Merge commit '6d5636ad'

* commit '6d5636ad':
  hevc: x86: Add add_residual() SIMD optimizations

See a6af4bf6

This merge is only cosmetics (renames, space shuffling, etc).

The functionnal changes in the ASM are *not* merged:
- unrolling with %rep is kept
- ADD_RES_MMX_4_8 is left untouched: this needs investigation
Merged-by: Clément Bœsch <u@pkh.me>

doc/libav-merge.txt

@@ -97,6 +97,7 @@ Stuff that didn't reach the codebase:
 - VAAPI VP8 decode hwaccel (currently under review: http://ffmpeg.org/pipermail/ffmpeg-devel/2017-February/thread.html#207348)
 - Removal of the custom atomic API (5cc0057f49, see http://ffmpeg.org/pipermail/ffmpeg-devel/2017-March/209003.html)
 - Use the new bitstream filter for extracting extradata (see 8e2ea69135 and 096a8effa3)
+- ADD_RES_MMX_4_8 in libavcodec/x86/hevc_add_res.asm probably needs updating (see 589880710)
 
 Collateral damage that needs work locally:
 ------------------------------------------

libavcodec/hevcdsp.h

@@ -46,7 +46,7 @@ typedef struct HEVCDSPContext {
     void (*put_pcm)(uint8_t *_dst, ptrdiff_t _stride, int width, int height,
                     struct GetBitContext *gb, int pcm_bit_depth);
 
-    void (*add_residual[4])(uint8_t *_dst, int16_t *coeffs, ptrdiff_t _stride);
+    void (*add_residual[4])(uint8_t *dst, int16_t *res, ptrdiff_t stride);
 
     void (*dequant)(int16_t *coeffs, int16_t log2_size);
 

libavcodec/x86/hevcdsp.h

@@ -236,23 +236,24 @@ WEIGHTING_PROTOTYPES(12, sse4);
 ///////////////////////////////////////////////////////////////////////////////
 // TRANSFORM_ADD
 ///////////////////////////////////////////////////////////////////////////////
-void ff_hevc_add_residual4_8_mmxext(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
-void ff_hevc_add_residual8_8_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
-void ff_hevc_add_residual16_8_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
-void ff_hevc_add_residual32_8_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
 
-void ff_hevc_add_residual8_8_avx(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
-void ff_hevc_add_residual16_8_avx(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
-void ff_hevc_add_residual32_8_avx(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
+void ff_hevc_add_residual_4_8_mmxext(uint8_t *dst, int16_t *res, ptrdiff_t stride);
+void ff_hevc_add_residual_8_8_sse2(uint8_t *dst, int16_t *res, ptrdiff_t stride);
+void ff_hevc_add_residual_16_8_sse2(uint8_t *dst, int16_t *res, ptrdiff_t stride);
+void ff_hevc_add_residual_32_8_sse2(uint8_t *dst, int16_t *res, ptrdiff_t stride);
 
-void ff_hevc_add_residual32_8_avx2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
+void ff_hevc_add_residual_8_8_avx(uint8_t *dst, int16_t *res, ptrdiff_t stride);
+void ff_hevc_add_residual_16_8_avx(uint8_t *dst, int16_t *res, ptrdiff_t stride);
+void ff_hevc_add_residual_32_8_avx(uint8_t *dst, int16_t *res, ptrdiff_t stride);
 
-void ff_hevc_add_residual4_10_mmxext(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
-void ff_hevc_add_residual8_10_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
-void ff_hevc_add_residual16_10_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
-void ff_hevc_add_residual32_10_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
+void ff_hevc_add_residual_32_8_avx2(uint8_t *dst, int16_t *res, ptrdiff_t stride);
 
-void ff_hevc_add_residual16_10_avx2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
-void ff_hevc_add_residual32_10_avx2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
+void ff_hevc_add_residual_4_10_mmxext(uint8_t *dst, int16_t *res, ptrdiff_t stride);
+void ff_hevc_add_residual_8_10_sse2(uint8_t *dst, int16_t *res, ptrdiff_t stride);
+void ff_hevc_add_residual_16_10_sse2(uint8_t *dst, int16_t *res, ptrdiff_t stride);
+void ff_hevc_add_residual_32_10_sse2(uint8_t *dst, int16_t *res, ptrdiff_t stride);
+
+void ff_hevc_add_residual_16_10_avx2(uint8_t *dst, int16_t *res, ptrdiff_t stride);
+void ff_hevc_add_residual_32_10_avx2(uint8_t *dst, int16_t *res, ptrdiff_t stride);
 
 #endif // AVCODEC_X86_HEVCDSP_H

libavcodec/x86/hevcdsp_init.c

@@ -713,7 +713,8 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
         if (EXTERNAL_MMXEXT(cpu_flags)) {
             c->idct_dc[0] = ff_hevc_idct_4x4_dc_8_mmxext;
             c->idct_dc[1] = ff_hevc_idct_8x8_dc_8_mmxext;
-            c->add_residual[0] = ff_hevc_add_residual4_8_mmxext;
+
+            c->add_residual[0] = ff_hevc_add_residual_4_8_mmxext;
         }
         if (EXTERNAL_SSE2(cpu_flags)) {
             c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_8_sse2;
@@ -734,9 +735,9 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
             c->idct[0]    = ff_hevc_idct_4x4_8_sse2;
             c->idct[1]    = ff_hevc_idct_8x8_8_sse2;
 
-            c->add_residual[1] = ff_hevc_add_residual8_8_sse2;
-            c->add_residual[2] = ff_hevc_add_residual16_8_sse2;
-            c->add_residual[3] = ff_hevc_add_residual32_8_sse2;
+            c->add_residual[1] = ff_hevc_add_residual_8_8_sse2;
+            c->add_residual[2] = ff_hevc_add_residual_16_8_sse2;
+            c->add_residual[3] = ff_hevc_add_residual_32_8_sse2;
         }
         if (EXTERNAL_SSSE3(cpu_flags)) {
             if(ARCH_X86_64) {
@@ -772,9 +773,9 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
             c->idct[0] = ff_hevc_idct_4x4_8_avx;
             c->idct[1] = ff_hevc_idct_8x8_8_avx;
 
-            c->add_residual[1] = ff_hevc_add_residual8_8_avx;
-            c->add_residual[2] = ff_hevc_add_residual16_8_avx;
-            c->add_residual[3] = ff_hevc_add_residual32_8_avx;
+            c->add_residual[1] = ff_hevc_add_residual_8_8_avx;
+            c->add_residual[2] = ff_hevc_add_residual_16_8_avx;
+            c->add_residual[3] = ff_hevc_add_residual_32_8_avx;
         }
         if (EXTERNAL_AVX2(cpu_flags)) {
             c->sao_band_filter[0] = ff_hevc_sao_band_filter_8_8_avx2;
@@ -874,11 +875,11 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
             c->sao_edge_filter[3] = ff_hevc_sao_edge_filter_48_8_avx2;
             c->sao_edge_filter[4] = ff_hevc_sao_edge_filter_64_8_avx2;
 
-            c->add_residual[3] = ff_hevc_add_residual32_8_avx2;
+            c->add_residual[3] = ff_hevc_add_residual_32_8_avx2;
         }
     } else if (bit_depth == 10) {
         if (EXTERNAL_MMXEXT(cpu_flags)) {
-            c->add_residual[0] = ff_hevc_add_residual4_10_mmxext;
+            c->add_residual[0] = ff_hevc_add_residual_4_10_mmxext;
             c->idct_dc[0] = ff_hevc_idct_4x4_dc_10_mmxext;
             c->idct_dc[1] = ff_hevc_idct_8x8_dc_10_mmxext;
         }
@@ -902,9 +903,9 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
             c->idct[0]    = ff_hevc_idct_4x4_10_sse2;
             c->idct[1]    = ff_hevc_idct_8x8_10_sse2;
 
-            c->add_residual[1] = ff_hevc_add_residual8_10_sse2;
-            c->add_residual[2] = ff_hevc_add_residual16_10_sse2;
-            c->add_residual[3] = ff_hevc_add_residual32_10_sse2;
+            c->add_residual[1] = ff_hevc_add_residual_8_10_sse2;
+            c->add_residual[2] = ff_hevc_add_residual_16_10_sse2;
+            c->add_residual[3] = ff_hevc_add_residual_32_10_sse2;
         }
         if (EXTERNAL_SSSE3(cpu_flags) && ARCH_X86_64) {
             c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_10_ssse3;
@@ -1090,9 +1091,8 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
             SAO_BAND_INIT(10, avx2);
             SAO_EDGE_INIT(10, avx2);
 
-            c->add_residual[2] = ff_hevc_add_residual16_10_avx2;
-            c->add_residual[3] = ff_hevc_add_residual32_10_avx2;
-
+            c->add_residual[2] = ff_hevc_add_residual_16_10_avx2;
+            c->add_residual[3] = ff_hevc_add_residual_32_10_avx2;
         }
     } else if (bit_depth == 12) {
         if (EXTERNAL_MMXEXT(cpu_flags)) {