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

using namespace metal;

struct SSBO
{
    float out_data[1];
};

constant uint3 gl_WorkGroupSize [[maybe_unused]] = uint3(4u, 4u, 1u);

kernel void main0(device SSBO& _67 [[buffer(0)]], uint3 gl_LocalInvocationID [[thread_position_in_threadgroup]], uint gl_LocalInvocationIndex [[thread_index_in_threadgroup]], uint3 gl_GlobalInvocationID [[thread_position_in_grid]])
{
    threadgroup float foo[4][4];
    foo[gl_LocalInvocationID.x][gl_LocalInvocationID.y] = float(gl_LocalInvocationIndex);
    threadgroup_barrier(mem_flags::mem_threadgroup);
    _67.out_data[gl_GlobalInvocationID.x] = ((foo[gl_LocalInvocationID.x][0] + foo[gl_LocalInvocationID.x][1]) + foo[gl_LocalInvocationID.x][2]) + foo[gl_LocalInvocationID.x][3];
}