test/tint/extensions/subgroup_matrix/runtime_array_workgroup.wgsl.expected.msl

#include <metal_stdlib>
using namespace metal;

struct tint_module_vars_struct {
  threadgroup array<uchar, 1024>* v;
};
typedef array<float, 1> __attribute__((__may_alias__)) tint_aliased_array_f32;

struct tint_symbol_1 {
  array<uchar, 1024> tint_symbol;
};

void main_inner(uint tint_local_index, tint_module_vars_struct tint_module_vars) {
  {
    uint idx = tint_local_index;
    while(true) {
      uint const v_1 = idx;
      if ((v_1 >= 1024u)) {
        break;
      }
      (*tint_module_vars.v)[v_1] = 0u;
      {
        idx = (idx + 32u);
      }
    }
  }
  (threadgroup_barrier(mem_flags::mem_threadgroup));
  threadgroup tint_aliased_array_f32* const v_2 = reinterpret_cast<threadgroup array<float, 1>*>(reinterpret_cast<threadgroup char*>(tint_module_vars.v) + 0u);
  threadgroup tint_aliased_array_f32* const view = v_2;
  bool const v_3 = (64u <= 256u);
  simdgroup_float8x8 v_4 = make_filled_simdgroup_matrix<float, 8, 8>(0.0f);
  (simdgroup_load(v_4, (&(*view)[select(0u, 0u, v_3)]), ulong(select(8u, 8u, v_3)), ulong2(0ul), true));
  simdgroup_float8x8 const m = v_4;
  bool const v_5 = (64u <= 256u);
  (simdgroup_store(m, (&(*view)[select(0u, 0u, v_5)]), ulong(select(8u, 8u, v_5)), ulong2(0ul), false));
}

[[max_total_threads_per_threadgroup(32)]]
kernel void v_6(uint tint_local_index [[thread_index_in_threadgroup]], threadgroup tint_symbol_1* v_7 [[threadgroup(0)]]) {
  tint_module_vars_struct const tint_module_vars = tint_module_vars_struct{.v=(&(*v_7).tint_symbol)};
  (main_inner(tint_local_index, tint_module_vars));
}