[WIP] Supporting arm_neon_optional flag for blink platform.

Source/platform/graphics/cpu/arm/gpu/WebGLImageConversionNEON.h

 /*
  * Copyright (C) 2012 Gabor Rapcsanyi (rgabor@inf.u-szeged.hu), University of Szeged
  *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
  * 1. Redistributions of source code must retain the above copyright
  *    notice, this list of conditions and the following disclaimer.
  * 2. Redistributions in binary form must reproduce the above copyright
  *    notice, this list of conditions and the following disclaimer in the
@@ skipping 16 matching lines @@
 #define WebGLImageConversionNEON_h
 
 #if HAVE(ARM_NEON_INTRINSICS)
 
 #include <arm_neon.h>
 
 namespace blink {
 
 namespace SIMD {
 
-ALWAYS_INLINE void unpackOneRowOfRGBA16LittleToRGBA8(const uint16_t*& source, uint8_t*& destination, unsigned& pixelsPerRow)
+ALWAYS_INLINE void unpackOneRowOfRGBA16LittleToRGBA8NEON(const uint16_t*& source, uint8_t*& destination, unsigned& pixelsPerRow)
 {
     unsigned componentsPerRow = pixelsPerRow * 4;
     unsigned tailComponents = componentsPerRow % 16;
     unsigned componentsSize = componentsPerRow - tailComponents;
     const uint8_t* src = reinterpret_cast<const uint8_t*>(source);
 
     for (unsigned i = 0; i < componentsSize; i += 16) {
         uint8x16x2_t components = vld2q_u8(src + i * 2);
         vst1q_u8(destination + i, components.val[1]);
     }
 
     source += componentsSize;
     destination += componentsSize;
     pixelsPerRow = tailComponents / 4;
 }
 
-ALWAYS_INLINE void unpackOneRowOfRGB16LittleToRGBA8(const uint16_t*& source, uint8_t*& destination, unsigned& pixelsPerRow)
+ALWAYS_INLINE void unpackOneRowOfRGB16LittleToRGBA8NEON(const uint16_t*& source, uint8_t*& destination, unsigned& pixelsPerRow)
 {
     unsigned componentsPerRow = pixelsPerRow * 3;
     unsigned tailComponents = componentsPerRow % 24;
     unsigned componentsSize = componentsPerRow - tailComponents;
 
     uint8x8_t componentA = vdup_n_u8(0xFF);
     for (unsigned i = 0; i < componentsSize; i += 24) {
         uint16x8x3_t RGB16 = vld3q_u16(source + i);
         uint8x8_t componentR = vqmovn_u16(vshrq_n_u16(RGB16.val[0], 8));
         uint8x8_t componentG = vqmovn_u16(vshrq_n_u16(RGB16.val[1], 8));
         uint8x8_t componentB = vqmovn_u16(vshrq_n_u16(RGB16.val[2], 8));
         uint8x8x4_t RGBA8 = {{componentR, componentG, componentB, componentA}};
         vst4_u8(destination, RGBA8);
         destination += 32;
     }
 
     source += componentsSize;
     pixelsPerRow = tailComponents / 3;
 }
 
-ALWAYS_INLINE void unpackOneRowOfARGB16LittleToRGBA8(const uint16_t*& source, uint8_t*& destination, unsigned& pixelsPerRow)
+ALWAYS_INLINE void unpackOneRowOfARGB16LittleToRGBA8NEON(const uint16_t*& source, uint8_t*& destination, unsigned& pixelsPerRow)
 {
     unsigned componentsPerRow = pixelsPerRow * 4;
     unsigned tailComponents = componentsPerRow % 32;
     unsigned componentsSize = componentsPerRow - tailComponents;
 
     for (unsigned i = 0; i < componentsSize; i += 32) {
         uint16x8x4_t ARGB16 = vld4q_u16(source + i);
         uint8x8_t componentA = vqmovn_u16(vshrq_n_u16(ARGB16.val[0], 8));
         uint8x8_t componentR = vqmovn_u16(vshrq_n_u16(ARGB16.val[1], 8));
         uint8x8_t componentG = vqmovn_u16(vshrq_n_u16(ARGB16.val[2], 8));
         uint8x8_t componentB = vqmovn_u16(vshrq_n_u16(ARGB16.val[3], 8));
         uint8x8x4_t RGBA8 = {{componentR, componentG, componentB, componentA}};
         vst4_u8(destination + i, RGBA8);
     }
 
     source += componentsSize;
     destination += componentsSize;
     pixelsPerRow = tailComponents / 4;
 }
 
-ALWAYS_INLINE void unpackOneRowOfBGRA16LittleToRGBA8(const uint16_t*& source, uint8_t*& destination, unsigned& pixelsPerRow)
+ALWAYS_INLINE void unpackOneRowOfBGRA16LittleToRGBA8NEON(const uint16_t*& source, uint8_t*& destination, unsigned& pixelsPerRow)
 {
     unsigned componentsPerRow = pixelsPerRow * 4;
     unsigned tailComponents = componentsPerRow % 32;
     unsigned componentsSize = componentsPerRow - tailComponents;
 
     for (unsigned i = 0; i < componentsSize; i += 32) {
         uint16x8x4_t ARGB16 = vld4q_u16(source + i);
         uint8x8_t componentB = vqmovn_u16(vshrq_n_u16(ARGB16.val[0], 8));
         uint8x8_t componentG = vqmovn_u16(vshrq_n_u16(ARGB16.val[1], 8));
         uint8x8_t componentR = vqmovn_u16(vshrq_n_u16(ARGB16.val[2], 8));
         uint8x8_t componentA = vqmovn_u16(vshrq_n_u16(ARGB16.val[3], 8));
         uint8x8x4_t RGBA8 = {{componentR, componentG, componentB, componentA}};
         vst4_u8(destination + i, RGBA8);
     }
 
     source += componentsSize;
     destination += componentsSize;
     pixelsPerRow = tailComponents / 4;
 }
 
-ALWAYS_INLINE void unpackOneRowOfRGBA4444ToRGBA8(const uint16_t*& source, uint8_t*& destination, unsigned& pixelsPerRow)
+ALWAYS_INLINE void unpackOneRowOfRGBA4444ToRGBA8NEON(const uint16_t*& source, uint8_t*& destination, unsigned& pixelsPerRow)
 {
     unsigned tailPixels = pixelsPerRow % 8;
     unsigned pixelSize = pixelsPerRow - tailPixels;
 
     uint16x8_t immediate0x0f = vdupq_n_u16(0x0F);
     for (unsigned i = 0; i < pixelSize; i += 8) {
         uint16x8_t eightPixels = vld1q_u16(source + i);
 
         uint8x8_t componentR = vqmovn_u16(vshrq_n_u16(eightPixels, 12));
         uint8x8_t componentG = vqmovn_u16(vandq_u16(vshrq_n_u16(eightPixels, 8), immediate0x0f));
         uint8x8_t componentB = vqmovn_u16(vandq_u16(vshrq_n_u16(eightPixels, 4), immediate0x0f));
         uint8x8_t componentA = vqmovn_u16(vandq_u16(eightPixels, immediate0x0f));
 
         componentR = vorr_u8(vshl_n_u8(componentR, 4), componentR);
         componentG = vorr_u8(vshl_n_u8(componentG, 4), componentG);
         componentB = vorr_u8(vshl_n_u8(componentB, 4), componentB);
         componentA = vorr_u8(vshl_n_u8(componentA, 4), componentA);
 
         uint8x8x4_t destComponents = {{componentR, componentG, componentB, componentA}};
         vst4_u8(destination, destComponents);
         destination += 32;
     }
 
     source += pixelSize;
     pixelsPerRow = tailPixels;
 }
 
-ALWAYS_INLINE void packOneRowOfRGBA8ToUnsignedShort4444(const uint8_t*& source, uint16_t*& destination, unsigned& pixelsPerRow)
+ALWAYS_INLINE void packOneRowOfRGBA8ToUnsignedShort4444NEON(const uint8_t*& source, uint16_t*& destination, unsigned& pixelsPerRow)
 {
     unsigned componentsPerRow = pixelsPerRow * 4;
     unsigned tailComponents = componentsPerRow % 32;
     unsigned componentsSize = componentsPerRow - tailComponents;
 
     uint8_t* dst = reinterpret_cast<uint8_t*>(destination);
     uint8x8_t immediate0xf0 = vdup_n_u8(0xF0);
     for (unsigned i = 0; i < componentsSize; i += 32) {
         uint8x8x4_t RGBA8 = vld4_u8(source + i);
 
         uint8x8_t componentR = vand_u8(RGBA8.val[0], immediate0xf0);
         uint8x8_t componentG = vshr_n_u8(vand_u8(RGBA8.val[1], immediate0xf0), 4);
         uint8x8_t componentB = vand_u8(RGBA8.val[2], immediate0xf0);
         uint8x8_t componentA = vshr_n_u8(vand_u8(RGBA8.val[3], immediate0xf0), 4);
 
         uint8x8x2_t RGBA4;
         RGBA4.val[0] = vorr_u8(componentB, componentA);
         RGBA4.val[1] = vorr_u8(componentR, componentG);
         vst2_u8(dst, RGBA4);
         dst += 16;
     }
 
     source += componentsSize;
     destination += componentsSize / 4;
     pixelsPerRow = tailComponents / 4;
 }
 
-ALWAYS_INLINE void unpackOneRowOfRGBA5551ToRGBA8(const uint16_t*& source, uint8_t*& destination, unsigned& pixelsPerRow)
+ALWAYS_INLINE void unpackOneRowOfRGBA5551ToRGBA8NEON(const uint16_t*& source, uint8_t*& destination, unsigned& pixelsPerRow)
 {
     unsigned tailPixels = pixelsPerRow % 8;
     unsigned pixelSize = pixelsPerRow - tailPixels;
 
     uint8x8_t immediate0x7 = vdup_n_u8(0x7);
     uint8x8_t immediate0xff = vdup_n_u8(0xFF);
     uint16x8_t immediate0x1f = vdupq_n_u16(0x1F);
     uint16x8_t immediate0x1 = vdupq_n_u16(0x1);
 
     for (unsigned i = 0; i < pixelSize; i += 8) {
@@ skipping 11 matching lines @@
 
         uint8x8x4_t destComponents = {{componentR, componentG, componentB, componentA}};
         vst4_u8(destination, destComponents);
         destination += 32;
     }
 
     source += pixelSize;
     pixelsPerRow = tailPixels;
 }
 
-ALWAYS_INLINE void packOneRowOfRGBA8ToUnsignedShort5551(const uint8_t*& source, uint16_t*& destination, unsigned& pixelsPerRow)
+ALWAYS_INLINE void packOneRowOfRGBA8ToUnsignedShort5551NEON(const uint8_t*& source, uint16_t*& destination, unsigned& pixelsPerRow)
 {
     unsigned componentsPerRow = pixelsPerRow * 4;
     unsigned tailComponents = componentsPerRow % 32;
     unsigned componentsSize = componentsPerRow - tailComponents;
 
     uint8_t* dst = reinterpret_cast<uint8_t*>(destination);
 
     uint8x8_t immediate0xf8 = vdup_n_u8(0xF8);
     uint8x8_t immediate0x18 = vdup_n_u8(0x18);
     for (unsigned i = 0; i < componentsSize; i += 32) {
@@ skipping 11 matching lines @@
         RGBA5551.val[1] = vorr_u8(componentR, componentG3bit);
         vst2_u8(dst, RGBA5551);
         dst += 16;
     }
 
     source += componentsSize;
     destination += componentsSize / 4;
     pixelsPerRow = tailComponents / 4;
 }
 
-ALWAYS_INLINE void unpackOneRowOfRGB565ToRGBA8(const uint16_t*& source, uint8_t*& destination, unsigned& pixelsPerRow)
+ALWAYS_INLINE void unpackOneRowOfRGB565ToRGBA8NEON(const uint16_t*& source, uint8_t*& destination, unsigned& pixelsPerRow)
 {
     unsigned tailPixels = pixelsPerRow % 8;
     unsigned pixelSize = pixelsPerRow - tailPixels;
 
     uint16x8_t immediate0x3f = vdupq_n_u16(0x3F);
     uint16x8_t immediate0x1f = vdupq_n_u16(0x1F);
     uint8x8_t immediate0x3 = vdup_n_u8(0x3);
     uint8x8_t immediate0x7 = vdup_n_u8(0x7);
 
     uint8x8_t componentA = vdup_n_u8(0xFF);
@@ skipping 11 matching lines @@
 
         uint8x8x4_t destComponents = {{componentR, componentG, componentB, componentA}};
         vst4_u8(destination, destComponents);
         destination += 32;
     }
 
     source += pixelSize;
     pixelsPerRow = tailPixels;
 }
 
-ALWAYS_INLINE void packOneRowOfRGBA8ToUnsignedShort565(const uint8_t*& source, uint16_t*& destination, unsigned& pixelsPerRow)
+ALWAYS_INLINE void packOneRowOfRGBA8ToUnsignedShort565NEON(const uint8_t*& source, uint16_t*& destination, unsigned& pixelsPerRow)
 {
     unsigned componentsPerRow = pixelsPerRow * 4;
     unsigned tailComponents = componentsPerRow % 32;
     unsigned componentsSize = componentsPerRow - tailComponents;
     uint8_t* dst = reinterpret_cast<uint8_t*>(destination);
 
     uint8x8_t immediate0xf8 = vdup_n_u8(0xF8);
     uint8x8_t immediate0x1c = vdup_n_u8(0x1C);
     for (unsigned i = 0; i < componentsSize; i += 32) {
         uint8x8x4_t RGBA8 = vld4_u8(source + i);
@@ skipping 15 matching lines @@
     pixelsPerRow = tailComponents / 4;
 }
 
 } // namespace SIMD
 
 } // namespace blink
 
 #endif // HAVE(ARM_NEON_INTRINSICS)
 
 #endif // WebGLImageConversionNEON_h