Update x86 H.264 deblock asm

Includes AVX versions from x264.

libavcodec/x86/h264dsp_mmx.c

@@ -219,11 +219,11 @@ static void h264_loop_filter_strength_mmx2( int16_t bS[2][4][4], uint8_t nnz[40]
 }
 
 #define LF_FUNC(DIR, TYPE, OPT) \
-void ff_x264_deblock_ ## DIR ## _ ## TYPE ## _ ## OPT (uint8_t *pix, int stride, \
-                                               int alpha, int beta, int8_t *tc0);
+void ff_deblock_ ## DIR ## _ ## TYPE ## _ ## OPT (uint8_t *pix, int stride, \
+                                                  int alpha, int beta, int8_t *tc0);
 #define LF_IFUNC(DIR, TYPE, OPT) \
-void ff_x264_deblock_ ## DIR ## _ ## TYPE ## _ ## OPT (uint8_t *pix, int stride, \
-                                               int alpha, int beta);
+void ff_deblock_ ## DIR ## _ ## TYPE ## _ ## OPT (uint8_t *pix, int stride, \
+                                                  int alpha, int beta);
 
 LF_FUNC (h,  chroma,       mmxext)
 LF_IFUNC(h,  chroma_intra, mmxext)
@@ -234,18 +234,18 @@ LF_FUNC (h,  luma,         mmxext)
 LF_IFUNC(h,  luma_intra,   mmxext)
 #if HAVE_YASM && ARCH_X86_32
 LF_FUNC (v8, luma,         mmxext)
-static void ff_x264_deblock_v_luma_mmxext(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
+static void ff_deblock_v_luma_mmxext(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
 {
     if((tc0[0] & tc0[1]) >= 0)
-        ff_x264_deblock_v8_luma_mmxext(pix+0, stride, alpha, beta, tc0);
+        ff_deblock_v8_luma_mmxext(pix+0, stride, alpha, beta, tc0);
     if((tc0[2] & tc0[3]) >= 0)
-        ff_x264_deblock_v8_luma_mmxext(pix+8, stride, alpha, beta, tc0+2);
+        ff_deblock_v8_luma_mmxext(pix+8, stride, alpha, beta, tc0+2);
 }
 LF_IFUNC(v8, luma_intra,   mmxext)
-static void ff_x264_deblock_v_luma_intra_mmxext(uint8_t *pix, int stride, int alpha, int beta)
+static void ff_deblock_v_luma_intra_mmxext(uint8_t *pix, int stride, int alpha, int beta)
 {
-    ff_x264_deblock_v8_luma_intra_mmxext(pix+0, stride, alpha, beta);
-    ff_x264_deblock_v8_luma_intra_mmxext(pix+8, stride, alpha, beta);
+    ff_deblock_v8_luma_intra_mmxext(pix+0, stride, alpha, beta);
+    ff_deblock_v8_luma_intra_mmxext(pix+8, stride, alpha, beta);
 }
 #endif
 
@@ -253,6 +253,10 @@ LF_FUNC (h,  luma,         sse2)
 LF_IFUNC(h,  luma_intra,   sse2)
 LF_FUNC (v,  luma,         sse2)
 LF_IFUNC(v,  luma_intra,   sse2)
+LF_FUNC (h,  luma,         avx)
+LF_IFUNC(h,  luma_intra,   avx)
+LF_FUNC (v,  luma,         avx)
+LF_IFUNC(v,  luma_intra,   avx)
 
 /***********************************/
 /* weighted prediction */
@@ -314,15 +318,15 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth)
             c->h264_idct_add8      = ff_h264_idct_add8_mmx2;
             c->h264_idct_add16intra= ff_h264_idct_add16intra_mmx2;
 
-            c->h264_v_loop_filter_chroma= ff_x264_deblock_v_chroma_mmxext;
-            c->h264_h_loop_filter_chroma= ff_x264_deblock_h_chroma_mmxext;
-            c->h264_v_loop_filter_chroma_intra= ff_x264_deblock_v_chroma_intra_mmxext;
-            c->h264_h_loop_filter_chroma_intra= ff_x264_deblock_h_chroma_intra_mmxext;
+            c->h264_v_loop_filter_chroma= ff_deblock_v_chroma_mmxext;
+            c->h264_h_loop_filter_chroma= ff_deblock_h_chroma_mmxext;
+            c->h264_v_loop_filter_chroma_intra= ff_deblock_v_chroma_intra_mmxext;
+            c->h264_h_loop_filter_chroma_intra= ff_deblock_h_chroma_intra_mmxext;
 #if ARCH_X86_32
-            c->h264_v_loop_filter_luma= ff_x264_deblock_v_luma_mmxext;
-            c->h264_h_loop_filter_luma= ff_x264_deblock_h_luma_mmxext;
-            c->h264_v_loop_filter_luma_intra = ff_x264_deblock_v_luma_intra_mmxext;
-            c->h264_h_loop_filter_luma_intra = ff_x264_deblock_h_luma_intra_mmxext;
+            c->h264_v_loop_filter_luma= ff_deblock_v_luma_mmxext;
+            c->h264_h_loop_filter_luma= ff_deblock_h_luma_mmxext;
+            c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_mmxext;
+            c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_mmxext;
 #endif
             c->weight_h264_pixels_tab[0]= ff_h264_weight_16x16_mmx2;
             c->weight_h264_pixels_tab[1]= ff_h264_weight_16x8_mmx2;
@@ -360,10 +364,10 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth)
                 c->biweight_h264_pixels_tab[4]= ff_h264_biweight_8x4_sse2;
 
 #if HAVE_ALIGNED_STACK
-                c->h264_v_loop_filter_luma = ff_x264_deblock_v_luma_sse2;
-                c->h264_h_loop_filter_luma = ff_x264_deblock_h_luma_sse2;
-                c->h264_v_loop_filter_luma_intra = ff_x264_deblock_v_luma_intra_sse2;
-                c->h264_h_loop_filter_luma_intra = ff_x264_deblock_h_luma_intra_sse2;
+                c->h264_v_loop_filter_luma = ff_deblock_v_luma_sse2;
+                c->h264_h_loop_filter_luma = ff_deblock_h_luma_sse2;
+                c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_sse2;
+                c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_sse2;
 #endif
 
                 c->h264_idct_add16 = ff_h264_idct_add16_sse2;
@@ -377,6 +381,14 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth)
                 c->biweight_h264_pixels_tab[3]= ff_h264_biweight_8x8_ssse3;
                 c->biweight_h264_pixels_tab[4]= ff_h264_biweight_8x4_ssse3;
             }
+            if (mm_flags&AV_CPU_FLAG_AVX) {
+#if HAVE_ALIGNED_STACK
+                c->h264_v_loop_filter_luma = ff_deblock_v_luma_avx;
+                c->h264_h_loop_filter_luma = ff_deblock_h_luma_avx;
+                c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_avx;
+                c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_avx;
+#endif
+            }
         }
     }
 #endif

libavcodec/x86/x86util.asm

@@ -24,16 +24,20 @@
 ;******************************************************************************
 
 %macro SBUTTERFLY 4
+%if avx_enabled == 0
     mova      m%4, m%2
     punpckl%1 m%2, m%3
     punpckh%1 m%4, m%3
+%else
+    punpckh%1 m%4, m%2, m%3
+    punpckl%1 m%2, m%3
+%endif
     SWAP %3, %4
 %endmacro
 
 %macro SBUTTERFLY2 4
-    mova      m%4, m%2
-    punpckh%1 m%2, m%3
-    punpckl%1 m%4, m%3
+    punpckl%1 m%4, m%2, m%3
+    punpckh%1 m%2, m%2, m%3
     SWAP %2, %4, %3
 %endmacro
 
@@ -444,3 +448,12 @@
 %macro PMINUB_MMXEXT 3 ; dst, src, ignored
     pminub   %1, %2
 %endmacro
+
+%macro SPLATW 2-3 0
+%if mmsize == 16
+    pshuflw    %1, %2, (%3)*0x55
+    punpcklqdq %1, %1
+%else
+    pshufw     %1, %2, (%3)*0x55
+%endif
+%endmacro