| #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 = false; if (VOLATILE_NAME) break;} |
| |
| struct tint_private_vars_struct { |
| uint local_invocation_index_1; |
| }; |
| |
| void tint_zero_workgroup_memory(uint local_idx, threadgroup tint_array<tint_array<tint_array<atomic_uint, 1>, 2>, 3>* const tint_symbol) { |
| for(uint idx_1 = local_idx; (idx_1 < 6u); idx_1 = (idx_1 + 1u)) { |
| TINT_ISOLATE_UB(tint_volatile_false); |
| uint const i = (idx_1 / 2u); |
| uint const i_1 = (idx_1 % 2u); |
| uint const i_2 = (idx_1 % 1u); |
| atomic_store_explicit(&((*(tint_symbol))[i][i_1][i_2]), 0u, memory_order_relaxed); |
| } |
| threadgroup_barrier(mem_flags::mem_threadgroup); |
| } |
| |
| uint tint_div(uint lhs, uint rhs) { |
| return (lhs / select(rhs, 1u, (rhs == 0u))); |
| } |
| |
| uint tint_mod(uint lhs, uint rhs) { |
| return (lhs % select(rhs, 1u, (rhs == 0u))); |
| } |
| |
| void compute_main_inner(uint local_invocation_index_2, threadgroup tint_array<tint_array<tint_array<atomic_uint, 1>, 2>, 3>* const tint_symbol_1) { |
| uint idx = 0u; |
| idx = local_invocation_index_2; |
| while(true) { |
| TINT_ISOLATE_UB(tint_volatile_false_1); |
| if (!((idx < 6u))) { |
| break; |
| } |
| uint const x_31 = idx; |
| uint const x_33 = idx; |
| uint const x_35 = idx; |
| atomic_store_explicit(&((*(tint_symbol_1))[min(tint_div(x_31, 2u), 2u)][min(tint_mod(x_33, 2u), 1u)][min(tint_mod(x_35, 1u), 0u)]), 0u, memory_order_relaxed); |
| { |
| idx = (idx + 1u); |
| } |
| } |
| threadgroup_barrier(mem_flags::mem_threadgroup); |
| atomic_store_explicit(&((*(tint_symbol_1))[2][1][0]), 1u, memory_order_relaxed); |
| return; |
| } |
| |
| void compute_main_1(thread tint_private_vars_struct* const tint_private_vars, threadgroup tint_array<tint_array<tint_array<atomic_uint, 1>, 2>, 3>* const tint_symbol_2) { |
| uint const x_57 = (*(tint_private_vars)).local_invocation_index_1; |
| compute_main_inner(x_57, tint_symbol_2); |
| return; |
| } |
| |
| void compute_main_inner_1(uint local_invocation_index_1_param, thread tint_private_vars_struct* const tint_private_vars, threadgroup tint_array<tint_array<tint_array<atomic_uint, 1>, 2>, 3>* const tint_symbol_3) { |
| tint_zero_workgroup_memory(local_invocation_index_1_param, tint_symbol_3); |
| (*(tint_private_vars)).local_invocation_index_1 = local_invocation_index_1_param; |
| compute_main_1(tint_private_vars, tint_symbol_3); |
| } |
| |
| kernel void compute_main(uint local_invocation_index_1_param [[thread_index_in_threadgroup]]) { |
| thread tint_private_vars_struct tint_private_vars = {}; |
| threadgroup tint_array<tint_array<tint_array<atomic_uint, 1>, 2>, 3> tint_symbol_4; |
| compute_main_inner_1(local_invocation_index_1_param, &(tint_private_vars), &(tint_symbol_4)); |
| return; |
| } |
| |