libvpx: Pull from upstream

source/libvpx/third_party/libyuv/source/scale_gcc.cc

 /*
  *  Copyright 2013 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 555 matching lines @@
     "jg        1b                              \n"
   : "+r"(src_ptr),    // %0
     "+r"(dst_ptr),    // %1
     "+r"(dst_width)   // %2
   : "r"((intptr_t)(src_stride))   // %3
   : "memory", "cc", NACL_R14
     "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7"
   );
 }
 
+// Reads 16xN bytes and produces 16 shorts at a time.
 void ScaleAddRows_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
                        uint16* dst_ptr, int src_width, int src_height) {
   int tmp_height = 0;
   intptr_t tmp_src = 0;
   asm volatile (
+    "mov       %0,%3                           \n"  // row pointer
+    "mov       %5,%2                           \n"  // height
+    "pxor      %%xmm0,%%xmm0                   \n"  // clear accumulators
+    "pxor      %%xmm1,%%xmm1                   \n"
     "pxor      %%xmm4,%%xmm4                   \n"
-    "sub       $0x1,%5                         \n"
 
     LABELALIGN
   "1:                                          \n"
-    "movdqu    " MEMACCESS(0) ",%%xmm0         \n"
-    "mov       %0,%3                           \n"
-    "add       %6,%0                           \n"
-    "movdqa    %%xmm0,%%xmm1                   \n"
-    "punpcklbw %%xmm4,%%xmm0                   \n"
-    "punpckhbw %%xmm4,%%xmm1                   \n"
-    "mov       %5,%2                           \n"
-    "test      %2,%2                           \n"
-    "je        3f                              \n"
-
-    LABELALIGN
-  "2:                                          \n"
-    "movdqu    " MEMACCESS(0) ",%%xmm2         \n"
-    "add       %6,%0                           \n"
+    "movdqu    " MEMACCESS(3) ",%%xmm2         \n"
+    "add       %6,%3                           \n"
     "movdqa    %%xmm2,%%xmm3                   \n"
     "punpcklbw %%xmm4,%%xmm2                   \n"
     "punpckhbw %%xmm4,%%xmm3                   \n"
     "paddusw   %%xmm2,%%xmm0                   \n"
     "paddusw   %%xmm3,%%xmm1                   \n"
     "sub       $0x1,%2                         \n"
-    "jg        2b                              \n"
+    "jg        1b                              \n"
 
-    LABELALIGN
-  "3:                                          \n"
     "movdqu    %%xmm0," MEMACCESS(1) "         \n"
     "movdqu    %%xmm1," MEMACCESS2(0x10,1) "   \n"
-    "lea       " MEMLEA(0x10,3) ",%0           \n"
     "lea       " MEMLEA(0x20,1) ",%1           \n"
+    "lea       " MEMLEA(0x10,0) ",%0           \n"  // src_ptr += 16
+    "mov       %0,%3                           \n"  // row pointer
+    "mov       %5,%2                           \n"  // height
+    "pxor      %%xmm0,%%xmm0                   \n"  // clear accumulators
+    "pxor      %%xmm1,%%xmm1                   \n"
     "sub       $0x10,%4                        \n"
     "jg        1b                              \n"
   : "+r"(src_ptr),     // %0
     "+r"(dst_ptr),     // %1
     "+r"(tmp_height),  // %2
     "+r"(tmp_src),     // %3
     "+r"(src_width),   // %4
     "+rm"(src_height)  // %5
   : "rm"((intptr_t)(src_stride))  // %6
   : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4"
@@ skipping 168 matching lines @@
     "+r"(dst_width)   // %2
   : "r"((intptr_t)(src_stride))   // %3
   : "memory", "cc", NACL_R14
     "xmm0", "xmm1", "xmm2", "xmm3"
   );
 }
 
 // Reads 4 pixels at a time.
 // Alignment requirement: dst_argb 16 byte aligned.
 void ScaleARGBRowDownEven_SSE2(const uint8* src_argb, ptrdiff_t src_stride,
-                               int src_stepx,
-                               uint8* dst_argb, int dst_width) {
+                               int src_stepx, uint8* dst_argb, int dst_width) {
   intptr_t src_stepx_x4 = (intptr_t)(src_stepx);
   intptr_t src_stepx_x12 = 0;
   asm volatile (
     "lea       " MEMLEA3(0x00,1,4) ",%1        \n"
     "lea       " MEMLEA4(0x00,1,1,2) ",%4      \n"
     LABELALIGN
   "1:                                          \n"
     "movd      " MEMACCESS(0) ",%%xmm0         \n"
     MEMOPREG(movd,0x00,0,1,1,xmm1)             //  movd      (%0,%1,1),%%xmm1
     "punpckldq %%xmm1,%%xmm0                   \n"
@@ skipping 273 matching lines @@
   );
   return num;
 }
 
 #endif  // defined(__x86_64__) || defined(__i386__)
 
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv
 #endif