Add ARGBExtractAlpha_AVX2 function

source/row_gcc.cc

 // VERSION 2
 /*
  *  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.
  */
@@ skipping 2842 matching lines @@
   : "+r"(src_argb),  // %0
     "+r"(dst_a),     // %1
     "+rm"(width)     // %2
   :
   : "memory", "cc"
     , "xmm0", "xmm1"
   );
 }
 #endif  // HAS_ARGBEXTRACTALPHAROW_SSE2
 
+#ifdef HAS_ARGBEXTRACTALPHAROW_AVX2
+static const uvec8 kShuffleAlphaShort_AVX2 = {
+  3u, 128u, 128u, 128u, 7u, 128u, 128u, 128u,
+  11u, 128u, 128u, 128u, 15u, 128u, 128u, 128u
+};
+
+void ARGBExtractAlphaRow_AVX2(const uint8* src_argb, uint8* dst_a, int width) {
+ asm volatile (
+    "vmovdqa    %3,%%ymm4                      \n"
+    "vbroadcastf128 %4,%%ymm5                  \n"
+    LABELALIGN
+  "1:                                          \n"
+    "vmovdqu   " MEMACCESS(0) ", %%ymm0        \n"
+    "vmovdqu   " MEMACCESS2(0x20, 0) ", %%ymm1 \n"
+    "vpshufb    %%ymm5,%%ymm0,%%ymm0           \n" // vpsrld $0x18, %%ymm0
+    "vpshufb    %%ymm5,%%ymm1,%%ymm1           \n"
+    "vmovdqu   " MEMACCESS2(0x40, 0) ", %%ymm2 \n"
+    "vmovdqu   " MEMACCESS2(0x60, 0) ", %%ymm3 \n"
+    "lea       " MEMLEA(0x80, 0) ", %0         \n"
+    "vpackssdw  %%ymm1, %%ymm0, %%ymm0         \n"  // mutates
+    "vpshufb    %%ymm5,%%ymm2,%%ymm2           \n"
+    "vpshufb    %%ymm5,%%ymm3,%%ymm3           \n"
+    "vpackssdw  %%ymm3, %%ymm2, %%ymm2         \n"  // mutates
+    "vpackuswb  %%ymm2,%%ymm0,%%ymm0           \n"  // mutates.
+    "vpermd     %%ymm0,%%ymm4,%%ymm0           \n"  // unmutate.
+    "vmovdqu    %%ymm0," MEMACCESS(1) "        \n"
+    "lea       " MEMLEA(0x20,1) ",%1           \n"
+    "sub        $0x20, %2                      \n"
+    "jg         1b                             \n"
+    "vzeroupper                                \n"
+  : "+r"(src_argb),  // %0
+    "+r"(dst_a),     // %1
+    "+rm"(width)     // %2
+  : "m"(kPermdARGBToY_AVX),  // %3
+    "m"(kShuffleAlphaShort_AVX2)  // %4
+  : "memory", "cc"
+    , "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
+  );
+}
+#endif  // HAS_ARGBEXTRACTALPHAROW_AVX2
+
 #ifdef HAS_ARGBCOPYYTOALPHAROW_SSE2
 // width in pixels
 void ARGBCopyYToAlphaRow_SSE2(const uint8* src, uint8* dst, int width) {
   asm volatile (
     "pcmpeqb   %%xmm0,%%xmm0                   \n"
     "pslld     $0x18,%%xmm0                    \n"
     "pcmpeqb   %%xmm1,%%xmm1                   \n"
     "psrld     $0x8,%%xmm1                     \n"
     LABELALIGN
   "1:                                          \n"
@@ skipping 2618 matching lines @@
   );
 }
 #endif  // HAS_ARGBLUMACOLORTABLEROW_SSSE3
 
 #endif  // defined(__x86_64__) || defined(__i386__)
 
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv
 #endif