2 put_pixels16_l2_mmxext
-#define avg_pixels16_l2_sse2 avg_pixels16_l2_mmxext
-#define put_pixels8_l2_ssse3 put_pixels8_l2_mmxext
-#define avg_pixels8_l2_ssse3 avg_pixels8_l2_mmxext
-#define put_pixels16_l2_ssse3 put_pixels16_l2_mmxext
-#define avg_pixels16_l2_ssse3 avg_pixels16_l2_mmxext
+#define ff_put_h264_qpel8_h_lowpass_l2_sse2  ff_put_h264_qpel8_h_lowpass_l2_mmxext
+#define ff_avg_h264_qpel8_h_lowpass_l2_sse2  ff_avg_h264_qpel8_h_lowpass_l2_mmxext
+#define ff_put_h264_qpel16_h_lowpass_l2_sse2 ff_put_h264_qpel16_h_lowpass_l2_mmxext
+#define ff_avg_h264_qpel16_h_lowpass_l2_sse2 ff_avg_h264_qpel16_h_lowpass_l2_mmxext
 
-#define put_pixels8_l2_shift5_sse2 put_pixels8_l2_shift5_mmxext
-#define avg_pixels8_l2_shift5_sse2 avg_pixels8_l2_shift5_mmxext
-#define put_pixels16_l2_shift5_sse2 put_pixels16_l2_shift5_mmxext
-#define avg_pixels16_l2_shift5_sse2 avg_pixels16_l2_shift5_mmxext
-#define put_pixels8_l2_shift5_ssse3 put_pixels8_l2_shift5_mmxext
-#define avg_pixels8_l2_shift5_ssse3 avg_pixels8_l2_shift5_mmxext
-#define put_pixels16_l2_shift5_ssse3 put_pixels16_l2_shift5_mmxext
-#define avg_pixels16_l2_shift5_ssse3 avg_pixels16_l2_shift5_mmxext
+#define ff_put_h264_qpel8_v_lowpass_ssse3  ff_put_h264_qpel8_v_lowpass_sse2
+#define ff_avg_h264_qpel8_v_lowpass_ssse3  ff_avg_h264_qpel8_v_lowpass_sse2
+#define ff_put_h264_qpel16_v_lowpass_ssse3 ff_put_h264_qpel16_v_lowpass_sse2
+#define ff_avg_h264_qpel16_v_lowpass_ssse3 ff_avg_h264_qpel16_v_lowpass_sse2
 
-#define put_h264_qpel8_h_lowpass_l2_sse2 put_h264_qpel8_h_lowpass_l2_mmxext
-#define avg_h264_qpel8_h_lowpass_l2_sse2 avg_h264_qpel8_h_lowpass_l2_mmxext
-#define put_h264_qpel16_h_lowpass_l2_sse2 put_h264_qpel16_h_lowpass_l2_mmxext
-#define avg_h264_qpel16_h_lowpass_l2_sse2 avg_h264_qpel16_h_lowpass_l2_mmxext
-
-#define put_h264_qpel8_v_lowpass_ssse3 put_h264_qpel8_v_lowpass_sse2
-#define avg_h264_qpel8_v_lowpass_ssse3 avg_h264_qpel8_v_lowpass_sse2
-#define put_h264_qpel16_v_lowpass_ssse3 put_h264_qpel16_v_lowpass_sse2
-#define avg_h264_qpel16_v_lowpass_ssse3 avg_h264_qpel16_v_lowpass_sse2
-
-#define put_h264_qpel8or16_hv2_lowpass_sse2 put_h264_qpel8or16_hv2_lowpass_mmxext
-#define avg_h264_qpel8or16_hv2_lowpass_sse2 avg_h264_qpel8or16_hv2_lowpass_mmxext
+#define ff_put_h264_qpel8or16_hv2_lowpass_sse2 ff_put_h264_qpel8or16_hv2_lowpass_mmxext
+#define ff_avg_h264_qpel8or16_hv2_lowpass_sse2 ff_avg_h264_qpel8or16_hv2_lowpass_mmxext
 
 #define H264_MC(OPNAME, SIZE, MMX, ALIGN) \
 H264_MC_C(OPNAME, SIZE, MMX, ALIGN)\
@@ -1040,77 +250,77 @@ H264_MC_H(OPNAME, SIZE, MMX, ALIGN)\
 H264_MC_HV(OPNAME, SIZE, MMX, ALIGN)\
 
 static void put_h264_qpel16_mc00_sse2 (uint8_t *dst, uint8_t *src, int stride){
-    put_pixels16_sse2(dst, src, stride, 16);
+    ff_put_pixels16_sse2(dst, src, stride, 16);
 }
 static void avg_h264_qpel16_mc00_sse2 (uint8_t *dst, uint8_t *src, int stride){
-    avg_pixels16_sse2(dst, src, stride, 16);
+    ff_avg_pixels16_sse2(dst, src, stride, 16);
 }
 #define put_h264_qpel8_mc00_sse2 put_h264_qpel8_mc00_mmxext
 #define avg_h264_qpel8_mc00_sse2 avg_h264_qpel8_mc00_mmxext
 
 #define H264_MC_C(OPNAME, SIZE, MMX, ALIGN) \
 static void OPNAME ## h264_qpel ## SIZE ## _mc00_ ## MMX (uint8_t *dst, uint8_t *src, int stride){\
-    OPNAME ## pixels ## SIZE ## _ ## MMX(dst, src, stride, SIZE);\
+    ff_ ## OPNAME ## pixels ## SIZE ## _ ## MMX(dst, src, stride, SIZE);\
 }\
 
 #define H264_MC_H(OPNAME, SIZE, MMX, ALIGN) \
 static void OPNAME ## h264_qpel ## SIZE ## _mc10_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, src, stride, stride);\
+    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, src, stride, stride);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc20_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_ ## MMX(dst, src, stride, stride);\
+    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_ ## MMX(dst, src, stride, stride);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc30_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, src+1, stride, stride);\
+    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, src+1, stride, stride);\
 }\
 
 #define H264_MC_V(OPNAME, SIZE, MMX, ALIGN) \
 static void OPNAME ## h264_qpel ## SIZE ## _mc01_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
     DECLARE_ALIGNED(ALIGN, uint8_t, temp)[SIZE*SIZE];\
-    put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
-    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src, temp, stride, stride, SIZE);\
+    ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
+    ff_ ## OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src, temp, stride, stride, SIZE);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc02_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
-    OPNAME ## h264_qpel ## SIZE ## _v_lowpass_ ## MMX(dst, src, stride, stride);\
+    ff_ ## OPNAME ## h264_qpel ## SIZE ## _v_lowpass_ ## MMX(dst, src, stride, stride);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc03_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
     DECLARE_ALIGNED(ALIGN, uint8_t, temp)[SIZE*SIZE];\
-    put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
-    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src+stride, temp, stride, stride, SIZE);\
+    ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
+    ff_ ## OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src+stride, temp, stride, stride, SIZE);\
 }\
 
 #define H264_MC_HV(OPNAME, SIZE, MMX, ALIGN) \
 static void OPNAME ## h264_qpel ## SIZE ## _mc11_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
     DECLARE_ALIGNED(ALIGN, uint8_t, temp)[SIZE*SIZE];\
-    put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
-    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, temp, stride, SIZE);\
+    ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
+    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, temp, stride, SIZE);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc31_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
     DECLARE_ALIGNED(ALIGN, uint8_t, temp)[SIZE*SIZE];\
-    put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src+1, SIZE, stride);\
-    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, temp, stride, SIZE);\
+    ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src+1, SIZE, stride);\
+    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, temp, stride, SIZE);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc13_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
     DECLARE_ALIGNED(ALIGN, uint8_t, temp)[SIZE*SIZE];\
-    put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
-    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, temp, stride, SIZE);\
+    ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
+    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, temp, stride, SIZE);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc33_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
     DECLARE_ALIGNED(ALIGN, uint8_t, temp)[SIZE*SIZE];\
-    put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src+1, SIZE, stride);\
-    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, temp, stride, SIZE);\
+    ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src+1, SIZE, stride);\
+    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, temp, stride, SIZE);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc22_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
     DECLARE_ALIGNED(ALIGN, uint16_t, temp)[SIZE*(SIZE<8?12:24)];\
-    OPNAME ## h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(dst, temp, src, stride, SIZE, stride);\
+    ff_ ## OPNAME ## h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(dst, temp, src, stride, SIZE, stride);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc21_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
@@ -1118,8 +328,8 @@ static void OPNAME ## h264_qpel ## SIZE ## _mc21_ ## MMX(uint8_t *dst, uint8_t *
     uint8_t * const halfHV= temp;\
     int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
     assert(((int)temp & 7) == 0);\
-    put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
-    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, halfHV, stride, SIZE);\
+    ff_put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
+    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, halfHV, stride, SIZE);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc23_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
@@ -1127,8 +337,8 @@ static void OPNAME ## h264_qpel ## SIZE ## _mc23_ ## MMX(uint8_t *dst, uint8_t *
     uint8_t * const halfHV= temp;\
     int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
     assert(((int)temp & 7) == 0);\
-    put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
-    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, halfHV, stride, SIZE);\
+    ff_put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
+    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, halfHV, stride, SIZE);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc12_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
@@ -1136,8 +346,8 @@ static void OPNAME ## h264_qpel ## SIZE ## _mc12_ ## MMX(uint8_t *dst, uint8_t *
     uint8_t * const halfHV= temp;\
     int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
     assert(((int)temp & 7) == 0);\
-    put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
-    OPNAME ## pixels ## SIZE ## _l2_shift5_ ## MMX(dst, halfV+2, halfHV, stride, SIZE, SIZE);\
+    ff_put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
+    ff_ ## OPNAME ## pixels ## SIZE ## _l2_shift5_mmxext(dst, halfV+2, halfHV, stride, SIZE, SIZE);\
 }\
 \
 static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
@@ -1145,8 +355,8 @@ static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## MMX(uint8_t *dst, uint8_t *
     uint8_t * const halfHV= temp;\
     int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
     assert(((int)temp & 7) == 0);\
-    put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
-    OPNAME ## pixels ## SIZE ## _l2_shift5_ ## MMX(dst, halfV+3, halfHV, stride, SIZE, SIZE);\
+    ff_put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
+    ff_ ## OPNAME ## pixels ## SIZE ## _l2_shift5_mmxext(dst, halfV+3, halfHV, stride, SIZE, SIZE);\
 }\
 
 #define H264_MC_4816(MMX)\
@@ -1171,25 +381,18 @@ QPEL_H264_V_XMM(put_,       PUT_OP, sse2)
 QPEL_H264_V_XMM(avg_,AVG_MMXEXT_OP, sse2)
 QPEL_H264_HV_XMM(put_,       PUT_OP, sse2)
 QPEL_H264_HV_XMM(avg_,AVG_MMXEXT_OP, sse2)
-#if HAVE_SSSE3_INLINE
 QPEL_H264_H_XMM(put_,       PUT_OP, ssse3)
 QPEL_H264_H_XMM(avg_,AVG_MMXEXT_OP, ssse3)
-QPEL_H264_HV2_XMM(put_,       PUT_OP, ssse3)
-QPEL_H264_HV2_XMM(avg_,AVG_MMXEXT_OP, ssse3)
 QPEL_H264_HV_XMM(put_,       PUT_OP, ssse3)
 QPEL_H264_HV_XMM(avg_,AVG_MMXEXT_OP, ssse3)
-#endif
 #undef PAVGB
 
 H264_MC_4816(mmxext)
 H264_MC_816(H264_MC_V, sse2)
 H264_MC_816(H264_MC_HV, sse2)
-#if HAVE_SSSE3_INLINE
 H264_MC_816(H264_MC_H, ssse3)
 H264_MC_816(H264_MC_HV, ssse3)
-#endif
 
-#endif /* HAVE_INLINE_ASM */
 
 //10bit
 #define LUMA_MC_OP(OP, NUM, DEPTH, TYPE, OPT) \
@@ -1285,3 +488,5 @@ QPEL16_OP(mc33, MMX)
 #if ARCH_X86_32 && HAVE_YASM && CONFIG_H264QPEL // ARCH_X86_64 implies SSE2+
 QPEL16(mmxext)
 #endif
+
+#endif /* HAVE_YASM */
diff --git a/libavcodec/x86/h264_qpel_8bit.asm b/libavcodec/x86/h264_qpel_8bit.asm
new file mode 100644
index 0000000..eef0715
--- /dev/null
+++ b/libavcodec/x86/h264_qpel_8bit.asm
@@ -0,0 +1,862 @@
+;*****************************************************************************
+;* MMX/SSE2/SSSE3-optimized H.264 QPEL code
+;*****************************************************************************
+;* Copyright (c) 2004-2005 Michael Niedermayer, Loren Merritt
+;* Copyright (C) 2012 Daniel Kang
+;*
+;* Authors: Daniel Kang <daniel.d.kang at gmail.com>
+;*
+;* This file is part of Libav.
+;*
+;* Libav is free software; you can redistribute it and/or
+;* modify it under the terms of the GNU Lesser General Public
+;* License as published by the Free Software Foundation; either
+;* version 2.1 of the License, or (at your option) any later version.
+;*
+;* Libav is distributed in the hope that it will be useful,
+;* but WITHOUT ANY WARRANTY; without even the implied warranty of
+;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+;* Lesser General Public License for more details.
+;*
+;* You should have received a copy of the GNU Lesser General Public
+;* License along with Libav; if not, write to the Free Software
+;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+;******************************************************************************
+
+%include "libavutil/x86/x86util.asm"
+
+SECTION_RODATA 32
+
+cextern pw_16
+cextern pw_5
+cextern pb_0
+
+SECTION .text
+
+
+%macro op_avgh 3
+    movh   %3, %2
+    pavgb  %1, %3
+    movh   %2, %1
+%endmacro
+
+%macro op_avg 2-3
+    pavgb  %1, %2
+    mova   %2, %1
+%endmacro
+
+%macro op_puth 2-3
+    movh   %2, %1
+%endmacro
+
+%macro op_put 2-3
+    mova   %2, %1
+%endmacro
+
+%macro QPEL4_H_LOWPASS_OP 1
+cglobal %1_h264_qpel4_h_lowpass, 4,5 ; dst, src, dstStride, srcStride
+    movsxdifnidn  r2, r2d
+    movsxdifnidn  r3, r3d
+    pxor          m7, m7
+    mova          m4, [pw_5]
+    mova          m5, [pw_16]
+    mov          r4d, 4
+.loop:
+    movh          m1, [r1-1]
+    movh          m2, [r1+0]
+    movh          m3, [r1+1]
+    movh          m0, [r1+2]
+    punpcklbw     m1, m7
+    punpcklbw     m2, m7
+    punpcklbw     m3, m7
+    punpcklbw     m0, m7
+    paddw         m1, m0
+    paddw         m2, m3
+    movh          m0, [r1-2]
+    movh          m3, [r1+3]
+    punpcklbw     m0, m7
+    punpcklbw     m3, m7
+    paddw         m0, m3
+    psllw         m2, 2
+    psubw         m2, m1
+    pmullw        m2, m4
+    paddw         m0, m5
+    paddw         m0, m2
+    psraw         m0, 5
+    packuswb      m0, m0
+    op_%1h        m0, [r0], m6
+    add           r0, r2
+    add           r1, r3
+    dec          r4d
+    jg         .loop
+    REP_RET
+%endmacro
+
+INIT_MMX mmxext
+QPEL4_H_LOWPASS_OP put
+QPEL4_H_LOWPASS_OP avg
+
+%macro QPEL8_H_LOWPASS_OP 1
+cglobal %1_h264_qpel8_h_lowpass, 4,5 ; dst, src, dstStride, srcStride
+    movsxdifnidn  r2, r2d
+    movsxdifnidn  r3, r3d
+    mov          r4d, 8
+    pxor          m7, m7
+    mova          m6, [pw_5]
+.loop:
+    mova          m0, [r1]
+    mova          m2, [r1+1]
+    mova          m1, m0
+    mova          m3, m2
+    punpcklbw     m0, m7
+    punpckhbw     m1, m7
+    punpcklbw     m2, m7
+    punpckhbw     m3, m7
+    paddw         m0, m2
+    paddw         m1, m3
+    psllw         m0, 2
+    psllw         m1, 2
+    mova          m2, [r1-1]
+    mova          m4, [r1+2]
+    mova          m3, m2
+    mova          m5, m4
+    punpcklbw     m2, m7
+    punpckhbw     m3, m7
+    punpcklbw     m4, m7
+    punpckhbw     m5, m7
+    paddw         m2, m4
+    paddw         m5, m3
+    psubw         m0, m2
+    psubw         m1, m5
+    pmullw        m0, m6
+    pmullw        m1, m6
+    movd          m2, [r1-2]
+    movd          m5, [r1+7]
+    punpcklbw     m2, m7
+    punpcklbw     m5, m7
+    paddw         m2, m3
+    paddw         m4, m5
+    mova          m5, [pw_16]
+    paddw         m2, m5
+    paddw         m4, m5
+    paddw         m0, m2
+    paddw         m1, m4
+    psraw         m0, 5
+    psraw         m1, 5
+    packuswb      m0, m1
+    op_%1         m0, [r0], m4
+    add           r0, r2
+    add           r1, r3
+    dec          r4d
+    jg         .loop
+    REP_RET
+%endmacro
+
+INIT_MMX mmxext
+QPEL8_H_LOWPASS_OP put
+QPEL8_H_LOWPASS_OP avg
+
+%macro QPEL8_H_LOWPASS_OP_XMM 1
+cglobal %1_h264_qpel8_h_lowpass, 4,5,7 ; dst, src, dstStride, srcStride
+    movsxdifnidn  r2, r2d
+    movsxdifnidn  r3, r3d
+    mov          r4d, 8
+    pxor          m7, m7
+    mova          m6, [pw_5]
+.loop:
+    movu          m1, [r1-2]
+    mova          m0, m1
+    punpckhbw     m1, m7
+    punpcklbw     m0, m7
+    mova          m2, m1
+    mova          m3, m1
+    mova          m4, m1
+    mova          m5, m1
+    palignr       m4, m0, 2
+    palignr       m3, m0, 4
+    palignr       m2, m0, 6
+    palignr       m1, m0, 8
+    palignr       m5, m0, 10
+    paddw         m0, m5
+    paddw         m2, m3
+    paddw         m1, m4
+    psllw         m2, 2
+    psubw         m2, m1
+    paddw         m0, [pw_16]
+    pmullw        m2, m6
+    paddw         m2, m0
+    psraw         m2, 5
+    packuswb      m2, m2
+    op_%1h        m2, [r0], m4
+    add           r1, r3
+    add           r0, r2
+    dec          r4d
+    jne        .loop
+    REP_RET
+%endmacro
+
+INIT_XMM ssse3
+QPEL8_H_LOWPASS_OP_XMM put
+QPEL8_H_LOWPASS_OP_XMM avg
+
+
+%macro QPEL4_H_LOWPASS_L2_OP 1
+cglobal %1_h264_qpel4_h_lowpass_l2, 5,6 ; dst, src, src2, dstStride, srcStride
+    movsxdifnidn  r3, r3d
+    movsxdifnidn  r4, r4d
+    pxor          m7, m7
+    mova          m4, [pw_5]
+    mova          m5, [pw_16]
+    mov          r5d, 4
+.loop:
+    movh          m1, [r1-1]
+    movh          m2, [r1+0]
+    movh          m3, [r1+1]
+    movh          m0, [r1+2]
+    punpcklbw     m1, m7
+    punpcklbw     m2, m7
+    punpcklbw     m3, m7
+    punpcklbw     m0, m7
+    paddw         m1, m0
+    paddw         m2, m3
+    movh          m0, [r1-2]
+    movh          m3, [r1+3]
+    punpcklbw     m0, m7
+    punpcklbw     m3, m7
+    paddw         m0, m3
+    psllw         m2, 2
+    psubw         m2, m1
+    pmullw        m2, m4
+    paddw         m0, m5
+    paddw         m0, m2
+    movh          m3, [r2]
+    psraw         m0, 5
+    packuswb      m0, m0
+    pavgb         m0, m3
+    op_%1h        m0, [r0], m6
+    add           r0, r3
+    add           r1, r3
+    add           r2, r4
+    dec          r5d
+    jg         .loop
+    REP_RET
+%endmacro
+
+INIT_MMX mmxext
+QPEL4_H_LOWPASS_L2_OP put
+QPEL4_H_LOWPASS_L2_OP avg
+
+
+%macro QPEL8_H_LOWPASS_L2_OP 1
+cglobal %1_h264_qpel8_h_lowpass_l2, 5,6 ; dst, src, src2, dstStride, srcStride
+    movsxdifnidn  r3, r3d
+    movsxdifnidn  r4, r4d
+    mov          r5d, 8
+    pxor          m7, m7
+    mova          m6, [pw_5]
+.loop:
+    mova          m0, [r1]
+    mova          m2, [r1+1]
+    mova          m1, m0
+    mova          m3, m2
+    punpcklbw     m0, m7
+    punpckhbw     m1, m7
+    punpcklbw     m2, m7
+    punpckhbw     m3, m7
+    paddw         m0, m2
+    paddw         m1, m3
+    psllw         m0, 2
+    psllw         m1, 2
+    mova          m2, [r1-1]
+    mova          m4, [r1+2]
+    mova          m3, m2
+    mova          m5, m4
+    punpcklbw     m2, m7
+    punpckhbw     m3, m7
+    punpcklbw     m4, m7
+    punpckhbw     m5, m7
+    paddw         m2, m4
+    paddw         m5, m3
+    psubw         m0, m2
+    psubw         m1, m5
+    pmullw        m0, m6
+    pmullw        m1, m6
+    movd          m2, [r1-2]
+    movd          m5, [r1+7]
+    punpcklbw     m2, m7
+    punpcklbw     m5, m7
+    paddw         m2, m3
+    paddw         m4, m5
+    mova          m5, [pw_16]
+    paddw         m2, m5
+    paddw         m4, m5
+    paddw         m0, m2
+    paddw         m1, m4
+    psraw         m0, 5
+    psraw         m1, 5
+    mova          m4, [r2]
+    packuswb      m0, m1
+    pavgb         m0, m4
+    op_%1         m0, [r0], m4
+    add           r0, r3
+    add           r1, r3
+    add           r2, r4
+    dec          r5d
+    jg         .loop
+    REP_RET
+%endmacro
+
+INIT_MMX mmxext
+QPEL8_H_LOWPASS_L2_OP put
+QPEL8_H_LOWPASS_L2_OP avg
+
+
+%macro QPEL8_H_LOWPASS_L2_OP_XMM 1
+cglobal %1_h264_qpel8_h_lowpass_l2, 5,6,7 ; dst, src, src2, dstStride, src2Stride
+    movsxdifnidn  r3, r3d
+    movsxdifnidn  r4, r4d
+    mov          r5d, 8
+    pxor          m7, m7
+    mova          m6, [pw_5]
+.loop:
+    lddqu         m1, [r1-2]
+    mova          m0, m1
+    punpckhbw     m1, m7
+    punpcklbw     m0, m7
+    mova          m2, m1
+    mova          m3, m1
+    mova          m4, m1
+    mova          m5, m1
+    palignr       m4, m0, 2
+    palignr       m3, m0, 4
+    palignr       m2, m0, 6
+    palignr       m1, m0, 8
+    palignr       m5, m0, 10
+    paddw         m0, m5
+    paddw         m2, m3
+    paddw         m1, m4
+    psllw         m2, 2
+    movh          m3, [r2]
+    psubw         m2, m1
+    paddw         m0, [pw_16]
+    pmullw        m2, m6
+    paddw         m2, m0
+    psraw         m2, 5
+    packuswb      m2, m2
+    pavgb         m2, m3
+    op_%1h        m2, [r0], m4
+    add           r1, r3
+    add           r0, r3
+    add           r2, r4
+    dec          r5d
+    jg         .loop
+    REP_RET
+%endmacro
+
+INIT_XMM ssse3
+QPEL8_H_LOWPASS_L2_OP_XMM put
+QPEL8_H_LOWPASS_L2_OP_XMM avg
+
+
+; All functions that call this are required to have function arguments of
+; dst, src, dstStride, srcStride
+%macro FILT_V 1
+    mova      m6, m2
+    movh      m5, [r1]
+    paddw     m6, m3
+    psllw     m6, 2
+    psubw     m6, m1
+    psubw     m6, m4
+    punpcklbw m5, m7
+    pmullw    m6, [pw_5]
+    paddw     m0, [pw_16]
+    add       r1, r3
+    paddw     m0, m5
+    paddw     m6, m0
+    psraw     m6, 5
+    packuswb  m6, m6
+    op_%1h    m6, [r0], m0 ; 1
+    add       r0, r2
+    SWAP       0, 1, 2, 3, 4, 5
+%endmacro
+
+%macro QPEL4_V_LOWPASS_OP 1
+cglobal %1_h264_qpel4_v_lowpass, 4,4 ; dst, src, dstStride, srcStride
+    movsxdifnidn  r2, r2d
+    movsxdifnidn  r3, r3d
+    sub           r1, r3
+    sub           r1, r3
+    pxor          m7, m7
+    movh          m0, [r1]
+    movh          m1, [r1+r3]
+    lea           r1, [r1+2*r3]
+    movh          m2, [r1]
+    movh          m3, [r1+r3]
+    lea           r1, [r1+2*r3]
+    movh          m4, [r1]
+    add           r1, r3
+    punpcklbw     m0, m7
+    punpcklbw     m1, m7
+    punpcklbw     m2, m7
+    punpcklbw     m3, m7
+    punpcklbw     m4, m7
+    FILT_V        %1
+    FILT_V        %1
+    FILT_V        %1
+    FILT_V        %1
+    RET
+%endmacro
+
+INIT_MMX mmxext
+QPEL4_V_LOWPASS_OP put
+QPEL4_V_LOWPASS_OP avg
+
+
+
+%macro QPEL8OR16_V_LOWPASS_OP 1
+%if cpuflag(sse2)
+cglobal %1_h264_qpel8or16_v_lowpass, 5,5,7 ; dst, src, dstStride, srcStride, h
+    movsxdifnidn  r2, r2d
+    movsxdifnidn  r3, r3d
+    sub           r1, r3
+    sub           r1, r3
+%else
+cglobal %1_h264_qpel8or16_v_lowpass_op, 5,5,7 ; dst, src, dstStride, srcStride, h
+    movsxdifnidn  r2, r2d
+    movsxdifnidn  r3, r3d
+%endif
+    pxor          m7, m7
+    movh          m0, [r1]
+    movh          m1, [r1+r3]
+    lea           r1, [r1+2*r3]
+    movh          m2, [r1]
+    movh          m3, [r1+r3]
+    lea           r1, [r1+2*r3]
+    movh          m4, [r1]
+    add           r1, r3
+    punpcklbw     m0, m7
+    punpcklbw     m1, m7
+    punpcklbw     m2, m7
+    punpcklbw     m3, m7
+    punpcklbw     m4, m7
+    FILT_V        %1
+    FILT_V        %1
+    FILT_V        %1
+    FILT_V        %1
+    FILT_V        %1
+    FILT_V        %1
+    FILT_V        %1
+    FILT_V        %1
+    cmp          r4d, 16
+    jne         .end
+    FILT_V        %1
+    FILT_V        %1
+    FILT_V        %1
+    FILT_V        %1
+    FILT_V        %1
+    FILT_V        %1
+    FILT_V        %1
+    FILT_V        %1
+.end:
+    REP_RET
+%endmacro
+
+INIT_MMX mmxext
+QPEL8OR16_V_LOWPASS_OP put
+QPEL8OR16_V_LOWPASS_OP avg
+
+INIT_XMM sse2
+QPEL8OR16_V_LOWPASS_OP put
+QPEL8OR16_V_LOWPASS_OP avg
+
+
+; All functions that use this are required to have args:
+; src, tmp, srcSize
+%macro FILT_HV 1 ; offset
+    mova           m6, m2
+    movh           m5, [r0]
+    paddw          m6, m3
+    psllw          m6, 2
+    paddw          m0, [pw_16]
+    psubw          m6, m1
+    psubw          m6, m4
+    punpcklbw      m5, m7
+    pmullw         m6, [pw_5]
+    paddw          m0, m5
+    add            r0, r2
+    paddw          m6, m0
+    mova      [r1+%1], m6
+    SWAP            0, 1, 2, 3, 4, 5
+%endmacro
+
+%macro QPEL4_HV1_LOWPASS_OP 1
+cglobal %1_h264_qpel4_hv_lowpass_v, 3,3 ; src, tmp, srcStride
+    movsxdifnidn  r2, r2d
+    pxor          m7, m7
+    movh          m0, [r0]
+    movh          m1, [r0+r2]
+    lea           r0, [r0+2*r2]
+    movh          m2, [r0]
+    movh          m3, [r0+r2]
+    lea           r0, [r0+2*r2]
+    movh          m4, [r0]
+    add           r0, r2
+    punpcklbw     m0, m7
+    punpcklbw     m1, m7
+    punpcklbw     m2, m7
+    punpcklbw     m3, m7
+    punpcklbw     m4, m7
+    FILT_HV       0*24
+    FILT_HV       1*24
+    FILT_HV       2*24
+    FILT_HV       3*24
+    RET
+
+cglobal %1_h264_qpel4_hv_lowpass_h, 3,4 ; tmp, dst, dstStride
+    movsxdifnidn  r2, r2d
+    mov          r3d, 4
+.loop:
+    mova          m0, [r0]
+    paddw         m0, [r0+10]
+    mova          m1, [r0+2]
+    paddw         m1, [r0+8]
+    mova          m2, [r0+4]
+    paddw         m2, [r0+6]
+    psubw         m0, m1
+    psraw         m0, 2
+    psubw         m0, m1
+    paddsw        m0, m2
+    psraw         m0, 2
+    paddw         m0, m2
+    psraw         m0, 6
+    packuswb      m0, m0
+    op_%1h        m0, [r1], m7
+    add           r0, 24
+    add           r1, r2
+    dec          r3d
+    jnz        .loop
+    REP_RET
+%endmacro
+
+INIT_MMX mmxext
+QPEL4_HV1_LOWPASS_OP put
+QPEL4_HV1_LOWPASS_OP avg
+
+%macro QPEL8OR16_HV1_LOWPASS_OP 1
+cglobal %1_h264_qpel8or16_hv1_lowpass_op, 4,4,7 ; src, tmp, srcStride, size
+    movsxdifnidn  r2, r2d
+    pxor          m7, m7
+    movh          m0, [r0]
+    movh          m1, [r0+r2]
+    lea           r0, [r0+2*r2]
+    movh          m2, [r0]
+    movh          m3, [r0+r2]
+    lea           r0, [r0+2*r2]
+    movh          m4, [r0]
+    add           r0, r2
+    punpcklbw     m0, m7
+    punpcklbw     m1, m7
+    punpcklbw     m2, m7
+    punpcklbw     m3, m7
+    punpcklbw     m4, m7
+    FILT_HV     0*48
+    FILT_HV     1*48
+    FILT_HV     2*48
+    FILT_HV     3*48
+    FILT_HV     4*48
+    FILT_HV     5*48
+    FILT_HV     6*48
+    FILT_HV     7*48
+    cmp          r3d, 16
+    jne         .end
+    FILT_HV     8*48
+    FILT_HV     9*48
+    FILT_HV    10*48
+    FILT_HV    11*48
+    FILT_HV    12*48
+    FILT_HV    13*48
+    FILT_HV    14*48
+    FILT_HV    15*48
+.end:
+    REP_RET
+%endmacro
+
+INIT_MMX mmxext
+QPEL8OR16_HV1_LOWPASS_OP put
+QPEL8OR16_HV1_LOWPASS_OP avg
+
+INIT_XMM sse2
+QPEL8OR16_HV1_LOWPASS_OP put
+
+
+
+%macro QPEL8OR16_HV2_LOWPASS_OP 1
+; unused is to match ssse3 and mmxext args
+cglobal %1_h264_qpel8or16_hv2_lowpass_op, 5,5 ; dst, tmp, dstStride, unused, h
+    movsxdifnidn  r2, r2d
+.loop:
+    mova          m0, [r1]
+    mova          m3, [r1+8]
+    mova          m1, [r1+2]
+    mova          m4, [r1+10]
+    paddw         m0, m4
+    paddw         m1, m3
+    paddw         m3, [r1+18]
+    paddw         m4, [r1+16]
+    mova          m2, [r1+4]
+    mova          m5, [r1+12]
+    paddw         m2, [r1+6]
+    paddw         m5, [r1+14]
+    psubw         m0, m1
+    psubw         m3, m4
+    psraw         m0, 2
+    psraw         m3, 2
+    psubw         m0, m1
+    psubw         m3, m4
+    paddsw        m0, m2
+    paddsw        m3, m5
+    psraw         m0, 2
+    psraw         m3, 2
+    paddw         m0, m2
+    paddw         m3, m5
+    psraw         m0, 6
+    psraw         m3, 6
+    packuswb      m0, m3
+    op_%1         m0, [r0], m7
+    add           r1, 48
+    add           r0, r2
+    dec          r4d
+    jne        .loop
+    REP_RET
+%endmacro
+
+INIT_MMX mmxext
+QPEL8OR16_HV2_LOWPASS_OP put
+QPEL8OR16_HV2_LOWPASS_OP avg
+
+%macro QPEL8OR16_HV2_LOWPASS_OP_XMM 1
+cglobal %1_h264_qpel8or16_hv2_lowpass, 5,5,7 ; dst, tmp, dstStride, tmpStride, size
+    movsxdifnidn  r2, r2d
+    movsxdifnidn  r3, r3d
+    cmp          r4d, 16
+    je         .op16
+.loop8:
+    mova          m1, [r1+16]
+    mova          m0, [r1]
+    mova          m2, m1
+    mova          m3, m1
+    mova          m4, m1
+    mova          m5, m1
+    palignr       m5, m0, 10
+    palignr       m4, m0, 8
+    palignr       m3, m0, 6
+    palignr       m2, m0, 4
+    palignr       m1, m0, 2
+    paddw         m0, m5
+    paddw         m1, m4
+    paddw         m2, m3
+    psubw         m0, m1
+    psraw         m0, 2
+    psubw         m0, m1
+    paddw         m0, m2
+    psraw         m0, 2
+    paddw         m0, m2
+    psraw         m0, 6
+    packuswb      m0, m0
+    op_%1h        m0, [r0], m7
+    add           r1, 48
+    add           r0, r2
+    dec          r4d
+    jne       .loop8
+    jmp        .done
+.op16:
+    mova          m4, [r1+32]
+    mova          m5, [r1+16]
+    mova          m7, [r1]
+    mova          m3, m4
+    mova          m2, m4
+    mova          m1, m4
+    mova          m0, m4
+    palignr       m0, m5, 10
+    palignr       m1, m5, 8
+    palignr       m2, m5, 6
+    palignr       m3, m5, 4
+    palignr       m4, m5, 2
+    paddw         m0, m5
+    paddw         m1, m4
+    paddw         m2, m3
+    mova          m6, m5
+    mova          m4, m5
+    mova          m3, m5
+    palignr       m4, m7, 8
+    palignr       m6, m7, 2
+    palignr       m3, m7, 10
+    paddw         m4, m6
+    mova          m6, m5
+    palignr       m5, m7, 6
+    palignr       m6, m7, 4
+    paddw         m3, m7
+    paddw         m5, m6
+    psubw         m0, m1
+    psubw         m3, m4
+    psraw         m0, 2
+    psraw         m3, 2
+    psubw         m0, m1
+    psubw         m3, m4
+    paddw         m0, m2
+    paddw         m3, m5
+    psraw         m0, 2
+    psraw         m3, 2
+    paddw         m0, m2
+    paddw         m3, m5
+    psraw         m0, 6
+    psraw         m3, 6
+    packuswb      m3, m0
+    op_%1         m3, [r0], m7
+    add           r1, 48
+    add           r0, r2
+    dec          r4d
+    jne        .op16
+.done:
+    REP_RET
+%endmacro
+
+INIT_XMM ssse3
+QPEL8OR16_HV2_LOWPASS_OP_XMM put
+QPEL8OR16_HV2_LOWPASS_OP_XMM avg
+
+
+%macro PIXELS4_L2_SHIFT5 1
+cglobal %1_pixels4_l2_shift5,6,6 ; dst, src16, src8, dstStride, src8Stride, h
+    movsxdifnidn  r3, r3d
+    movsxdifnidn  r4, r4d
+    mova          m0, [r1]
+    mova          m1, [r1+24]
+    psraw         m0, 5
+    psraw         m1, 5
+    packuswb      m0, m0
+    packuswb      m1, m1
+    pavgb         m0, [r2]
+    pavgb         m1, [r2+r4]
+    op_%1h        m0, [r0], m4
+    op_%1h        m1, [r0+r3], m5
+    lea           r2, [r2+r4*2]
+    lea           r0, [r0+r3*2]
+    mova          m0, [r1+48]
+    mova          m1, [r1+72]
+    psraw         m0, 5
+    psraw         m1, 5
+    packuswb      m0, m0
+    packuswb      m1, m1
+    pavgb         m0, [r2]
+    pavgb         m1, [r2+r4]
+    op_%1h        m0, [r0], m4
+    op_%1h        m1, [r0+r3], m5
+    RET
+%endmacro
+
+INIT_MMX mmxext
+PIXELS4_L2_SHIFT5 put
+PIXELS4_L2_SHIFT5 avg
+
+
+%macro PIXELS8_L2_SHIFT5 1
+cglobal %1_pixels8_l2_shift5, 6, 6 ; dst, src16, src8, dstStride, src8Stride, h
+    movsxdifnidn  r3, r3d
+    movsxdifnidn  r4, r4d
+.loop:
+    mova          m0, [r1]
+    mova          m1, [r1+8]
+    mova          m2, [r1+48]
+    mova          m3, [r1+48+8]
+    psraw         m0, 5
+    psraw         m1, 5
+    psraw         m2, 5
+    psraw         m3, 5
+    packuswb      m0, m1
+    packuswb      m2, m3
+    pavgb         m0, [r2]
+    pavgb         m2, [r2+r4]
+    op_%1         m0, [r0], m4
+    op_%1         m2, [r0+r3], m5
+    lea           r2, [r2+2*r4]
+    add           r1, 48*2
+    lea           r0, [r0+2*r3]
+    sub          r5d, 2
+    jne        .loop
+    REP_RET
+%endmacro
+
+INIT_MMX mmxext
+PIXELS8_L2_SHIFT5 put
+PIXELS8_L2_SHIFT5 avg
+
+
+%if ARCH_X86_64
+%macro QPEL16_H_LOWPASS_L2_OP 1
+cglobal %1_h264_qpel16_h_lowpass_l2, 5, 6, 16 ; dst, src, src2, dstStride, src2Stride
+    movsxdifnidn  r3, r3d
+    movsxdifnidn  r4, r4d
+    mov          r5d, 16
+    pxor         m15, m15
+    mova         m14, [pw_5]
+    mova         m13, [pw_16]
+.loop:
+    lddqu         m1, [r1+6]
+    lddqu         m7, [r1-2]
+    mova          m0, m1
+    punpckhbw     m1, m15
+    punpcklbw     m0, m15
+    punpcklbw     m7, m15
+    mova          m2, m1
+    mova          m6, m0
+    mova          m3, m1
+    mova          m8, m0
+    mova          m4, m1
+    mova          m9, m0
+    mova         m12, m0
+    mova         m11, m1
+    palignr      m11, m0, 10
+    palignr      m12, m7, 10
+    palignr       m4, m0, 2
+    palignr       m9, m7, 2
+    palignr       m3, m0, 4
+    palignr       m8, m7, 4
+    palignr       m2, m0, 6
+    palignr       m6, m7, 6
+    paddw        m11, m0
+    palignr       m1, m0, 8
+    palignr       m0, m7, 8
+    paddw         m7, m12
+    paddw         m2, m3
+    paddw         m6, m8
+    paddw         m1, m4
+    paddw         m0, m9
+    psllw         m2, 2
+    psllw         m6, 2
+    psubw         m2, m1
+    psubw         m6, m0
+    paddw        m11, m13
+    paddw         m7, m13
+    pmullw        m2, m14
+    pmullw        m6, m14
+    lddqu         m3, [r2]
+    paddw         m2, m11
+    paddw         m6, m7
+    psraw         m2, 5
+    psraw         m6, 5
+    packuswb      m6, m2
+    pavgb         m6, m3
+    op_%1         m6, [r0], m11
+    add           r1, r3
+    add           r0, r3
+    add           r2, r4
+    dec          r5d
+    jg         .loop
+    REP_RET
+%endmacro
+
+INIT_XMM ssse3
+QPEL16_H_LOWPASS_L2_OP put
+QPEL16_H_LOWPASS_L2_OP avg
+%endif



More information about the ffmpeg-cvslog mailing list