#pragma clang diagnostic ignored "-Wmissing-prototypes"
#pragma clang diagnostic ignored "-Wmissing-braces"

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

using namespace metal;

template<typename T, size_t Num>
struct spvUnsafeArray
{
    T elements[Num ? Num : 1];
    
    thread T& operator [] (size_t pos) thread
    {
        return elements[pos];
    }
    constexpr const thread T& operator [] (size_t pos) const thread
    {
        return elements[pos];
    }
    
    device T& operator [] (size_t pos) device
    {
        return elements[pos];
    }
    constexpr const device T& operator [] (size_t pos) const device
    {
        return elements[pos];
    }
    
    constexpr const constant T& operator [] (size_t pos) const constant
    {
        return elements[pos];
    }
    
    threadgroup T& operator [] (size_t pos) threadgroup
    {
        return elements[pos];
    }
    constexpr const threadgroup T& operator [] (size_t pos) const threadgroup
    {
        return elements[pos];
    }
};

struct DrawMonoid
{
    uint path_ix;
    uint clip_ix;
};

struct DrawMonoid_1
{
    uint path_ix;
    uint clip_ix;
};

struct DataBuf
{
    DrawMonoid_1 data[1];
};

constant uint3 gl_WorkGroupSize [[maybe_unused]] = uint3(512u, 1u, 1u);

static inline __attribute__((always_inline))
DrawMonoid combine_tag_monoid(thread const DrawMonoid& a, thread const DrawMonoid& b)
{
    DrawMonoid c;
    c.path_ix = a.path_ix + b.path_ix;
    c.clip_ix = a.clip_ix + b.clip_ix;
    return c;
}

static inline __attribute__((always_inline))
DrawMonoid tag_monoid_identity()
{
    return DrawMonoid{ 0u, 0u };
}

kernel void main0(device DataBuf& _57 [[buffer(0)]], uint3 gl_GlobalInvocationID [[thread_position_in_grid]], uint3 gl_LocalInvocationID [[thread_position_in_threadgroup]])
{
    threadgroup DrawMonoid sh_scratch[512];
    uint ix = gl_GlobalInvocationID.x * 8u;
    spvUnsafeArray<DrawMonoid, 8> local;
    local[0].path_ix = _57.data[ix].path_ix;
    local[0].clip_ix = _57.data[ix].clip_ix;
    DrawMonoid param_1;
    for (uint i = 1u; i < 8u; i++)
    {
        uint _82 = ix + i;
        DrawMonoid param = local[i - 1u];
        param_1.path_ix = _57.data[_82].path_ix;
        param_1.clip_ix = _57.data[_82].clip_ix;
        local[i] = combine_tag_monoid(param, param_1);
    }
    DrawMonoid agg = local[7];
    sh_scratch[gl_LocalInvocationID.x] = agg;
    for (uint i_1 = 0u; i_1 < 9u; i_1++)
    {
        threadgroup_barrier(mem_flags::mem_threadgroup);
        if (gl_LocalInvocationID.x >= (1u << i_1))
        {
            DrawMonoid other = sh_scratch[gl_LocalInvocationID.x - (1u << i_1)];
            DrawMonoid param_2 = other;
            DrawMonoid param_3 = agg;
            agg = combine_tag_monoid(param_2, param_3);
        }
        threadgroup_barrier(mem_flags::mem_threadgroup);
        sh_scratch[gl_LocalInvocationID.x] = agg;
    }
    threadgroup_barrier(mem_flags::mem_threadgroup);
    DrawMonoid row = tag_monoid_identity();
    if (gl_LocalInvocationID.x > 0u)
    {
        row = sh_scratch[gl_LocalInvocationID.x - 1u];
    }
    for (uint i_2 = 0u; i_2 < 8u; i_2++)
    {
        DrawMonoid param_4 = row;
        DrawMonoid param_5 = local[i_2];
        DrawMonoid m = combine_tag_monoid(param_4, param_5);
        uint _178 = ix + i_2;
        _57.data[_178].path_ix = m.path_ix;
        _57.data[_178].clip_ix = m.clip_ix;
    }
}