Add SplitUVPlanes and MergeUVPlanes

source/row_common.cc

 /*
  *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS. All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
 #include "libyuv/row.h"
 
 #include <string.h>  // For memcpy and memset.
 
 #include "libyuv/basic_types.h"
+#include "libyuv/cpu_id.h"
 
 #ifdef __cplusplus
 namespace libyuv {
 extern "C" {
 #endif
 
 // llvm x86 is poor at ternary operator, so use branchless min/max.
 
 #define USE_BRANCHLESS 1
 #if USE_BRANCHLESS
@@ skipping 1767 matching lines @@
     dst_v[x] = src_uv[1];
     dst_v[x + 1] = src_uv[3];
     src_uv += 4;
   }
   if (width & 1) {
     dst_u[width - 1] = src_uv[0];
     dst_v[width - 1] = src_uv[1];
   }
 }
 
+SplitUVRowFunction GetOptimizedSplitUVRowFunction(

[fbarchard1, 2016/08/24 17:42:24]

suggest removing.  common is not meant for this sort of function.
For each row function, there should be a _NEON, _SSE2, and _C version, which
goes here.
If we do getters, I think this belongs in its own file, or row_any.c

+    const uint8 *src_uv, int src_stride_uv,
+    const uint8 *dst_u, int dst_stride_u,
+    const uint8 *dst_v, int dst_stride_v,
+    int width) {
+#if defined(HAS_SPLITUVROW_DSPR2)
+  if (TestCpuFlag(kCpuHasDSPR2) &&
+      IS_ALIGNED(src_uv, 4) && IS_ALIGNED(src_stride_uv, 4) &&
+      IS_ALIGNED(dst_u, 4) && IS_ALIGNED(dst_stride_u, 4) &&
+      IS_ALIGNED(dst_v, 4) && IS_ALIGNED(dst_stride_v, 4)) {
+    return IS_ALIGNED(width, 16) ? SplitUVRow_DSPR2 : SplitUVRow_Any_DSPR2;
+  }
+#endif
+#if defined(HAS_SPLITUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    return IS_ALIGNED(width, 16) ? SplitUVRow_NEON : SplitUVRow_Any_NEON;
+  }
+#endif
+#if defined(HAS_SPLITUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    return IS_ALIGNED(width, 32) ? SplitUVRow_AVX2 : SplitUVRow_Any_AVX2;
+  }
+#endif
+#if defined(HAS_SPLITUVROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    return IS_ALIGNED(width, 16) ? SplitUVRow_SSE2 : SplitUVRow_Any_SSE2;
+  }
+#endif
+  return SplitUVRow_C;
+}
+
 void MergeUVRow_C(const uint8* src_u, const uint8* src_v, uint8* dst_uv,
                   int width) {
   int x;
   for (x = 0; x < width - 1; x += 2) {
     dst_uv[0] = src_u[x];
     dst_uv[1] = src_v[x];
     dst_uv[2] = src_u[x + 1];
     dst_uv[3] = src_v[x + 1];
     dst_uv += 4;
   }
   if (width & 1) {
     dst_uv[0] = src_u[width - 1];
     dst_uv[1] = src_v[width - 1];
   }
 }
 
+MergeUVRowFunction GetOptimizedMergeUVRowFunction(int width) {
+#if defined(HAS_MERGEUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    return IS_ALIGNED(width, 16) ? MergeUVRow_NEON : MergeUVRow_Any_NEON;
+  }
+#endif
+#if defined(HAS_MERGEUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    return IS_ALIGNED(width, 32) ? MergeUVRow_AVX2 : MergeUVRow_Any_AVX2;
+  }
+#endif
+#if defined(HAS_MERGEUVROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    return IS_ALIGNED(width, 16) ? MergeUVRow_SSE2 : MergeUVRow_Any_SSE2;
+  }
+#endif
+  return MergeUVRow_C;
+}
+
 void CopyRow_C(const uint8* src, uint8* dst, int count) {
   memcpy(dst, src, count);
 }
 
 void CopyRow_16_C(const uint16* src, uint16* dst, int count) {
   memcpy(dst, src, count * 2);
 }
 
 void SetRow_C(uint8* dst, uint8 v8, int width) {
   memset(dst, v8, width);
@@ skipping 789 matching lines @@
     dst_rgb565 += twidth * 2;
     width -= twidth;
   }
 }
 #endif
 
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv
 #endif