#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_symbol_5 {
  tint_array<float4x4, 4> w;
};

void f_inner(uint local_invocation_index, threadgroup tint_array<float4x4, 4>* const tint_symbol, const constant tint_array<float4x4, 4>* const tint_symbol_1) {
  for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
    uint const i = idx;
    (*(tint_symbol))[i] = float4x4(float4(0.0f), float4(0.0f), float4(0.0f), float4(0.0f));
  }
  threadgroup_barrier(mem_flags::mem_threadgroup);
  *(tint_symbol) = *(tint_symbol_1);
  (*(tint_symbol))[1] = (*(tint_symbol_1))[2];
  (*(tint_symbol))[1][0] = float4((*(tint_symbol_1))[0][1]).ywxz;
  (*(tint_symbol))[1][0][0] = (*(tint_symbol_1))[0][1][0];
}

kernel void f(const constant tint_array<float4x4, 4>* tint_symbol_4 [[buffer(0)]], threadgroup tint_symbol_5* tint_symbol_3 [[threadgroup(0)]], uint local_invocation_index [[thread_index_in_threadgroup]]) {
  threadgroup tint_array<float4x4, 4>* const tint_symbol_2 = &((*(tint_symbol_3)).w);
  f_inner(local_invocation_index, tint_symbol_2, tint_symbol_4);
  return;
}