Add x86 asm functions for VP8 put_pixels

Originally committed as revision 23858 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/x86/vp8dsp-init.c

@@ -87,6 +87,16 @@ extern void ff_put_vp8_bilinear8_h_ssse3 (uint8_t *dst, int dststride,
                                           uint8_t *src, int srcstride,
                                           int height, int mx, int my);
 
+extern void ff_put_vp8_pixels8_mmx (uint8_t *dst, int dststride,
+                                    uint8_t *src, int srcstride,
+                                    int height, int mx, int my);
+extern void ff_put_vp8_pixels16_mmx(uint8_t *dst, int dststride,
+                                    uint8_t *src, int srcstride,
+                                    int height, int mx, int my);
+extern void ff_put_vp8_pixels16_sse(uint8_t *dst, int dststride,
+                                    uint8_t *src, int srcstride,
+                                    int height, int mx, int my);
+
 #define TAP_W16(OPT, FILTERTYPE, TAPTYPE) \
 static void ff_put_vp8_ ## FILTERTYPE ## 16_ ## TAPTYPE ## _ ## OPT( \
     uint8_t *dst,  int dststride, uint8_t *src, \
@@ -218,6 +228,10 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
 #if HAVE_YASM
     if (mm_flags & FF_MM_MMX) {
         c->vp8_idct_dc_add                  = ff_vp8_idct_dc_add_mmx;
+        c->put_vp8_epel_pixels_tab[0][0][0]     =
+        c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_mmx;
+        c->put_vp8_epel_pixels_tab[1][0][0]     =
+        c->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_mmx;
     }
 
     /* note that 4-tap width=16 functions are missing because w=16
@@ -231,6 +245,11 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
         VP8_BILINEAR_MC_FUNC(1, 4, mmxext);
     }
 
+    if (mm_flags & FF_MM_SSE) {
+        c->put_vp8_epel_pixels_tab[0][0][0]     =
+        c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_sse;
+    }
+
     if (mm_flags & FF_MM_SSE2) {
         VP8_LUMA_MC_FUNC(0, 16, sse2);
         VP8_MC_FUNC(1, 8, sse2);

libavcodec/x86/vp8dsp.asm

@@ -810,6 +810,46 @@ cglobal put_vp8_bilinear8_h_ssse3, 7,7,5
     jg .nextrow
     REP_RET
 
+cglobal put_vp8_pixels8_mmx, 5,5
+.nextrow:
+    movq  mm0, [r2+r3*0]
+    movq  mm1, [r2+r3*1]
+    lea    r2, [r2+r3*2]
+    movq [r0+r1*0], mm0
+    movq [r0+r1*1], mm1
+    lea    r0, [r0+r1*2]
+    sub   r4d, 2
+    jg .nextrow
+    REP_RET
+
+cglobal put_vp8_pixels16_mmx, 5,5
+.nextrow:
+    movq  mm0, [r2+r3*0+0]
+    movq  mm1, [r2+r3*0+8]
+    movq  mm2, [r2+r3*1+0]
+    movq  mm3, [r2+r3*1+8]
+    lea    r2, [r2+r3*2]
+    movq [r0+r1*0+0], mm0
+    movq [r0+r1*0+8], mm1
+    movq [r0+r1*1+0], mm2
+    movq [r0+r1*1+8], mm3
+    lea    r0, [r0+r1*2]
+    sub   r4d, 2
+    jg .nextrow
+    REP_RET
+
+cglobal put_vp8_pixels16_sse, 5,5,2
+.nextrow:
+    movups xmm0, [r2+r3*0]
+    movups xmm1, [r2+r3*1]
+    lea     r2, [r2+r3*2]
+    movaps [r0+r1*0], xmm0
+    movaps [r0+r1*1], xmm1
+    lea     r0, [r0+r1*2]
+    sub    r4d, 2
+    jg .nextrow
+    REP_RET
+
 ;-----------------------------------------------------------------------------
 ; IDCT functions:
 ;