YUY2 16 pixels at a time is 2x faster

BUG=116
TEST=libyuv_unittest
Review URL: https://webrtc-codereview.appspot.com/870005

git-svn-id: http://libyuv.googlecode.com/svn/trunk@393 16f28f9a-4ce2-e073-06de-1de4eb20be90

README.chromium

 Name: libyuv
 URL: http://code.google.com/p/libyuv/
-Version: 392
+Version: 393
 License: BSD
 License File: LICENSE
 

include/libyuv/version.h

@@ -11,6 +11,6 @@
 #ifndef INCLUDE_LIBYUV_VERSION_H_  // NOLINT
 #define INCLUDE_LIBYUV_VERSION_H_
 
-#define LIBYUV_VERSION 392
+#define LIBYUV_VERSION 393
 
 #endif  // INCLUDE_LIBYUV_VERSION_H_  NOLINT

source/convert.cc

@@ -558,11 +558,9 @@ int Q420ToI420(const uint8* src_y, int src_stride_y,
         YUY2ToUV422Row = YUY2ToUV422Row_Any_NEON;
       }
     }
-    if (IS_ALIGNED(width, 8)) {
+    if (IS_ALIGNED(width, 16)) {
       YUY2ToYRow = YUY2ToYRow_NEON;
-      if (IS_ALIGNED(width, 16)) {
-        YUY2ToUV422Row = YUY2ToUV422Row_NEON;
-      }
+      YUY2ToUV422Row = YUY2ToUV422Row_NEON;
     }
   }
 #endif
@@ -683,11 +681,9 @@ int YUY2ToI420(const uint8* src_yuy2, int src_stride_yuy2,
         YUY2ToUVRow = YUY2ToUVRow_Any_NEON;
       }
     }
-    if (IS_ALIGNED(width, 8)) {
+    if (IS_ALIGNED(width, 16)) {
       YUY2ToYRow = YUY2ToYRow_NEON;
-      if (IS_ALIGNED(width, 16)) {
-        YUY2ToUVRow = YUY2ToUVRow_NEON;
-      }
+      YUY2ToUVRow = YUY2ToUVRow_NEON;
     }
   }
 #endif
@@ -752,11 +748,9 @@ int UYVYToI420(const uint8* src_uyvy, int src_stride_uyvy,
         UYVYToUVRow = UYVYToUVRow_Any_NEON;
       }
     }
-    if (IS_ALIGNED(width, 8)) {
+    if (IS_ALIGNED(width, 16)) {
       UYVYToYRow = UYVYToYRow_NEON;
-      if (IS_ALIGNED(width, 16)) {
-        UYVYToUVRow = UYVYToUVRow_NEON;
-      }
+      UYVYToUVRow = UYVYToUVRow_NEON;
     }
   }
 #endif
@@ -872,11 +866,9 @@ int V210ToI420(const uint8* src_v210, int src_stride_v210,
         UYVYToUVRow = UYVYToUVRow_Any_NEON;
       }
     }
-    if (IS_ALIGNED(width, 8)) {
+    if (IS_ALIGNED(width, 16)) {
       UYVYToYRow = UYVYToYRow_NEON;
-      if (IS_ALIGNED(width, 16)) {
-        UYVYToUVRow = UYVYToUVRow_NEON;
-      }
+      UYVYToUVRow = UYVYToUVRow_NEON;
     }
   }
 #endif
@@ -903,11 +895,9 @@ int V210ToI420(const uint8* src_v210, int src_stride_v210,
         UYVYToUVRow = UYVYToUVRow_Any_NEON;
       }
     }
-    if (IS_ALIGNED(width, 8)) {
+    if (IS_ALIGNED(width, 16)) {
       UYVYToYRow = UYVYToYRow_NEON;
-      if (IS_ALIGNED(width, 16)) {
-        UYVYToUVRow = UYVYToUVRow_NEON;
-      }
+      UYVYToUVRow = UYVYToUVRow_NEON;
     }
   }
 #endif

source/row_neon.cc

@@ -509,15 +509,15 @@ void ARGBToRAWRow_NEON(const uint8* src_argb, uint8* dst_raw, int pix) {
 void YUY2ToYRow_NEON(const uint8* src_yuy2, uint8* dst_y, int pix) {
   asm volatile (
   "1:                                          \n"
-    "vld2.u8    {d0, d1}, [%0]!                \n"  // load 8 pixels of YUY2.
-    "subs       %2, %2, #8                     \n"  // 8 processed per loop.
-    "vst1.u8    {d0}, [%1]!                    \n"  // store 8 pixels of Y.
+    "vld2.u8    {q0, q1}, [%0]!                \n"  // load 16 pixels of YUY2.
+    "subs       %2, %2, #16                    \n"  // 16 processed per loop.
+    "vst1.u8    {q0}, [%1]!                    \n"  // store 16 pixels of Y.
     "bgt        1b                             \n"
   : "+r"(src_yuy2),  // %0
     "+r"(dst_y),     // %1
     "+r"(pix)        // %2
   :
-  : "memory", "cc", "d0", "d1"  // Clobber List
+  : "memory", "cc", "q0", "q1"  // Clobber List
   );
 }
 #endif  // HAS_YUY2TOYROW_NEON
@@ -526,19 +526,22 @@ void YUY2ToYRow_NEON(const uint8* src_yuy2, uint8* dst_y, int pix) {
 void UYVYToYRow_NEON(const uint8* src_uyvy, uint8* dst_y, int pix) {
   asm volatile (
   "1:                                          \n"
-    "vld2.u8    {d0, d1}, [%0]!                \n"  // load 8 pixels of UYVY.
-    "subs       %2, %2, #8                     \n"  // 8 processed per loop.
-    "vst1.u8    {d1}, [%1]!                    \n"  // store 8 pixels of Y.
+    "vld2.u8    {q0, q1}, [%0]!                \n"  // load 16 pixels of UYVY.
+    "subs       %2, %2, #16                    \n"  // 16 processed per loop.
+    "vst1.u8    {q1}, [%1]!                    \n"  // store 16 pixels of Y.
     "bgt        1b                             \n"
   : "+r"(src_uyvy),  // %0
     "+r"(dst_y),     // %1
     "+r"(pix)        // %2
   :
-  : "memory", "cc", "d0", "d1"  // Clobber List
+  : "memory", "cc", "q0", "q1"  // Clobber List
   );
 }
 #endif  // HAS_UYVYTOYROW_NEON
 
+#endif  // HAS_UYVYTOYROW_NEON
+
+
 #ifdef HAS_YUY2TOYROW_NEON
 void YUY2ToUV422Row_NEON(const uint8* src_yuy2, uint8* dst_u, uint8* dst_v,
                          int pix) {
@@ -627,8 +630,6 @@ void UYVYToUVRow_NEON(const uint8* src_uyvy, int stride_uyvy,
   : "memory", "cc", "d0", "d1", "d2", "d3", "d4", "d5", "d6", "d7"  // Clobber List
   );
 }
-#endif  // HAS_UYVYTOYROW_NEON
-
 #endif  // __ARM_NEON__
 
 #ifdef __cplusplus