| #include <metal_stdlib> |
| |
| using namespace metal; |
| |
| template<typename T, size_t N> |
| struct tint_array { |
| const constant T& operator[](size_t i) const constant { return elements[i]; } |
| device T& operator[](size_t i) device { return elements[i]; } |
| const device T& operator[](size_t i) const device { return elements[i]; } |
| thread T& operator[](size_t i) thread { return elements[i]; } |
| const thread T& operator[](size_t i) const thread { return elements[i]; } |
| threadgroup T& operator[](size_t i) threadgroup { return elements[i]; } |
| const threadgroup T& operator[](size_t i) const threadgroup { return elements[i]; } |
| T elements[N]; |
| }; |
| |
| #define TINT_ISOLATE_UB(VOLATILE_NAME) \ |
| volatile bool VOLATILE_NAME = true; \ |
| if (VOLATILE_NAME) |
| |
| uint4 tint_ftou(float4 v) { |
| return select(uint4(4294967295u), select(uint4(v), uint4(0u), (v < float4(0.0f))), (v <= float4(4294967040.0f))); |
| } |
| |
| struct Uniforms { |
| /* 0x0000 */ uint dstTextureFlipY; |
| /* 0x0004 */ uint isFloat16; |
| /* 0x0008 */ uint isRGB10A2Unorm; |
| /* 0x000c */ uint channelCount; |
| }; |
| |
| struct OutputBuf { |
| /* 0x0000 */ tint_array<uint, 1> result; |
| }; |
| |
| uint ConvertToFp16FloatValue(float fp32) { |
| return 1u; |
| } |
| |
| void tint_symbol_inner(uint3 GlobalInvocationID, texture2d<float, access::sample> tint_symbol_2, const constant Uniforms* const tint_symbol_3, texture2d<float, access::sample> tint_symbol_4, device OutputBuf* const tint_symbol_5) { |
| uint2 size = uint2(tint_symbol_2.get_width(), tint_symbol_2.get_height()); |
| uint2 dstTexCoord = GlobalInvocationID.xy; |
| uint2 srcTexCoord = dstTexCoord; |
| if (((*(tint_symbol_3)).dstTextureFlipY == 1u)) { |
| srcTexCoord[1] = ((size[1] - dstTexCoord[1]) - 1u); |
| } |
| float4 srcColor = tint_symbol_2.read(uint2(srcTexCoord), 0); |
| float4 dstColor = tint_symbol_4.read(uint2(dstTexCoord), 0); |
| bool success = true; |
| uint4 srcColorBits = 0u; |
| uint4 dstColorBits = tint_ftou(dstColor); |
| TINT_ISOLATE_UB(tint_volatile_true) for(uint i = 0u; (i < (*(tint_symbol_3)).channelCount); i = (i + 1u)) { |
| uint const tint_symbol_1 = i; |
| srcColorBits[tint_symbol_1] = ConvertToFp16FloatValue(srcColor[i]); |
| success = (success && (srcColorBits[i] == dstColorBits[i])); |
| } |
| uint outputIndex = ((GlobalInvocationID[1] * uint(size[0])) + GlobalInvocationID[0]); |
| if (success) { |
| (*(tint_symbol_5)).result[outputIndex] = 1u; |
| } else { |
| (*(tint_symbol_5)).result[outputIndex] = 0u; |
| } |
| } |
| |
| kernel void tint_symbol(texture2d<float, access::sample> tint_symbol_6 [[texture(0)]], const constant Uniforms* tint_symbol_7 [[buffer(0)]], texture2d<float, access::sample> tint_symbol_8 [[texture(1)]], device OutputBuf* tint_symbol_9 [[buffer(1)]], uint3 GlobalInvocationID [[thread_position_in_grid]]) { |
| tint_symbol_inner(GlobalInvocationID, tint_symbol_6, tint_symbol_7, tint_symbol_8, tint_symbol_9); |
| return; |
| } |
| |