75 lines
2.8 KiB
Plaintext
75 lines
2.8 KiB
Plaintext
#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];
|
|
};
|
|
|
|
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_3) {
|
|
uint idx = 0u;
|
|
idx = local_invocation_index_2;
|
|
while (true) {
|
|
uint const x_25 = idx;
|
|
if (!((x_25 < 6u))) {
|
|
break;
|
|
}
|
|
uint const x_31 = idx;
|
|
uint const x_33 = idx;
|
|
uint const x_35 = idx;
|
|
uint const tint_symbol = tint_div(x_31, 2u);
|
|
uint const tint_symbol_1 = tint_mod(x_33, 2u);
|
|
uint const tint_symbol_2 = tint_mod(x_35, 1u);
|
|
atomic_store_explicit(&((*(tint_symbol_3))[tint_symbol][tint_symbol_1][tint_symbol_2]), 0u, memory_order_relaxed);
|
|
{
|
|
uint const x_42 = idx;
|
|
idx = (x_42 + 1u);
|
|
}
|
|
}
|
|
threadgroup_barrier(mem_flags::mem_threadgroup);
|
|
atomic_store_explicit(&((*(tint_symbol_3))[2][1][0]), 1u, memory_order_relaxed);
|
|
return;
|
|
}
|
|
|
|
void compute_main_1(thread uint* const tint_symbol_4, threadgroup tint_array<tint_array<tint_array<atomic_uint, 1>, 2>, 3>* const tint_symbol_5) {
|
|
uint const x_57 = *(tint_symbol_4);
|
|
compute_main_inner(x_57, tint_symbol_5);
|
|
return;
|
|
}
|
|
|
|
void compute_main_inner_1(uint local_invocation_index_1_param, threadgroup tint_array<tint_array<tint_array<atomic_uint, 1>, 2>, 3>* const tint_symbol_6, thread uint* const tint_symbol_7) {
|
|
for(uint idx_1 = local_invocation_index_1_param; (idx_1 < 6u); idx_1 = (idx_1 + 1u)) {
|
|
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_6))[i][i_1][i_2]), 0u, memory_order_relaxed);
|
|
}
|
|
threadgroup_barrier(mem_flags::mem_threadgroup);
|
|
*(tint_symbol_7) = local_invocation_index_1_param;
|
|
compute_main_1(tint_symbol_7, tint_symbol_6);
|
|
}
|
|
|
|
kernel void compute_main(uint local_invocation_index_1_param [[thread_index_in_threadgroup]]) {
|
|
threadgroup tint_array<tint_array<tint_array<atomic_uint, 1>, 2>, 3> tint_symbol_8;
|
|
thread uint tint_symbol_9 = 0u;
|
|
compute_main_inner_1(local_invocation_index_1_param, &(tint_symbol_8), &(tint_symbol_9));
|
|
return;
|
|
}
|
|
|