Add MSA (MIPS SIMD Arch) optimized WebGL image conversion functions

third_party/WebKit/Source/platform/graphics/gpu/WebGLImageConversion.cpp

 // Copyright 2014 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "platform/graphics/gpu/WebGLImageConversion.h"
 
 #include "platform/CheckedInt.h"
 #include "platform/graphics/ImageObserver.h"
 #include "platform/graphics/cpu/arm/WebGLImageConversionNEON.h"
 #include "platform/graphics/cpu/mips/WebGLImageConversionMSA.h"
@@ skipping 409 matching lines @@
 }
 
 template<> void unpack<WebGLImageConversion::DataFormatBGRA8, uint8_t, uint8_t>(const uint8_t* source, uint8_t* destination, unsigned pixelsPerRow)
 {
     const uint32_t* source32 = reinterpret_cast_ptr<const uint32_t*>(source);
     uint32_t* destination32 = reinterpret_cast_ptr<uint32_t*>(destination);
 
 #if CPU(X86) || CPU(X86_64)
     SIMD::unpackOneRowOfBGRA8LittleToRGBA8(source32, destination32, pixelsPerRow);
 #endif
+#if HAVE(MIPS_MSA_INTRINSICS)
+    SIMD::unpackOneRowOfBGRA8LittleToRGBA8MSA(source32, destination32, pixelsPerRow);
+#endif
     for (unsigned i = 0; i < pixelsPerRow; ++i) {
         uint32_t bgra = source32[i];
 #if CPU(BIG_ENDIAN)
         uint32_t brMask = 0xff00ff00;
         uint32_t gaMask = 0x00ff00ff;
 #else
         uint32_t brMask = 0x00ff00ff;
         uint32_t gaMask = 0xff00ff00;
 #endif
         uint32_t rgba = (((bgra >> 16) | (bgra << 16)) & brMask) | (bgra & gaMask);
@@ skipping 28 matching lines @@
 }
 
 template<> void unpack<WebGLImageConversion::DataFormatRGBA4444, uint16_t, uint8_t>(const uint16_t* source, uint8_t* destination, unsigned pixelsPerRow)
 {
 #if CPU(X86) || CPU(X86_64)
     SIMD::unpackOneRowOfRGBA4444LittleToRGBA8(source, destination, pixelsPerRow);
 #endif
 #if HAVE(ARM_NEON_INTRINSICS)
     SIMD::unpackOneRowOfRGBA4444ToRGBA8(source, destination, pixelsPerRow);
 #endif
+#if HAVE(MIPS_MSA_INTRINSICS)
+    SIMD::unpackOneRowOfRGBA4444ToRGBA8MSA(source, destination, pixelsPerRow);
+#endif
     for (unsigned i = 0; i < pixelsPerRow; ++i) {
         uint16_t packedValue = source[0];
         uint8_t r = packedValue >> 12;
         uint8_t g = (packedValue >> 8) & 0x0F;
         uint8_t b = (packedValue >> 4) & 0x0F;
         uint8_t a = packedValue & 0x0F;
         destination[0] = r << 4 | r;
         destination[1] = g << 4 | g;
         destination[2] = b << 4 | b;
         destination[3] = a << 4 | a;
@@ skipping 253 matching lines @@
         source += 4;
         destination += 4;
     }
 }
 
 // FIXME: this routine is lossy and must be removed.
 template<> void pack<WebGLImageConversion::DataFormatRGBA8, WebGLImageConversion::AlphaDoUnmultiply, uint8_t, uint8_t>(const uint8_t* source, uint8_t* destination, unsigned pixelsPerRow)
 {
 #if CPU(X86) || CPU(X86_64)
     SIMD::packOneRowOfRGBA8LittleToRGBA8(source, destination, pixelsPerRow);
-#else
+#endif
+#if HAVE(MIPS_MSA_INTRINSICS)
+    SIMD::packOneRowOfRGBA8LittleToRGBA8MSA(source, destination, pixelsPerRow);
+#endif
     for (unsigned i = 0; i < pixelsPerRow; ++i) {
         float scaleFactor = source[3] ? 255.0f / source[3] : 1.0f;
         uint8_t sourceR = static_cast<uint8_t>(static_cast<float>(source[0]) * scaleFactor);
         uint8_t sourceG = static_cast<uint8_t>(static_cast<float>(source[1]) * scaleFactor);
         uint8_t sourceB = static_cast<uint8_t>(static_cast<float>(source[2]) * scaleFactor);
         destination[0] = sourceR;
         destination[1] = sourceG;
         destination[2] = sourceB;
         destination[3] = source[3];
         source += 4;
         destination += 4;
     }
-#endif
 }
 
 template<> void pack<WebGLImageConversion::DataFormatRGBA4444, WebGLImageConversion::AlphaDoNothing, uint8_t, uint16_t>(const uint8_t* source, uint16_t* destination, unsigned pixelsPerRow)
 {
 #if HAVE(ARM_NEON_INTRINSICS)
     SIMD::packOneRowOfRGBA8ToUnsignedShort4444(source, destination, pixelsPerRow);
 #endif
     for (unsigned i = 0; i < pixelsPerRow; ++i) {
         *destination = (((source[0] & 0xF0) << 8)
                         | ((source[1] & 0xF0) << 4)
@@ skipping 35 matching lines @@
         source += 4;
         destination += 1;
     }
 }
 
 template<> void pack<WebGLImageConversion::DataFormatRGBA5551, WebGLImageConversion::AlphaDoNothing, uint8_t, uint16_t>(const uint8_t* source, uint16_t* destination, unsigned pixelsPerRow)
 {
 #if HAVE(ARM_NEON_INTRINSICS)
     SIMD::packOneRowOfRGBA8ToUnsignedShort5551(source, destination, pixelsPerRow);
 #endif
+#if HAVE(MIPS_MSA_INTRINSICS)
+    SIMD::packOneRowOfRGBA8ToUnsignedShort5551MSA(source, destination, pixelsPerRow);
+#endif
     for (unsigned i = 0; i < pixelsPerRow; ++i) {
         *destination = (((source[0] & 0xF8) << 8)
                         | ((source[1] & 0xF8) << 3)
                         | ((source[2] & 0xF8) >> 2)
                         | (source[3] >> 7));
         source += 4;
         destination += 1;
     }
 }
 
@@ skipping 28 matching lines @@
         source += 4;
         destination += 1;
     }
 }
 
 template<> void pack<WebGLImageConversion::DataFormatRGB565, WebGLImageConversion::AlphaDoNothing, uint8_t, uint16_t>(const uint8_t* source, uint16_t* destination, unsigned pixelsPerRow)
 {
 #if HAVE(ARM_NEON_INTRINSICS)
     SIMD::packOneRowOfRGBA8ToUnsignedShort565(source, destination, pixelsPerRow);
 #endif
+#if HAVE(MIPS_MSA_INTRINSICS)
+    SIMD::packOneRowOfRGBA8ToUnsignedShort565MSA(source, destination, pixelsPerRow);
+#endif
     for (unsigned i = 0; i < pixelsPerRow; ++i) {
         *destination = (((source[0] & 0xF8) << 8)
                         | ((source[1] & 0xFC) << 3)
                         | ((source[2] & 0xF8) >> 3));
         source += 4;
         destination += 1;
     }
 }
 
 template<> void pack<WebGLImageConversion::DataFormatRGB565, WebGLImageConversion::AlphaDoPremultiply, uint8_t, uint16_t>(const uint8_t* source, uint16_t* destination, unsigned pixelsPerRow)
@@ skipping 1529 matching lines @@
     }
 
     FormatConverter converter(width, height, sourceData, destinationData, srcStride, dstStride);
     converter.convert(sourceDataFormat, dstDataFormat, alphaOp);
     if (!converter.Success())
         return false;
     return true;
 }
 
 } // namespace blink