| #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 tint_private_vars_struct { |
| half4x3 m; |
| }; |
| |
| struct tint_packed_vec3_f16_array_element { |
| /* 0x0000 */ packed_half3 elements; |
| /* 0x0006 */ tint_array<int8_t, 2> tint_pad; |
| }; |
| |
| void assign_and_preserve_padding(device tint_array<tint_packed_vec3_f16_array_element, 4>* const dest, half4x3 value) { |
| (*(dest))[0u].elements = packed_half3(value[0u]); |
| (*(dest))[1u].elements = packed_half3(value[1u]); |
| (*(dest))[2u].elements = packed_half3(value[2u]); |
| (*(dest))[3u].elements = packed_half3(value[3u]); |
| } |
| |
| kernel void f(device tint_array<tint_packed_vec3_f16_array_element, 4>* tint_symbol [[buffer(0)]]) { |
| thread tint_private_vars_struct tint_private_vars = {}; |
| tint_private_vars.m = half4x3(half3(0.0h, 1.0h, 2.0h), half3(3.0h, 4.0h, 5.0h), half3(6.0h, 7.0h, 8.0h), half3(9.0h, 10.0h, 11.0h)); |
| assign_and_preserve_padding(tint_symbol, tint_private_vars.m); |
| return; |
| } |
| |