test/tint/extensions/subgroup_matrix/construct.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, 1>* tint_member;
  const constant tint_array<uint4, 1>* tint_storage_buffer_sizes;
};

struct tint_array_lengths_struct {
  uint tint_array_length_0_0;
};

struct S {
  simdgroup_float8x8 l;
  simdgroup_float8x8 r;
};

struct S_Nested {
  S s;
};

[[max_total_threads_per_threadgroup(64)]]
kernel void v(device tint_array<float, 1>* v_1 [[buffer(0)]], const constant tint_array<uint4, 1>* tint_storage_buffer_sizes [[buffer(30)]]) {
  tint_module_vars_struct const tint_module_vars = tint_module_vars_struct{.tint_member=v_1, .tint_storage_buffer_sizes=tint_storage_buffer_sizes};
  tint_array_lengths_struct const v_2 = tint_array_lengths_struct{.tint_array_length_0_0=((*tint_module_vars.tint_storage_buffer_sizes)[0u].x / 4u)};
  uint const v_3 = as_type<uint>(0);
  uint const v_4 = as_type<uint>(64);
  bool const v_5 = (((v_3 + (v_4 * 7u)) + 8u) <= v_2.tint_array_length_0_0);
  (simdgroup_store(make_filled_simdgroup_matrix<float, 8, 8>(0.0f), (&(*tint_module_vars.tint_member)[select(0u, v_3, v_5)]), ulong(select(8u, v_4, v_5)), ulong2(0ul), false));
  uint const v_6 = as_type<uint>(0);
  uint const v_7 = as_type<uint>(64);
  bool const v_8 = (((v_6 + (v_7 * 7u)) + 8u) <= v_2.tint_array_length_0_0);
  (simdgroup_store(tint_array<simdgroup_float8x8, 4>{}[1u], (&(*tint_module_vars.tint_member)[select(0u, v_6, v_8)]), ulong(select(8u, v_7, v_8)), ulong2(0ul), false));
  uint const v_9 = as_type<uint>(0);
  uint const v_10 = as_type<uint>(64);
  bool const v_11 = (((v_9 + (v_10 * 7u)) + 8u) <= v_2.tint_array_length_0_0);
  (simdgroup_store(tint_array<tint_array<simdgroup_float8x8, 4>, 4>{}[2u][3u], (&(*tint_module_vars.tint_member)[select(0u, v_9, v_11)]), ulong(select(8u, v_10, v_11)), ulong2(0ul), false));
  uint const v_12 = as_type<uint>(0);
  uint const v_13 = as_type<uint>(64);
  bool const v_14 = (((v_12 + (v_13 * 7u)) + 8u) <= v_2.tint_array_length_0_0);
  (simdgroup_store(S{}.l, (&(*tint_module_vars.tint_member)[select(0u, v_12, v_14)]), ulong(select(8u, v_13, v_14)), ulong2(0ul), false));
  uint const v_15 = as_type<uint>(0);
  uint const v_16 = as_type<uint>(64);
  bool const v_17 = (((v_15 + (v_16 * 7u)) + 8u) <= v_2.tint_array_length_0_0);
  (simdgroup_store(S_Nested{}.s.r, (&(*tint_module_vars.tint_member)[select(0u, v_15, v_17)]), ulong(select(8u, v_16, v_17)), ulong2(0ul), false));
  uint const v_18 = as_type<uint>(0);
  uint const v_19 = as_type<uint>(64);
  bool const v_20 = (((v_18 + (v_19 * 7u)) + 8u) <= v_2.tint_array_length_0_0);
  (simdgroup_store(make_filled_simdgroup_matrix<float, 8, 8>(42.0f), (&(*tint_module_vars.tint_member)[select(0u, v_18, v_20)]), ulong(select(8u, v_19, v_20)), ulong2(0ul), false));
  uint const v_21 = as_type<uint>(0);
  uint const v_22 = as_type<uint>(64);
  bool const v_23 = (((v_21 + (v_22 * 7u)) + 8u) <= v_2.tint_array_length_0_0);
  (simdgroup_store(tint_array<simdgroup_float8x8, 2>{make_filled_simdgroup_matrix<float, 8, 8>(42.0f), make_filled_simdgroup_matrix<float, 8, 8>(100.0f)}[1u], (&(*tint_module_vars.tint_member)[select(0u, v_21, v_23)]), ulong(select(8u, v_22, v_23)), ulong2(0ul), false));
  tint_array<simdgroup_float8x8, 2> const v_24 = tint_array<simdgroup_float8x8, 2>{make_filled_simdgroup_matrix<float, 8, 8>(42.0f), make_filled_simdgroup_matrix<float, 8, 8>(100.0f)};
  uint const v_25 = as_type<uint>(0);
  uint const v_26 = as_type<uint>(64);
  bool const v_27 = (((v_25 + (v_26 * 7u)) + 8u) <= v_2.tint_array_length_0_0);
  (simdgroup_store(tint_array<tint_array<simdgroup_float8x8, 2>, 2>{v_24, tint_array<simdgroup_float8x8, 2>{make_filled_simdgroup_matrix<float, 8, 8>(-7.0f), make_filled_simdgroup_matrix<float, 8, 8>(-42.0f)}}[1u][0u], (&(*tint_module_vars.tint_member)[select(0u, v_25, v_27)]), ulong(select(8u, v_26, v_27)), ulong2(0ul), false));
  uint const v_28 = as_type<uint>(0);
  uint const v_29 = as_type<uint>(64);
  bool const v_30 = (((v_28 + (v_29 * 7u)) + 8u) <= v_2.tint_array_length_0_0);
  (simdgroup_store(S{.l=make_filled_simdgroup_matrix<float, 8, 8>(42.0f), .r=make_filled_simdgroup_matrix<float, 8, 8>(100.0f)}.l, (&(*tint_module_vars.tint_member)[select(0u, v_28, v_30)]), ulong(select(8u, v_29, v_30)), ulong2(0ul), false));
  uint const v_31 = as_type<uint>(0);
  uint const v_32 = as_type<uint>(64);
  bool const v_33 = (((v_31 + (v_32 * 7u)) + 8u) <= v_2.tint_array_length_0_0);
  (simdgroup_store(S_Nested{.s=S{.l=make_filled_simdgroup_matrix<float, 8, 8>(42.0f), .r=make_filled_simdgroup_matrix<float, 8, 8>(100.0f)}}.s.r, (&(*tint_module_vars.tint_member)[select(0u, v_31, v_33)]), ulong(select(8u, v_32, v_33)), ulong2(0ul), false));
}