#include <metal_stdlib>
#include <simd/simd.h>

using namespace metal;

struct Foo0
{
    float a;
};

struct Foo1
{
    Foo0 a;
};

struct Foo2
{
    Foo1 a;
    float weight;
};

struct Foo0_1
{
    float a;
};

struct Foo1_1
{
    Foo0_1 a;
};

struct Foo2_1
{
    Foo1_1 a;
    float weight;
};

struct SSBO
{
    Foo2_1 outputs[1];
};

constant uint3 gl_WorkGroupSize [[maybe_unused]] = uint3(8u, 8u, 1u);

kernel void main0(device SSBO& _53 [[buffer(0)]], uint gl_LocalInvocationIndex [[thread_index_in_threadgroup]], uint3 gl_WorkGroupID [[threadgroup_position_in_grid]])
{
    threadgroup Foo2 coeffs[64];
    coeffs[gl_LocalInvocationIndex] = Foo2{ Foo1{ Foo0{ 0.0 } }, 0.0 };
    threadgroup_barrier(mem_flags::mem_threadgroup);
    if (gl_LocalInvocationIndex == 0u)
    {
        _53.outputs[gl_WorkGroupID.x].a.a.a = coeffs[0].a.a.a;
        _53.outputs[gl_WorkGroupID.x].weight = coeffs[0].weight;
    }
}