test/tint/builtins/gen/var/subgroupMatrixScalarAdd/a00a6a.wgsl.expected.msl

#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 {
  device tint_array<float, 1024>* prevent_dce;
};

simdgroup_float8x8 subgroupMatrixScalarAdd_a00a6a() {
  simdgroup_float8x8 arg_0 = make_filled_simdgroup_matrix<float, 8, 8>(0.0f);
  float arg_1 = 8.0f;
  simdgroup_float8x8 v = make_filled_simdgroup_matrix<float, 8, 8>(0.0f);
  simdgroup_multiply_accumulate(v, arg_0, simdgroup_float8x8(1.0f), make_filled_simdgroup_matrix<float, 8, 8>(arg_1));
  simdgroup_float8x8 res = v;
  return res;
}

[[max_total_threads_per_threadgroup(1)]]
kernel void compute_main(device tint_array<float, 1024>* prevent_dce [[buffer(0)]]) {
  tint_module_vars_struct const tint_module_vars = tint_module_vars_struct{.prevent_dce=prevent_dce};
  simdgroup_store(subgroupMatrixScalarAdd_a00a6a(), (&(*tint_module_vars.prevent_dce)[0u]), ulong(64u), ulong2(0ul), false);
}