78 lines
2.4 KiB
Plaintext
78 lines
2.4 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];
|
|
};
|
|
|
|
struct S_atomic {
|
|
int x;
|
|
tint_array<atomic_uint, 10> a;
|
|
uint y;
|
|
};
|
|
|
|
struct S {
|
|
int x;
|
|
tint_array<uint, 10> a;
|
|
uint y;
|
|
};
|
|
|
|
void compute_main_inner(uint local_invocation_index_2, threadgroup S_atomic* const tint_symbol) {
|
|
uint idx = 0u;
|
|
(*(tint_symbol)).x = 0;
|
|
(*(tint_symbol)).y = 0u;
|
|
idx = local_invocation_index_2;
|
|
while (true) {
|
|
uint const x_30 = idx;
|
|
if (!((x_30 < 10u))) {
|
|
break;
|
|
}
|
|
uint const x_35 = idx;
|
|
atomic_store_explicit(&((*(tint_symbol)).a[x_35]), 0u, memory_order_relaxed);
|
|
{
|
|
uint const x_41 = idx;
|
|
idx = (x_41 + 1u);
|
|
}
|
|
}
|
|
threadgroup_barrier(mem_flags::mem_threadgroup);
|
|
atomic_store_explicit(&((*(tint_symbol)).a[4]), 1u, memory_order_relaxed);
|
|
return;
|
|
}
|
|
|
|
void compute_main_1(thread uint* const tint_symbol_1, threadgroup S_atomic* const tint_symbol_2) {
|
|
uint const x_53 = *(tint_symbol_1);
|
|
compute_main_inner(x_53, tint_symbol_2);
|
|
return;
|
|
}
|
|
|
|
void compute_main_inner_1(uint local_invocation_index_1_param, threadgroup S_atomic* const tint_symbol_3, thread uint* const tint_symbol_4) {
|
|
{
|
|
(*(tint_symbol_3)).x = 0;
|
|
(*(tint_symbol_3)).y = 0u;
|
|
}
|
|
for(uint idx_1 = local_invocation_index_1_param; (idx_1 < 10u); idx_1 = (idx_1 + 1u)) {
|
|
uint const i = idx_1;
|
|
atomic_store_explicit(&((*(tint_symbol_3)).a[i]), 0u, memory_order_relaxed);
|
|
}
|
|
threadgroup_barrier(mem_flags::mem_threadgroup);
|
|
*(tint_symbol_4) = local_invocation_index_1_param;
|
|
compute_main_1(tint_symbol_4, tint_symbol_3);
|
|
}
|
|
|
|
kernel void compute_main(uint local_invocation_index_1_param [[thread_index_in_threadgroup]]) {
|
|
threadgroup S_atomic tint_symbol_5;
|
|
thread uint tint_symbol_6 = 0u;
|
|
compute_main_inner_1(local_invocation_index_1_param, &(tint_symbol_5), &(tint_symbol_6));
|
|
return;
|
|
}
|
|
|