test/tint/builtins/gen/literal/subgroupMatrixLoad/128bf4.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 SB_RW {
  /* 0x0000 */ tint_array<half, 1> arg_0;
};

struct tint_module_vars_struct {
  device tint_array<half, 1024>* prevent_dce;
  device SB_RW* sb_rw;
};

simdgroup_half8x8 subgroupMatrixLoad_128bf4(tint_module_vars_struct tint_module_vars) {
  simdgroup_half8x8 v = make_filled_simdgroup_matrix<half, 8, 8>(0.0h);
  simdgroup_load(v, (&(*tint_module_vars.sb_rw).arg_0[1u]), ulong(1u), ulong2(0ul), true);
  simdgroup_half8x8 res = v;
  return res;
}

kernel void compute_main(device tint_array<half, 1024>* prevent_dce [[buffer(0)]], device SB_RW* sb_rw [[buffer(1)]]) {
  tint_module_vars_struct const tint_module_vars = tint_module_vars_struct{.prevent_dce=prevent_dce, .sb_rw=sb_rw};
  simdgroup_store(subgroupMatrixLoad_128bf4(tint_module_vars), (&(*tint_module_vars.prevent_dce)[0u]), ulong(64u), ulong2(0ul), false);
}