test for unaligned vs aligned for CopyRow_SSE2

source/row_win.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.
  */
 
@@ skipping 3358 matching lines @@
 #endif  //  HAS_MERGEUVROW_AVX2
 
 #ifdef HAS_COPYROW_SSE2
 // CopyRow copys 'count' bytes using a 16 byte load/store, 32 bytes at time.
 __declspec(naked)
 void CopyRow_SSE2(const uint8* src, uint8* dst, int count) {
   __asm {
     mov        eax, [esp + 4]   // src
     mov        edx, [esp + 8]   // dst
     mov        ecx, [esp + 12]  // count
+    test       eax, 15
+    jne        convertloopu
+    test       edx, 15
+    jne        convertloopu
 
-  convertloop:
+  convertloopa:
+    movdqa     xmm0, [eax]
+    movdqa     xmm1, [eax + 16]
+    lea        eax, [eax + 32]
+    movdqa     [edx], xmm0
+    movdqa     [edx + 16], xmm1
+    lea        edx, [edx + 32]
+    sub        ecx, 32
+    jg         convertloopa
+    ret
+
+  convertloopu:
     movdqu     xmm0, [eax]
     movdqu     xmm1, [eax + 16]
     lea        eax, [eax + 32]
     movdqu     [edx], xmm0
     movdqu     [edx + 16], xmm1
     lea        edx, [edx + 32]
     sub        ecx, 32
-    jg         convertloop
+    jg         convertloopu
     ret
   }
 }
 #endif  // HAS_COPYROW_SSE2
 
 #ifdef HAS_COPYROW_AVX
 // CopyRow copys 'count' bytes using a 32 byte load/store, 64 bytes at time.
 __declspec(naked)
 void CopyRow_AVX(const uint8* src, uint8* dst, int count) {
   __asm {
@@ skipping 2829 matching lines @@
 }
 #endif  // HAS_ARGBLUMACOLORTABLEROW_SSSE3
 
 #endif  // defined(_M_X64)
 #endif  // !defined(LIBYUV_DISABLE_X86) && (defined(_M_IX86) || defined(_M_X64))
 
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv
 #endif