| #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]; |
| }; |
| |
| struct S { |
| /* 0x0000 */ int before; |
| /* 0x0004 */ tint_array<int8_t, 4> tint_pad; |
| /* 0x0008 */ float4x2 m; |
| /* 0x0028 */ int after; |
| /* 0x002c */ tint_array<int8_t, 4> tint_pad_1; |
| }; |
| |
| struct tint_symbol_6 { |
| tint_array<S, 4> w; |
| }; |
| |
| void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S, 4>* const tint_symbol_2) { |
| for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) { |
| uint const i = idx; |
| S const tint_symbol = S{}; |
| (*(tint_symbol_1))[i] = tint_symbol; |
| } |
| threadgroup_barrier(mem_flags::mem_threadgroup); |
| *(tint_symbol_1) = *(tint_symbol_2); |
| (*(tint_symbol_1))[1] = (*(tint_symbol_2))[2]; |
| (*(tint_symbol_1))[3].m = (*(tint_symbol_2))[2].m; |
| (*(tint_symbol_1))[1].m[0] = float2((*(tint_symbol_2))[0].m[1]).yx; |
| } |
| |
| kernel void f(const constant tint_array<S, 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<S, 4>* const tint_symbol_3 = &((*(tint_symbol_4)).w); |
| f_inner(local_invocation_index, tint_symbol_3, tint_symbol_5); |
| return; |
| } |
| |