|  | #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 { | 
|  | threadgroup tint_array<tint_array<tint_array<atomic_uint, 1>, 2>, 3>* wg; | 
|  | }; | 
|  |  | 
|  | struct tint_symbol_1 { | 
|  | tint_array<tint_array<tint_array<atomic_uint, 1>, 2>, 3> tint_symbol; | 
|  | }; | 
|  |  | 
|  | void compute_main_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 >= 6u)) { | 
|  | break; | 
|  | } | 
|  | atomic_store_explicit((&(*tint_module_vars.wg)[(v_1 / 2u)][(v_1 % 2u)][0u]), 0u, memory_order_relaxed); | 
|  | { | 
|  | v = (v_1 + 1u); | 
|  | } | 
|  | continue; | 
|  | } | 
|  | } | 
|  | threadgroup_barrier(mem_flags::mem_threadgroup); | 
|  | atomic_store_explicit((&(*tint_module_vars.wg)[2u][1u][0u]), 1u, memory_order_relaxed); | 
|  | } | 
|  |  | 
|  | kernel void compute_main(uint tint_local_index [[thread_index_in_threadgroup]], threadgroup tint_symbol_1* v_2 [[threadgroup(0)]]) { | 
|  | tint_module_vars_struct const tint_module_vars = tint_module_vars_struct{.wg=(&(*v_2).tint_symbol)}; | 
|  | compute_main_inner(tint_local_index, tint_module_vars); | 
|  | } |