ARGBToRGBA implemented as alias

BUG=78
TEST=unittest
Review URL: https://webrtc-codereview.appspot.com/788009

git-svn-id: http://libyuv.googlecode.com/svn/trunk@353 16f28f9a-4ce2-e073-06de-1de4eb20be90

README.chromium

 Name: libyuv
 URL: http://code.google.com/p/libyuv/
-Version: 352
+Version: 353
 License: BSD
 License File: LICENSE
 

include/libyuv/planar_functions.h

@@ -68,6 +68,11 @@ int NV21ToRGB565(const uint8* src_y, int src_stride_y,
 #define ARGBToBGRA BGRAToARGB
 #define ARGBToABGR ABGRToARGB
 
+// Convert ARGB To RGBA.
+int ARGBToRGBA(const uint8* src_frame, int src_stride_frame,
+               uint8* dst_argb, int dst_stride_argb,
+               int width, int height);
+
 // Convert ARGB To RGB24.
 int ARGBToRGB24(const uint8* src_argb, int src_stride_argb,
                 uint8* dst_rgb24, int dst_stride_rgb24,

include/libyuv/version.h

@@ -11,6 +11,6 @@
 #ifndef INCLUDE_LIBYUV_VERSION_H_  // NOLINT
 #define INCLUDE_LIBYUV_VERSION_H_
 
-#define LIBYUV_VERSION 352
+#define LIBYUV_VERSION 353
 
 #endif  // INCLUDE_LIBYUV_VERSION_H_  NOLINT

source/planar_functions.cc

@@ -296,6 +296,39 @@ int ARGBToI422(const uint8* src_argb, int src_stride_argb,
   return 0;
 }
 
+// Convert ARGB to RGBA.
+int ARGBToRGBA(const uint8* src_argb, int src_stride_argb,
+               uint8* dst_rgba, int dst_stride_rgba,
+               int width, int height) {
+  if (!src_argb || !dst_rgba ||
+      width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  void (*ARGBToRGBARow)(const uint8* src_argb, uint8* dst_rgba, int pix) =
+      ARGBToRGBARow_C;
+#if defined(HAS_ARGBTORGBAROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3) &&
+      IS_ALIGNED(width, 4) &&
+      IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16) &&
+      IS_ALIGNED(dst_rgba, 16) && IS_ALIGNED(dst_stride_rgba, 16)) {
+    ARGBToRGBARow = ARGBToRGBARow_SSSE3;
+  }
+#endif
+
+  for (int y = 0; y < height; ++y) {
+    ARGBToRGBARow(src_argb, dst_rgba, width);
+    src_argb += src_stride_argb;
+    dst_rgba += dst_stride_rgba;
+  }
+  return 0;
+}
+
 // Convert ARGB To RGB24.
 int ARGBToRGB24(const uint8* src_argb, int src_stride_argb,
                 uint8* dst_rgb24, int dst_stride_rgb24,

source/row.h

@@ -38,6 +38,7 @@ extern "C" {
 #define HAS_ABGRTOUVROW_SSSE3
 #define HAS_ABGRTOYROW_SSSE3
 #define HAS_RGBATOARGBROW_SSSE3
+#define HAS_ARGBTORGBAROW_SSSE3
 #define HAS_ARGB1555TOARGBROW_SSE2
 #define HAS_ARGB4444TOARGBROW_SSE2
 #define HAS_ARGBATTENUATEROW_SSSE3
@@ -232,12 +233,14 @@ void RGB565ToARGBRow_C(const uint8* src_rgb, uint8* dst_argb, int pix);
 void ARGB1555ToARGBRow_C(const uint8* src_argb, uint8* dst_argb, int pix);
 void ARGB4444ToARGBRow_C(const uint8* src_argb, uint8* dst_argb, int pix);
 
+void ARGBToRGBARow_SSSE3(const uint8* src_argb, uint8* dst_rgb, int pix);
 void ARGBToRGB24Row_SSSE3(const uint8* src_argb, uint8* dst_rgb, int pix);
 void ARGBToRAWRow_SSSE3(const uint8* src_argb, uint8* dst_rgb, int pix);
 void ARGBToRGB565Row_SSE2(const uint8* src_argb, uint8* dst_rgb, int pix);
 void ARGBToARGB1555Row_SSE2(const uint8* src_argb, uint8* dst_rgb, int pix);
 void ARGBToARGB4444Row_SSE2(const uint8* src_argb, uint8* dst_rgb, int pix);
 
+void ARGBToRGBARow_C(const uint8* src_argb, uint8* dst_rgb, int pix);
 void ARGBToRGB24Row_C(const uint8* src_argb, uint8* dst_rgb, int pix);
 void ARGBToRAWRow_C(const uint8* src_argb, uint8* dst_rgb, int pix);
 void ARGBToRGB565Row_C(const uint8* src_argb, uint8* dst_rgb, int pix);

source/row_common.cc

@@ -139,6 +139,21 @@ void ARGB4444ToARGBRow_C(const uint8* src_rgb, uint8* dst_argb, int width) {
   }
 }
 
+void ARGBToRGBARow_C(const uint8* src_argb, uint8* dst_rgb, int width) {
+  for (int x = 0; x < width; ++x) {
+    uint8 b = src_argb[0];
+    uint8 g = src_argb[1];
+    uint8 r = src_argb[2];
+    uint8 a = src_argb[3];
+    dst_rgb[0] = a;
+    dst_rgb[1] = b;
+    dst_rgb[2] = g;
+    dst_rgb[3] = r;
+    dst_rgb += 4;
+    src_argb += 4;
+  }
+}
+
 void ARGBToRGB24Row_C(const uint8* src_argb, uint8* dst_rgb, int width) {
   for (int x = 0; x < width; ++x) {
     uint8 b = src_argb[0];

source/row_posix.cc

@@ -103,6 +103,11 @@ CONST uvec8 kShuffleMaskRGBAToARGB = {
   1u, 2u, 3u, 0u, 5u, 6u, 7u, 4u, 9u, 10u, 11u, 8u, 13u, 14u, 15u, 12u
 };
 
+// Shuffle table for converting ARGB to RGBA.
+CONST uvec8 kShuffleMaskARGBToRGBA = {
+  3u, 0u, 1u, 2u, 7u, 4u, 5u, 6u, 11u, 8u, 9u, 10u, 15u, 12u, 13u, 14u
+};
+
 // Shuffle table for converting ARGB to RGB24.
 CONST uvec8 kShuffleMaskARGBToRGB24 = {
   0u, 1u, 2u, 4u, 5u, 6u, 8u, 9u, 10u, 12u, 13u, 14u, 128u, 128u, 128u, 128u
@@ -214,6 +219,30 @@ void RGBAToARGBRow_SSSE3(const uint8* src_rgba, uint8* dst_argb, int pix) {
   );
 }
 
+void ARGBToRGBARow_SSSE3(const uint8* src_argb, uint8* dst_rgba, int pix) {
+  asm volatile (
+    "movdqa    %3,%%xmm5                       \n"
+    "sub       %0,%1                           \n"
+    ".p2align  4                               \n"
+  "1:                                          \n"
+    "movdqa    (%0),%%xmm0                     \n"
+    "pshufb    %%xmm5,%%xmm0                   \n"
+    "sub       $0x4,%2                         \n"
+    "movdqa    %%xmm0,(%0,%1,1)                \n"
+    "lea       0x10(%0),%0                     \n"
+    "jg        1b                              \n"
+
+  : "+r"(src_argb),  // %0
+    "+r"(dst_rgba),  // %1
+    "+r"(pix)        // %2
+  : "m"(kShuffleMaskARGBToRGBA)  // %3
+  : "memory", "cc"
+#if defined(__SSE2__)
+    , "xmm0", "xmm5"
+#endif
+  );
+}
+
 void RGB24ToARGBRow_SSSE3(const uint8* src_rgb24, uint8* dst_argb, int pix) {
   asm volatile (
     "pcmpeqb   %%xmm5,%%xmm5                   \n"  // generate mask 0xff000000

source/row_win.cc

@@ -93,6 +93,11 @@ static const uvec8 kShuffleMaskRGBAToARGB = {
   1u, 2u, 3u, 0u, 5u, 6u, 7u, 4u, 9u, 10u, 11u, 8u, 13u, 14u, 15u, 12u
 };
 
+// Shuffle table for converting ARGB to RGBA.
+static const uvec8 kShuffleMaskARGBToRGBA = {
+  3u, 0u, 1u, 2u, 7u, 4u, 5u, 6u, 11u, 8u, 9u, 10u, 15u, 12u, 13u, 14u
+};
+
 // Shuffle table for converting ARGB to RGB24.
 static const uvec8 kShuffleMaskARGBToRGB24 = {
   0u, 1u, 2u, 4u, 5u, 6u, 8u, 9u, 10u, 12u, 13u, 14u, 128u, 128u, 128u, 128u
@@ -194,6 +199,27 @@ __asm {
   }
 }
 
+__declspec(naked) __declspec(align(16))
+void ARGBToRGBARow_SSSE3(const uint8* src_argb, uint8* dst_rgba, int pix) {
+__asm {
+    mov       eax, [esp + 4]   // src_argb
+    mov       edx, [esp + 8]   // dst_rgba
+    mov       ecx, [esp + 12]  // pix
+    movdqa    xmm5, kShuffleMaskARGBToRGBA
+    sub       edx, eax
+
+    align      16
+ convertloop:
+    movdqa    xmm0, [eax]
+    pshufb    xmm0, xmm5
+    sub       ecx, 4
+    movdqa    [eax + edx], xmm0
+    lea       eax, [eax + 16]
+    jg        convertloop
+    ret
+  }
+}
+
 __declspec(naked) __declspec(align(16))
 void RGB24ToARGBRow_SSSE3(const uint8* src_rgb24, uint8* dst_argb, int pix) {
 __asm {

unit_test/planar_test.cc

@@ -299,6 +299,7 @@ TEST_F(libyuvTest, FMT_A##To##FMT_B##N##_OptVsC) {                             \
 TESTATOB(ARGB, 4, 4, ARGB, 4)
 TESTATOB(ARGB, 4, 4, BGRA, 4)
 TESTATOB(ARGB, 4, 4, ABGR, 4)
+TESTATOB(ARGB, 4, 4, RGBA, 4)
 TESTATOB(ARGB, 4, 4, RAW, 3)
 TESTATOB(ARGB, 4, 4, RGB24, 3)
 TESTATOB(ARGB, 4, 4, RGB565, 2)
@@ -355,6 +356,7 @@ TEST_F(libyuvTest, FMT_A##To##FMT_B##_Random) {                                \
 TESTATOBRANDOM(ARGB, 4, 4, ARGB, 4)
 TESTATOBRANDOM(ARGB, 4, 4, BGRA, 4)
 TESTATOBRANDOM(ARGB, 4, 4, ABGR, 4)
+TESTATOBRANDOM(ARGB, 4, 4, RGBA, 4)
 TESTATOBRANDOM(ARGB, 4, 4, RAW, 3)
 TESTATOBRANDOM(ARGB, 4, 4, RGB24, 3)
 TESTATOBRANDOM(ARGB, 4, 4, RGB565, 2)