| #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_module_vars_struct { |
| const constant tint_array<float3x4, 4>* u; |
| threadgroup tint_array<float3x4, 4>* w; |
| }; |
| |
| struct tint_symbol_1 { |
| tint_array<float3x4, 4> tint_symbol; |
| }; |
| |
| void f_inner(uint tint_local_index, tint_module_vars_struct tint_module_vars) { |
| { |
| uint v = 0u; |
| v = tint_local_index; |
| while(true) { |
| uint const v_1 = v; |
| if ((v_1 >= 4u)) { |
| break; |
| } |
| (*tint_module_vars.w)[v_1] = float3x4(float4(0.0f), float4(0.0f), float4(0.0f)); |
| { |
| v = (v_1 + 1u); |
| } |
| continue; |
| } |
| } |
| threadgroup_barrier(mem_flags::mem_threadgroup); |
| (*tint_module_vars.w) = (*tint_module_vars.u); |
| (*tint_module_vars.w)[1u] = (*tint_module_vars.u)[2u]; |
| (*tint_module_vars.w)[1u][0u] = (*tint_module_vars.u)[0u][1u].ywxz; |
| (*tint_module_vars.w)[1u][0u].x = (*tint_module_vars.u)[0u][1u].x; |
| } |
| |
| kernel void f(uint tint_local_index [[thread_index_in_threadgroup]], const constant tint_array<float3x4, 4>* u [[buffer(0)]], threadgroup tint_symbol_1* v_2 [[threadgroup(0)]]) { |
| tint_module_vars_struct const tint_module_vars = tint_module_vars_struct{.u=u, .w=(&(*v_2).tint_symbol)}; |
| f_inner(tint_local_index, tint_module_vars); |
| } |