test/tint/builtins/gen/literal/subgroupMatrixLoad/5de410.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_RO {
  /* 0x0000 */ tint_array<half, 1> arg_0;
};

struct tint_module_vars_struct {
  device tint_array<half, 1024>* prevent_dce;
  const device SB_RO* sb_ro;
  const constant tint_array<uint4, 1>* tint_storage_buffer_sizes;
};

simdgroup_half8x8 subgroupMatrixLoad_5de410(tint_module_vars_struct tint_module_vars) {
  simdgroup_half8x8 v = make_filled_simdgroup_matrix<half, 8, 8>(0.0h);
  if ((((1u + (8u * 7u)) + 8u) <= (((*tint_module_vars.tint_storage_buffer_sizes)[0u].x - 0u) / 2u))) {
    simdgroup_half8x8 v_1 = make_filled_simdgroup_matrix<half, 8, 8>(0.0h);
    simdgroup_load(v_1, (&(*tint_module_vars.sb_ro).arg_0[1u]), ulong(8u), ulong2(0ul), true);
    v = v_1;
  }
  simdgroup_half8x8 res = v;
  return res;
}

kernel void compute_main(device tint_array<half, 1024>* prevent_dce [[buffer(0)]], const device SB_RO* sb_ro [[buffer(1)]], const constant tint_array<uint4, 1>* tint_storage_buffer_sizes [[buffer(30)]]) {
  tint_module_vars_struct const tint_module_vars = tint_module_vars_struct{.prevent_dce=prevent_dce, .sb_ro=sb_ro, .tint_storage_buffer_sizes=tint_storage_buffer_sizes};
  simdgroup_store(subgroupMatrixLoad_5de410(tint_module_vars), (&(*tint_module_vars.prevent_dce)[0u]), ulong(64u), ulong2(0ul), false);
}