| #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) |
| |
| struct tint_symbol_6 { |
| tint_array<float4x3, 4> w; |
| }; |
| |
| struct tint_packed_vec3_f32_array_element { |
| /* 0x0000 */ packed_float3 elements; |
| /* 0x000c */ tint_array<int8_t, 4> tint_pad; |
| }; |
| |
| float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) { |
| float4x3 result = float4x3(float3(in[0].elements), float3(in[1].elements), float3(in[2].elements), float3(in[3].elements)); |
| return result; |
| } |
| |
| tint_array<float4x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 4>, 4> in) { |
| tint_array<float4x3, 4> result = tint_array<float4x3, 4>{tint_unpack_vec3_in_composite(in[0]), tint_unpack_vec3_in_composite(in[1]), tint_unpack_vec3_in_composite(in[2]), tint_unpack_vec3_in_composite(in[3])}; |
| return result; |
| } |
| |
| void tint_zero_workgroup_memory(uint local_idx, threadgroup tint_array<float4x3, 4>* const tint_symbol) { |
| TINT_ISOLATE_UB(tint_volatile_true) for(uint idx = local_idx; (idx < 4u); idx = (idx + 1u)) { |
| uint const i = idx; |
| (*(tint_symbol))[i] = float4x3(float3(0.0f), float3(0.0f), float3(0.0f), float3(0.0f)); |
| } |
| threadgroup_barrier(mem_flags::mem_threadgroup); |
| } |
| |
| void f_inner(uint local_invocation_index, threadgroup tint_array<float4x3, 4>* const tint_symbol_1, const constant tint_array<tint_array<tint_packed_vec3_f32_array_element, 4>, 4>* const tint_symbol_2) { |
| tint_zero_workgroup_memory(local_invocation_index, tint_symbol_1); |
| *(tint_symbol_1) = tint_unpack_vec3_in_composite_1(*(tint_symbol_2)); |
| (*(tint_symbol_1))[1] = tint_unpack_vec3_in_composite((*(tint_symbol_2))[2]); |
| (*(tint_symbol_1))[1][0] = float3((*(tint_symbol_2))[0][1].elements).zxy; |
| (*(tint_symbol_1))[1][0][0] = (*(tint_symbol_2))[0][1].elements[0]; |
| } |
| |
| kernel void f(const constant tint_array<tint_array<tint_packed_vec3_f32_array_element, 4>, 4>* tint_symbol_5 [[buffer(0)]], threadgroup tint_symbol_6* tint_symbol_4 [[threadgroup(0)]], uint local_invocation_index [[thread_index_in_threadgroup]]) { |
| threadgroup tint_array<float4x3, 4>* const tint_symbol_3 = &((*(tint_symbol_4)).w); |
| f_inner(local_invocation_index, tint_symbol_3, tint_symbol_5); |
| return; |
| } |
| |