RGBAToARGB conversion. SSSE3 optimized.

BUG=78
TEST=RGBA unittests
Review URL: https://webrtc-codereview.appspot.com/788008

git-svn-id: http://libyuv.googlecode.com/svn/trunk@351 16f28f9a-4ce2-e073-06de-1de4eb20be90

README.chromium

@@ -1,6 +1,6 @@
 Name: libyuv
 URL: http://code.google.com/p/libyuv/
-Version: 350
+Version: 351
 License: BSD
 License File: LICENSE
 

include/libyuv/convert_argb.h

@@ -122,6 +122,11 @@ int ABGRToARGB(const uint8* src_frame, int src_stride_frame,
                uint8* dst_argb, int dst_stride_argb,
                int width, int height);
 
+// RGBA little endian (abgr in memory) to ARGB
+int RGBAToARGB(const uint8* src_frame, int src_stride_frame,
+               uint8* dst_argb, int dst_stride_argb,
+               int width, int height);
+
 // Deprecated function name.
 #define BG24ToARGB RGB24ToARGB
 

include/libyuv/version.h

@@ -11,6 +11,6 @@
 #ifndef INCLUDE_LIBYUV_VERSION_H_  // NOLINT
 #define INCLUDE_LIBYUV_VERSION_H_
 
-#define LIBYUV_VERSION 350
+#define LIBYUV_VERSION 351
 
 #endif  // INCLUDE_LIBYUV_VERSION_H_  NOLINT

include/libyuv/video_common.h

@@ -36,6 +36,7 @@ extern "C" {
 //   http://v4l2spec.bytesex.org/spec/book1.htm
 //   http://developer.apple.com/quicktime/icefloe/dispatch020.html
 //   http://msdn.microsoft.com/library/windows/desktop/dd206750.aspx#nv12
+//   http://people.xiph.org/~xiphmont/containers/nut/nut4cc.txt
 
 enum FourCC {
   // Canonical fourcc codes used in our code.
@@ -53,9 +54,10 @@ enum FourCC {
   FOURCC_Q420 = FOURCC('Q', '4', '2', '0'),
   FOURCC_V210 = FOURCC('V', '2', '1', '0'),
   FOURCC_24BG = FOURCC('2', '4', 'B', 'G'),
-  FOURCC_ABGR = FOURCC('A', 'B', 'G', 'R'),
-  FOURCC_BGRA = FOURCC('B', 'G', 'R', 'A'),
   FOURCC_ARGB = FOURCC('A', 'R', 'G', 'B'),
+  FOURCC_BGRA = FOURCC('B', 'G', 'R', 'A'),
+  FOURCC_ABGR = FOURCC('A', 'B', 'G', 'R'),
+  FOURCC_RGBA = FOURCC('R', 'G', 'B', 'A'),
   FOURCC_RGBP = FOURCC('R', 'G', 'B', 'P'),  // bgr565.
   FOURCC_RGBO = FOURCC('R', 'G', 'B', 'O'),  // abgr1555.
   FOURCC_R444 = FOURCC('R', '4', '4', '4'),  // argb4444.
@@ -107,9 +109,10 @@ enum FourCCBpp {
   FOURCC_BPP_Q420 = 12,
   FOURCC_BPP_V210 = 22,  // 22.5 actually
   FOURCC_BPP_24BG = 24,
-  FOURCC_BPP_ABGR = 32,
-  FOURCC_BPP_BGRA = 32,
   FOURCC_BPP_ARGB = 32,
+  FOURCC_BPP_BGRA = 32,
+  FOURCC_BPP_ABGR = 32,
+  FOURCC_BPP_RGBA = 32,
   FOURCC_BPP_RGBP = 16,
   FOURCC_BPP_RGBO = 16,
   FOURCC_BPP_R444 = 16,

source/convert_argb.cc

@@ -317,6 +317,39 @@ int BGRAToARGB(const uint8* src_bgra, int src_stride_bgra,
   return 0;
 }
 
+// Convert RGBA to ARGB.
+int RGBAToARGB(const uint8* src_rgba, int src_stride_rgba,
+               uint8* dst_argb, int dst_stride_argb,
+               int width, int height) {
+  if (!src_rgba || !dst_argb ||
+      width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_rgba = src_rgba + (height - 1) * src_stride_rgba;
+    src_stride_rgba = -src_stride_rgba;
+  }
+  void (*RGBAToARGBRow)(const uint8* src_rgba, uint8* dst_argb, int pix) =
+      RGBAToARGBRow_C;
+#if defined(HAS_RGBATOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3) &&
+      IS_ALIGNED(width, 4) &&
+      IS_ALIGNED(src_rgba, 16) && IS_ALIGNED(src_stride_rgba, 16) &&
+      IS_ALIGNED(dst_argb, 16) && IS_ALIGNED(dst_stride_argb, 16)) {
+    RGBAToARGBRow = RGBAToARGBRow_SSSE3;
+  }
+#endif
+
+  for (int y = 0; y < height; ++y) {
+    RGBAToARGBRow(src_rgba, dst_argb, width);
+    src_rgba += src_stride_rgba;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
 // Convert RAW to ARGB.
 int RAWToARGB(const uint8* src_raw, int src_stride_raw,
               uint8* dst_argb, int dst_stride_argb,
@@ -1006,6 +1039,12 @@ int ConvertToARGB(const uint8* sample, size_t sample_size,
                      dst_argb, argb_stride,
                      dst_width, inv_dst_height);
       break;
+    case FOURCC_RGBA:
+      src = sample + (src_width * crop_y + crop_x) * 4;
+      r = RGBAToARGB(src, src_width * 4,
+                     dst_argb, argb_stride,
+                     dst_width, inv_dst_height);
+      break;
     case FOURCC_RGBP:
       src = sample + (src_width * crop_y + crop_x) * 2;
       r = RGB565ToARGB(src, src_width * 2,

source/row.h

@@ -37,6 +37,7 @@ extern "C" {
 #define HAS_ABGRTOARGBROW_SSSE3
 #define HAS_ABGRTOUVROW_SSSE3
 #define HAS_ABGRTOYROW_SSSE3
+#define HAS_RGBATOARGBROW_SSSE3
 #define HAS_ARGB1555TOARGBROW_SSE2
 #define HAS_ARGB4444TOARGBROW_SSE2
 #define HAS_ARGBATTENUATEROW_SSSE3
@@ -215,6 +216,7 @@ void ABGRToUVRow_C(const uint8* src_argb0, int src_stride_argb,
 
 void ABGRToARGBRow_SSSE3(const uint8* src_abgr, uint8* dst_argb, int pix);
 void BGRAToARGBRow_SSSE3(const uint8* src_bgra, uint8* dst_argb, int pix);
+void RGBAToARGBRow_SSSE3(const uint8* src_abgr, uint8* dst_argb, int pix);
 void RGB24ToARGBRow_SSSE3(const uint8* src_rgb24, uint8* dst_argb, int pix);
 void RAWToARGBRow_SSSE3(const uint8* src_rgb24, uint8* dst_argb, int pix);
 void ARGB1555ToARGBRow_SSE2(const uint8* src_argb, uint8* dst_argb, int pix);
@@ -223,6 +225,7 @@ void ARGB4444ToARGBRow_SSE2(const uint8* src_argb, uint8* dst_argb, int pix);
 
 void ABGRToARGBRow_C(const uint8* src_abgr, uint8* dst_argb, int pix);
 void BGRAToARGBRow_C(const uint8* src_bgra, uint8* dst_argb, int pix);
+void RGBAToARGBRow_C(const uint8* src_abgr, uint8* dst_argb, int pix);
 void RGB24ToARGBRow_C(const uint8* src_rgb24, uint8* dst_argb, int pix);
 void RAWToARGBRow_C(const uint8* src_rgb24, uint8* dst_argb, int pix);
 void RGB565ToARGBRow_C(const uint8* src_rgb, uint8* dst_argb, int pix);

source/row_common.cc

@@ -51,6 +51,22 @@ void BGRAToARGBRow_C(const uint8* src_bgra, uint8* dst_argb, int width) {
   }
 }
 
+void RGBAToARGBRow_C(const uint8* src_abgr, uint8* dst_argb, int width) {
+  for (int x = 0; x < width; ++x) {
+    // To support in-place conversion.
+    uint8 a = src_abgr[0];
+    uint8 b = src_abgr[1];
+    uint8 g = src_abgr[2];
+    uint8 r = src_abgr[3];
+    dst_argb[0] = b;
+    dst_argb[1] = g;
+    dst_argb[2] = r;
+    dst_argb[3] = a;
+    dst_argb += 4;
+    src_abgr += 4;
+  }
+}
+
 void RGB24ToARGBRow_C(const uint8* src_rgb24, uint8* dst_argb, int width) {
   for (int x = 0; x < width; ++x) {
     uint8 b = src_rgb24[0];

source/row_posix.cc

@@ -98,6 +98,11 @@ CONST uvec8 kShuffleMaskBGRAToARGB = {
   3u, 2u, 1u, 0u, 7u, 6u, 5u, 4u, 11u, 10u, 9u, 8u, 15u, 14u, 13u, 12u
 };
 
+// Shuffle table for converting RGBA to ARGB.
+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 RGB24.
 CONST uvec8 kShuffleMaskARGBToRGB24 = {
   0u, 1u, 2u, 4u, 5u, 6u, 8u, 9u, 10u, 12u, 13u, 14u, 128u, 128u, 128u, 128u
@@ -185,6 +190,30 @@ void BGRAToARGBRow_SSSE3(const uint8* src_bgra, uint8* dst_argb, int pix) {
   );
 }
 
+void RGBAToARGBRow_SSSE3(const uint8* src_rgba, uint8* dst_argb, 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_rgba),  // %0
+    "+r"(dst_argb),  // %1
+    "+r"(pix)        // %2
+  : "m"(kShuffleMaskRGBAToARGB)  // %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

@@ -88,6 +88,11 @@ static const uvec8 kShuffleMaskBGRAToARGB = {
   3u, 2u, 1u, 0u, 7u, 6u, 5u, 4u, 11u, 10u, 9u, 8u, 15u, 14u, 13u, 12u
 };
 
+// Shuffle table for converting RGBA to ARGB.
+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 RGB24.
 static const uvec8 kShuffleMaskARGBToRGB24 = {
   0u, 1u, 2u, 4u, 5u, 6u, 8u, 9u, 10u, 12u, 13u, 14u, 128u, 128u, 128u, 128u
@@ -168,6 +173,27 @@ __asm {
   }
 }
 
+__declspec(naked) __declspec(align(16))
+void RGBAToARGBRow_SSSE3(const uint8* src_rgba, uint8* dst_argb, int pix) {
+__asm {
+    mov       eax, [esp + 4]   // src_rgba
+    mov       edx, [esp + 8]   // dst_argb
+    mov       ecx, [esp + 12]  // pix
+    movdqa    xmm5, kShuffleMaskRGBAToARGB
+    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

@@ -306,6 +306,7 @@ TESTATOB(ARGB, 4, 4, ARGB1555, 2)
 TESTATOB(ARGB, 4, 4, ARGB4444, 2)
 TESTATOB(BGRA, 4, 4, ARGB, 4)
 TESTATOB(ABGR, 4, 4, ARGB, 4)
+TESTATOB(RGBA, 4, 4, ARGB, 4)
 TESTATOB(RAW, 3, 3, ARGB, 4)
 TESTATOB(RGB24, 3, 3, ARGB, 4)
 TESTATOB(RGB565, 2, 2, ARGB, 4)
@@ -362,6 +363,7 @@ TESTATOBRANDOM(ARGB, 4, 4, ARGB4444, 2)
 
 TESTATOBRANDOM(BGRA, 4, 4, ARGB, 4)
 TESTATOBRANDOM(ABGR, 4, 4, ARGB, 4)
+TESTATOBRANDOM(RGBA, 4, 4, ARGB, 4)
 TESTATOBRANDOM(RAW, 3, 3, ARGB, 4)
 TESTATOBRANDOM(RGB24, 3, 3, ARGB, 4)
 TESTATOBRANDOM(RGB565, 2, 2, ARGB, 4)