test/tint/extensions/subgroup_matrix/function_param.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 S {
  simdgroup_float8x8 l;
  simdgroup_float8x8 r;
};

struct S_Nested {
  S s;
};

struct tint_array_lengths_struct {
  uint tint_array_length_0_0;
};

struct tint_module_vars_struct {
  device tint_array<float, 1>* tint_member;
  const constant tint_array<uint4, 1>* tint_storage_buffer_sizes;
};

void foo(simdgroup_float8x8 m, tint_array<simdgroup_float8x8, 4> m_array, tint_array<tint_array<simdgroup_float8x8, 4>, 4> m_nested_array, S m_struct, S_Nested m_nested_struct, tint_array_lengths_struct tint_array_lengths, tint_module_vars_struct tint_module_vars) {
  uint const v = as_type<uint>(0);
  uint const v_1 = as_type<uint>(64);
  bool const v_2 = (addsat(madsat(v_1, 7u, v), 8u) <= tint_array_lengths.tint_array_length_0_0);
  (simdgroup_store(m, (&(*tint_module_vars.tint_member)[select(0u, v, v_2)]), ulong(select(8u, v_1, v_2)), ulong2(0ul), false));
  uint const v_3 = as_type<uint>(0);
  uint const v_4 = as_type<uint>(64);
  bool const v_5 = (addsat(madsat(v_4, 7u, v_3), 8u) <= tint_array_lengths.tint_array_length_0_0);
  (simdgroup_store(m_array[0u], (&(*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 = (addsat(madsat(v_7, 7u, v_6), 8u) <= tint_array_lengths.tint_array_length_0_0);
  (simdgroup_store(m_nested_array[1u][2u], (&(*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 = (addsat(madsat(v_10, 7u, v_9), 8u) <= tint_array_lengths.tint_array_length_0_0);
  (simdgroup_store(m_struct.l, (&(*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 = (addsat(madsat(v_13, 7u, v_12), 8u) <= tint_array_lengths.tint_array_length_0_0);
  (simdgroup_store(m_nested_struct.s.r, (&(*tint_module_vars.tint_member)[select(0u, v_12, v_14)]), ulong(select(8u, v_13, v_14)), ulong2(0ul), false));
}

[[max_total_threads_per_threadgroup(64)]]
kernel void v_15(device tint_array<float, 1>* v_16 [[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_16, .tint_storage_buffer_sizes=tint_storage_buffer_sizes};
  tint_array_lengths_struct const v_17 = tint_array_lengths_struct{.tint_array_length_0_0=((*tint_module_vars.tint_storage_buffer_sizes)[0u].x / 4u)};
  simdgroup_float8x8 m = make_filled_simdgroup_matrix<float, 8, 8>(0.0f);
  tint_array<simdgroup_float8x8, 4> m_array = {};
  tint_array<tint_array<simdgroup_float8x8, 4>, 4> m_nested_array = {};
  S m_struct = {};
  S_Nested m_nested_struct = {};
  (foo(m, m_array, m_nested_array, m_struct, m_nested_struct, v_17, tint_module_vars));
}