#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];
    }
};

template<typename T, int Cols, int Rows=Cols>
struct spvStorageMatrix
{
    vec<T, Rows> columns[Cols];
    
    spvStorageMatrix() thread = default;
    thread spvStorageMatrix& operator=(initializer_list<vec<T, Rows>> cols) thread
    {
        size_t i;
        thread vec<T, Rows>* col;
        for (i = 0, col = cols.begin(); i < Cols; ++i, ++col)
            columns[i] = *col;
        return *this;
    }
    
    spvStorageMatrix(const thread matrix<T, Cols, Rows>& m) thread
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const thread spvStorageMatrix& m) thread = default;
    thread spvStorageMatrix& operator=(const thread matrix<T, Cols, Rows>& m) thread
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    thread spvStorageMatrix& operator=(const thread spvStorageMatrix& m) thread = default;
    
    spvStorageMatrix(const constant matrix<T, Cols, Rows>& m) thread
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const constant spvStorageMatrix& m) thread = default;
    thread spvStorageMatrix& operator=(const constant matrix<T, Cols, Rows>& m) thread
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    thread spvStorageMatrix& operator=(const constant spvStorageMatrix& m) thread = default;
    
    spvStorageMatrix(const device matrix<T, Cols, Rows>& m) thread
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const device spvStorageMatrix& m) thread = default;
    thread spvStorageMatrix& operator=(const device matrix<T, Cols, Rows>& m) thread
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    thread spvStorageMatrix& operator=(const device spvStorageMatrix& m) thread = default;
    
    spvStorageMatrix(const threadgroup matrix<T, Cols, Rows>& m) thread
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const threadgroup spvStorageMatrix& m) thread = default;
    thread spvStorageMatrix& operator=(const threadgroup matrix<T, Cols, Rows>& m) thread
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    thread spvStorageMatrix& operator=(const threadgroup spvStorageMatrix& m) thread = default;
    
    #ifdef __HAVE_IMAGEBLOCKS__
    spvStorageMatrix(const threadgroup_imageblock matrix<T, Cols, Rows>& m) thread
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const threadgroup_imageblock spvStorageMatrix& m) thread = default;
    thread spvStorageMatrix& operator=(const threadgroup_imageblock matrix<T, Cols, Rows>& m) thread
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    thread spvStorageMatrix& operator=(const threadgroup_imageblock spvStorageMatrix& m) thread = default;
    #endif
    
    #ifdef __HAVE_RAYTRACING__
    spvStorageMatrix(const ray_data matrix<T, Cols, Rows>& m) thread
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const ray_data spvStorageMatrix& m) thread = default;
    thread spvStorageMatrix& operator=(const ray_data matrix<T, Cols, Rows>& m) thread
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    thread spvStorageMatrix& operator=(const ray_data spvStorageMatrix& m) thread = default;
    #endif
    
    #ifdef __HAVE_MESH__
    spvStorageMatrix(const object_data matrix<T, Cols, Rows>& m) thread
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const object_data spvStorageMatrix& m) thread = default;
    thread spvStorageMatrix& operator=(const object_data matrix<T, Cols, Rows>& m) thread
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    thread spvStorageMatrix& operator=(const object_data spvStorageMatrix& m) thread = default;
    #endif
    
    operator matrix<T, Cols, Rows>() const thread
    {
        matrix<T, Cols, Rows> m;
        for (int i = 0; i < Cols; ++i)
            m.columns[i] = columns[i];
        return m;
    }
    
    vec<T, Rows> operator[](size_t idx) const thread
    {
        return columns[idx];
    }
    thread vec<T, Rows>& operator[](size_t idx) thread
    {
        return columns[idx];
    }
    
    spvStorageMatrix() constant = default;
    
    spvStorageMatrix(const thread matrix<T, Cols, Rows>& m) constant
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const thread spvStorageMatrix& m) constant = default;
    
    spvStorageMatrix(const constant matrix<T, Cols, Rows>& m) constant
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const constant spvStorageMatrix& m) constant = default;
    
    spvStorageMatrix(const device matrix<T, Cols, Rows>& m) constant
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const device spvStorageMatrix& m) constant = default;
    
    spvStorageMatrix(const threadgroup matrix<T, Cols, Rows>& m) constant
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const threadgroup spvStorageMatrix& m) constant = default;
    
    #ifdef __HAVE_IMAGEBLOCKS__
    spvStorageMatrix(const threadgroup_imageblock matrix<T, Cols, Rows>& m) constant
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const threadgroup_imageblock spvStorageMatrix& m) constant = default;
    #endif
    
    #ifdef __HAVE_RAYTRACING__
    spvStorageMatrix(const ray_data matrix<T, Cols, Rows>& m) constant
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const ray_data spvStorageMatrix& m) constant = default;
    #endif
    
    #ifdef __HAVE_MESH__
    spvStorageMatrix(const object_data matrix<T, Cols, Rows>& m) constant
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const object_data spvStorageMatrix& m) constant = default;
    #endif
    
    operator matrix<T, Cols, Rows>() const constant
    {
        matrix<T, Cols, Rows> m;
        for (int i = 0; i < Cols; ++i)
            m.columns[i] = columns[i];
        return m;
    }
    
    vec<T, Rows> operator[](size_t idx) const constant
    {
        return columns[idx];
    }
    
    spvStorageMatrix() device = default;
    device spvStorageMatrix& operator=(initializer_list<vec<T, Rows>> cols) device
    {
        size_t i;
        thread vec<T, Rows>* col;
        for (i = 0, col = cols.begin(); i < Cols; ++i, ++col)
            columns[i] = *col;
        return *this;
    }
    
    spvStorageMatrix(const thread matrix<T, Cols, Rows>& m) device
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const thread spvStorageMatrix& m) device = default;
    device spvStorageMatrix& operator=(const thread matrix<T, Cols, Rows>& m) device
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    device spvStorageMatrix& operator=(const thread spvStorageMatrix& m) device = default;
    
    spvStorageMatrix(const constant matrix<T, Cols, Rows>& m) device
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const constant spvStorageMatrix& m) device = default;
    device spvStorageMatrix& operator=(const constant matrix<T, Cols, Rows>& m) device
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    device spvStorageMatrix& operator=(const constant spvStorageMatrix& m) device = default;
    
    spvStorageMatrix(const device matrix<T, Cols, Rows>& m) device
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const device spvStorageMatrix& m) device = default;
    device spvStorageMatrix& operator=(const device matrix<T, Cols, Rows>& m) device
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    device spvStorageMatrix& operator=(const device spvStorageMatrix& m) device = default;
    
    spvStorageMatrix(const threadgroup matrix<T, Cols, Rows>& m) device
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const threadgroup spvStorageMatrix& m) device = default;
    device spvStorageMatrix& operator=(const threadgroup matrix<T, Cols, Rows>& m) device
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    device spvStorageMatrix& operator=(const threadgroup spvStorageMatrix& m) device = default;
    
    #ifdef __HAVE_IMAGEBLOCKS__
    spvStorageMatrix(const threadgroup_imageblock matrix<T, Cols, Rows>& m) device
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const threadgroup_imageblock spvStorageMatrix& m) device = default;
    device spvStorageMatrix& operator=(const threadgroup_imageblock matrix<T, Cols, Rows>& m) device
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    device spvStorageMatrix& operator=(const threadgroup_imageblock spvStorageMatrix& m) device = default;
    #endif
    
    #ifdef __HAVE_RAYTRACING__
    spvStorageMatrix(const ray_data matrix<T, Cols, Rows>& m) device
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const ray_data spvStorageMatrix& m) device = default;
    device spvStorageMatrix& operator=(const ray_data matrix<T, Cols, Rows>& m) device
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    device spvStorageMatrix& operator=(const ray_data spvStorageMatrix& m) device = default;
    #endif
    
    #ifdef __HAVE_MESH__
    spvStorageMatrix(const object_data matrix<T, Cols, Rows>& m) device
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const object_data spvStorageMatrix& m) device = default;
    device spvStorageMatrix& operator=(const object_data matrix<T, Cols, Rows>& m) device
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    device spvStorageMatrix& operator=(const object_data spvStorageMatrix& m) device = default;
    #endif
    
    operator matrix<T, Cols, Rows>() const device
    {
        matrix<T, Cols, Rows> m;
        for (int i = 0; i < Cols; ++i)
            m.columns[i] = columns[i];
        return m;
    }
    
    vec<T, Rows> operator[](size_t idx) const device
    {
        return columns[idx];
    }
    device vec<T, Rows>& operator[](size_t idx) device
    {
        return columns[idx];
    }
    
    spvStorageMatrix() threadgroup = default;
    threadgroup spvStorageMatrix& operator=(initializer_list<vec<T, Rows>> cols) threadgroup
    {
        size_t i;
        thread vec<T, Rows>* col;
        for (i = 0, col = cols.begin(); i < Cols; ++i, ++col)
            columns[i] = *col;
        return *this;
    }
    
    spvStorageMatrix(const thread matrix<T, Cols, Rows>& m) threadgroup
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const thread spvStorageMatrix& m) threadgroup = default;
    threadgroup spvStorageMatrix& operator=(const thread matrix<T, Cols, Rows>& m) threadgroup
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    threadgroup spvStorageMatrix& operator=(const thread spvStorageMatrix& m) threadgroup = default;
    
    spvStorageMatrix(const constant matrix<T, Cols, Rows>& m) threadgroup
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const constant spvStorageMatrix& m) threadgroup = default;
    threadgroup spvStorageMatrix& operator=(const constant matrix<T, Cols, Rows>& m) threadgroup
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    threadgroup spvStorageMatrix& operator=(const constant spvStorageMatrix& m) threadgroup = default;
    
    spvStorageMatrix(const device matrix<T, Cols, Rows>& m) threadgroup
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const device spvStorageMatrix& m) threadgroup = default;
    threadgroup spvStorageMatrix& operator=(const device matrix<T, Cols, Rows>& m) threadgroup
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    threadgroup spvStorageMatrix& operator=(const device spvStorageMatrix& m) threadgroup = default;
    
    spvStorageMatrix(const threadgroup matrix<T, Cols, Rows>& m) threadgroup
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const threadgroup spvStorageMatrix& m) threadgroup = default;
    threadgroup spvStorageMatrix& operator=(const threadgroup matrix<T, Cols, Rows>& m) threadgroup
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    threadgroup spvStorageMatrix& operator=(const threadgroup spvStorageMatrix& m) threadgroup = default;
    
    #ifdef __HAVE_IMAGEBLOCKS__
    spvStorageMatrix(const threadgroup_imageblock matrix<T, Cols, Rows>& m) threadgroup
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const threadgroup_imageblock spvStorageMatrix& m) threadgroup = default;
    threadgroup spvStorageMatrix& operator=(const threadgroup_imageblock matrix<T, Cols, Rows>& m) threadgroup
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    threadgroup spvStorageMatrix& operator=(const threadgroup_imageblock spvStorageMatrix& m) threadgroup = default;
    #endif
    
    #ifdef __HAVE_RAYTRACING__
    spvStorageMatrix(const ray_data matrix<T, Cols, Rows>& m) threadgroup
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const ray_data spvStorageMatrix& m) threadgroup = default;
    threadgroup spvStorageMatrix& operator=(const ray_data matrix<T, Cols, Rows>& m) threadgroup
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    threadgroup spvStorageMatrix& operator=(const ray_data spvStorageMatrix& m) threadgroup = default;
    #endif
    
    #ifdef __HAVE_MESH__
    spvStorageMatrix(const object_data matrix<T, Cols, Rows>& m) threadgroup
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const object_data spvStorageMatrix& m) threadgroup = default;
    threadgroup spvStorageMatrix& operator=(const object_data matrix<T, Cols, Rows>& m) threadgroup
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    threadgroup spvStorageMatrix& operator=(const object_data spvStorageMatrix& m) threadgroup = default;
    #endif
    
    operator matrix<T, Cols, Rows>() const threadgroup
    {
        matrix<T, Cols, Rows> m;
        for (int i = 0; i < Cols; ++i)
            m.columns[i] = columns[i];
        return m;
    }
    
    vec<T, Rows> operator[](size_t idx) const threadgroup
    {
        return columns[idx];
    }
    threadgroup vec<T, Rows>& operator[](size_t idx) threadgroup
    {
        return columns[idx];
    }
    
    #ifdef __HAVE_IMAGEBLOCKS__
    spvStorageMatrix() threadgroup_imageblock = default;
    threadgroup_imageblock spvStorageMatrix& operator=(initializer_list<vec<T, Rows>> cols) threadgroup_imageblock
    {
        size_t i;
        thread vec<T, Rows>* col;
        for (i = 0, col = cols.begin(); i < Cols; ++i, ++col)
            columns[i] = *col;
        return *this;
    }
    
    spvStorageMatrix(const thread matrix<T, Cols, Rows>& m) threadgroup_imageblock
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const thread spvStorageMatrix& m) threadgroup_imageblock = default;
    threadgroup_imageblock spvStorageMatrix& operator=(const thread matrix<T, Cols, Rows>& m) threadgroup_imageblock
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    threadgroup_imageblock spvStorageMatrix& operator=(const thread spvStorageMatrix& m) threadgroup_imageblock = default;
    
    spvStorageMatrix(const constant matrix<T, Cols, Rows>& m) threadgroup_imageblock
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const constant spvStorageMatrix& m) threadgroup_imageblock = default;
    threadgroup_imageblock spvStorageMatrix& operator=(const constant matrix<T, Cols, Rows>& m) threadgroup_imageblock
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    threadgroup_imageblock spvStorageMatrix& operator=(const constant spvStorageMatrix& m) threadgroup_imageblock = default;
    
    spvStorageMatrix(const device matrix<T, Cols, Rows>& m) threadgroup_imageblock
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const device spvStorageMatrix& m) threadgroup_imageblock = default;
    threadgroup_imageblock spvStorageMatrix& operator=(const device matrix<T, Cols, Rows>& m) threadgroup_imageblock
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    threadgroup_imageblock spvStorageMatrix& operator=(const device spvStorageMatrix& m) threadgroup_imageblock = default;
    
    spvStorageMatrix(const threadgroup matrix<T, Cols, Rows>& m) threadgroup_imageblock
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const threadgroup spvStorageMatrix& m) threadgroup_imageblock = default;
    threadgroup_imageblock spvStorageMatrix& operator=(const threadgroup matrix<T, Cols, Rows>& m) threadgroup_imageblock
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    threadgroup_imageblock spvStorageMatrix& operator=(const threadgroup spvStorageMatrix& m) threadgroup_imageblock = default;
    
    spvStorageMatrix(const threadgroup_imageblock matrix<T, Cols, Rows>& m) threadgroup_imageblock
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const threadgroup_imageblock spvStorageMatrix& m) threadgroup_imageblock = default;
    threadgroup_imageblock spvStorageMatrix& operator=(const threadgroup_imageblock matrix<T, Cols, Rows>& m) threadgroup_imageblock
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    threadgroup_imageblock spvStorageMatrix& operator=(const threadgroup_imageblock spvStorageMatrix& m) threadgroup_imageblock = default;
    
    #ifdef __HAVE_RAYTRACING__
    spvStorageMatrix(const ray_data matrix<T, Cols, Rows>& m) threadgroup_imageblock
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const ray_data spvStorageMatrix& m) threadgroup_imageblock = default;
    threadgroup_imageblock spvStorageMatrix& operator=(const ray_data matrix<T, Cols, Rows>& m) threadgroup_imageblock
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    threadgroup_imageblock spvStorageMatrix& operator=(const ray_data spvStorageMatrix& m) threadgroup_imageblock = default;
    #endif
    
    #ifdef __HAVE_MESH__
    spvStorageMatrix(const object_data matrix<T, Cols, Rows>& m) threadgroup_imageblock
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const object_data spvStorageMatrix& m) threadgroup_imageblock = default;
    threadgroup_imageblock spvStorageMatrix& operator=(const object_data matrix<T, Cols, Rows>& m) threadgroup_imageblock
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    threadgroup_imageblock spvStorageMatrix& operator=(const object_data spvStorageMatrix& m) threadgroup_imageblock = default;
    #endif
    
    operator matrix<T, Cols, Rows>() const threadgroup_imageblock
    {
        matrix<T, Cols, Rows> m;
        for (int i = 0; i < Cols; ++i)
            m.columns[i] = columns[i];
        return m;
    }
    
    vec<T, Rows> operator[](size_t idx) const threadgroup_imageblock
    {
        return columns[idx];
    }
    threadgroup_imageblock vec<T, Rows>& operator[](size_t idx) threadgroup_imageblock
    {
        return columns[idx];
    }
    #endif
    
    #ifdef __HAVE_RAYTRACING__
    spvStorageMatrix() ray_data = default;
    ray_data spvStorageMatrix& operator=(initializer_list<vec<T, Rows>> cols) ray_data
    {
        size_t i;
        thread vec<T, Rows>* col;
        for (i = 0, col = cols.begin(); i < Cols; ++i, ++col)
            columns[i] = *col;
        return *this;
    }
    
    spvStorageMatrix(const thread matrix<T, Cols, Rows>& m) ray_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const thread spvStorageMatrix& m) ray_data = default;
    ray_data spvStorageMatrix& operator=(const thread matrix<T, Cols, Rows>& m) ray_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    ray_data spvStorageMatrix& operator=(const thread spvStorageMatrix& m) ray_data = default;
    
    spvStorageMatrix(const constant matrix<T, Cols, Rows>& m) ray_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const constant spvStorageMatrix& m) ray_data = default;
    ray_data spvStorageMatrix& operator=(const constant matrix<T, Cols, Rows>& m) ray_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    ray_data spvStorageMatrix& operator=(const constant spvStorageMatrix& m) ray_data = default;
    
    spvStorageMatrix(const device matrix<T, Cols, Rows>& m) ray_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const device spvStorageMatrix& m) ray_data = default;
    ray_data spvStorageMatrix& operator=(const device matrix<T, Cols, Rows>& m) ray_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    ray_data spvStorageMatrix& operator=(const device spvStorageMatrix& m) ray_data = default;
    
    spvStorageMatrix(const threadgroup matrix<T, Cols, Rows>& m) ray_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const threadgroup spvStorageMatrix& m) ray_data = default;
    ray_data spvStorageMatrix& operator=(const threadgroup matrix<T, Cols, Rows>& m) ray_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    ray_data spvStorageMatrix& operator=(const threadgroup spvStorageMatrix& m) ray_data = default;
    
    #ifdef __HAVE_IMAGEBLOCKS__
    spvStorageMatrix(const threadgroup_imageblock matrix<T, Cols, Rows>& m) ray_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const threadgroup_imageblock spvStorageMatrix& m) ray_data = default;
    ray_data spvStorageMatrix& operator=(const threadgroup_imageblock matrix<T, Cols, Rows>& m) ray_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    ray_data spvStorageMatrix& operator=(const threadgroup_imageblock spvStorageMatrix& m) ray_data = default;
    #endif
    
    spvStorageMatrix(const ray_data matrix<T, Cols, Rows>& m) ray_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const ray_data spvStorageMatrix& m) ray_data = default;
    ray_data spvStorageMatrix& operator=(const ray_data matrix<T, Cols, Rows>& m) ray_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    ray_data spvStorageMatrix& operator=(const ray_data spvStorageMatrix& m) ray_data = default;
    
    #ifdef __HAVE_MESH__
    spvStorageMatrix(const object_data matrix<T, Cols, Rows>& m) ray_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const object_data spvStorageMatrix& m) ray_data = default;
    ray_data spvStorageMatrix& operator=(const object_data matrix<T, Cols, Rows>& m) ray_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    ray_data spvStorageMatrix& operator=(const object_data spvStorageMatrix& m) ray_data = default;
    #endif
    
    operator matrix<T, Cols, Rows>() const ray_data
    {
        matrix<T, Cols, Rows> m;
        for (int i = 0; i < Cols; ++i)
            m.columns[i] = columns[i];
        return m;
    }
    
    vec<T, Rows> operator[](size_t idx) const ray_data
    {
        return columns[idx];
    }
    ray_data vec<T, Rows>& operator[](size_t idx) ray_data
    {
        return columns[idx];
    }
    #endif
    
    #ifdef __HAVE_MESH__
    spvStorageMatrix() object_data = default;
    object_data spvStorageMatrix& operator=(initializer_list<vec<T, Rows>> cols) object_data
    {
        size_t i;
        thread vec<T, Rows>* col;
        for (i = 0, col = cols.begin(); i < Cols; ++i, ++col)
            columns[i] = *col;
        return *this;
    }
    
    spvStorageMatrix(const thread matrix<T, Cols, Rows>& m) object_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const thread spvStorageMatrix& m) object_data = default;
    object_data spvStorageMatrix& operator=(const thread matrix<T, Cols, Rows>& m) object_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    object_data spvStorageMatrix& operator=(const thread spvStorageMatrix& m) object_data = default;
    
    spvStorageMatrix(const constant matrix<T, Cols, Rows>& m) object_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const constant spvStorageMatrix& m) object_data = default;
    object_data spvStorageMatrix& operator=(const constant matrix<T, Cols, Rows>& m) object_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    object_data spvStorageMatrix& operator=(const constant spvStorageMatrix& m) object_data = default;
    
    spvStorageMatrix(const device matrix<T, Cols, Rows>& m) object_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const device spvStorageMatrix& m) object_data = default;
    object_data spvStorageMatrix& operator=(const device matrix<T, Cols, Rows>& m) object_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    object_data spvStorageMatrix& operator=(const device spvStorageMatrix& m) object_data = default;
    
    spvStorageMatrix(const threadgroup matrix<T, Cols, Rows>& m) object_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const threadgroup spvStorageMatrix& m) object_data = default;
    object_data spvStorageMatrix& operator=(const threadgroup matrix<T, Cols, Rows>& m) object_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    object_data spvStorageMatrix& operator=(const threadgroup spvStorageMatrix& m) object_data = default;
    
    #ifdef __HAVE_IMAGEBLOCKS__
    spvStorageMatrix(const threadgroup_imageblock matrix<T, Cols, Rows>& m) object_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const threadgroup_imageblock spvStorageMatrix& m) object_data = default;
    object_data spvStorageMatrix& operator=(const threadgroup_imageblock matrix<T, Cols, Rows>& m) object_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    object_data spvStorageMatrix& operator=(const threadgroup_imageblock spvStorageMatrix& m) object_data = default;
    #endif
    
    #ifdef __HAVE_RAYTRACING__
    spvStorageMatrix(const ray_data matrix<T, Cols, Rows>& m) object_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const ray_data spvStorageMatrix& m) object_data = default;
    object_data spvStorageMatrix& operator=(const ray_data matrix<T, Cols, Rows>& m) object_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    object_data spvStorageMatrix& operator=(const ray_data spvStorageMatrix& m) object_data = default;
    #endif
    
    spvStorageMatrix(const object_data matrix<T, Cols, Rows>& m) object_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
    }
    spvStorageMatrix(const object_data spvStorageMatrix& m) object_data = default;
    object_data spvStorageMatrix& operator=(const object_data matrix<T, Cols, Rows>& m) object_data
    {
        for (size_t i = 0; i < Cols; ++i)
            columns[i] = m.columns[i];
        return *this;
    }
    object_data spvStorageMatrix& operator=(const object_data spvStorageMatrix& m) object_data = default;
    
    operator matrix<T, Cols, Rows>() const object_data
    {
        matrix<T, Cols, Rows> m;
        for (int i = 0; i < Cols; ++i)
            m.columns[i] = columns[i];
        return m;
    }
    
    vec<T, Rows> operator[](size_t idx) const object_data
    {
        return columns[idx];
    }
    object_data vec<T, Rows>& operator[](size_t idx) object_data
    {
        return columns[idx];
    }
    #endif
    
};

template<typename T, int Cols, int Rows>
matrix<T, Rows, Cols> transpose(spvStorageMatrix<T, Cols, Rows> m)
{
    return transpose(matrix<T, Cols, Rows>(m));
}

typedef spvStorageMatrix<half, 2, 2> spvStorage_half2x2;
typedef spvStorageMatrix<half, 2, 3> spvStorage_half2x3;
typedef spvStorageMatrix<half, 2, 4> spvStorage_half2x4;
typedef spvStorageMatrix<half, 3, 2> spvStorage_half3x2;
typedef spvStorageMatrix<half, 3, 3> spvStorage_half3x3;
typedef spvStorageMatrix<half, 3, 4> spvStorage_half3x4;
typedef spvStorageMatrix<half, 4, 2> spvStorage_half4x2;
typedef spvStorageMatrix<half, 4, 3> spvStorage_half4x3;
typedef spvStorageMatrix<half, 4, 4> spvStorage_half4x4;
typedef spvStorageMatrix<float, 2, 2> spvStorage_float2x2;
typedef spvStorageMatrix<float, 2, 3> spvStorage_float2x3;
typedef spvStorageMatrix<float, 2, 4> spvStorage_float2x4;
typedef spvStorageMatrix<float, 3, 2> spvStorage_float3x2;
typedef spvStorageMatrix<float, 3, 3> spvStorage_float3x3;
typedef spvStorageMatrix<float, 3, 4> spvStorage_float3x4;
typedef spvStorageMatrix<float, 4, 2> spvStorage_float4x2;
typedef spvStorageMatrix<float, 4, 3> spvStorage_float4x3;
typedef spvStorageMatrix<float, 4, 4> spvStorage_float4x4;

struct sA
{
    spvStorage_float2x3 mA;
};

struct sB
{
    spvStorage_float2x2 mA;
    spvStorage_float3x2 mB;
    uint3 mC;
};

struct sC
{
    sA mA;
    sB mB;
};

struct sD
{
    sC mA;
};

struct sE
{
    spvStorage_float3x2 mA;
    spvStorage_float4x3 mB;
};

struct sF
{
    sE mA;
};

struct sG
{
    sF mA;
};

struct sH
{
    spvUnsafeArray<short3, 2> mA;
};

struct S1
{
    sD a;
    sG b;
    spvUnsafeArray<sH, 2> c;
};

struct block
{
    uint passed;
};

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

kernel void main0(device block& _424 [[buffer(0)]])
{
    threadgroup S1 s1;
    s1.a.mA.mA.mA = spvStorage_float2x3(float2x3(float3(6.0, 8.0, 8.0), float3(0.0, -4.0, -5.0)));
    s1.a.mA.mB.mA = spvStorage_float2x2(float2x2(float2(9.0, -4.0), float2(-6.0, -1.0)));
    s1.a.mA.mB.mB = spvStorage_float3x2(float3x2(float2(-1.0, -2.0), float2(1.0, 6.0), float2(5.0, 7.0)));
    s1.a.mA.mB.mC = uint3(3u, 1u, 5u);
    s1.b.mA.mA.mA = spvStorage_float3x2(float3x2(float2(8.0, 3.0), float2(0.0, 2.0), float2(1.0, 8.0)));
    s1.b.mA.mA.mB = spvStorage_float4x3(float4x3(float3(0.0, 9.0, -1.0), float3(-1.0, -7.0, 7.0), float3(-4.0, -3.0, 1.0), float3(-4.0, -9.0, 1.0)));
    s1.c[0].mA[0] = short3(bool3(true, false, false));
    s1.c[0].mA[1] = short3(bool3(true, false, false));
    s1.c[1].mA[0] = short3(bool3(false));
    s1.c[1].mA[1] = short3(bool3(false));
    threadgroup_barrier(mem_flags::mem_threadgroup);
    threadgroup_barrier(mem_flags::mem_device | mem_flags::mem_threadgroup | mem_flags::mem_texture);
    bool _484 = abs(6.0 - float2x3(s1.a.mA.mA.mA)[0].x) < 0.0500000007450580596923828125;
    bool _469;
    if (_484)
    {
        _469 = abs(8.0 - float2x3(s1.a.mA.mA.mA)[0].y) < 0.0500000007450580596923828125;
    }
    else
    {
        _469 = _484;
    }
    bool _477;
    if (_469)
    {
        _477 = abs(8.0 - float2x3(s1.a.mA.mA.mA)[0].z) < 0.0500000007450580596923828125;
    }
    else
    {
        _477 = _469;
    }
    bool _448;
    if (_477)
    {
        bool _534 = abs(-float2x3(s1.a.mA.mA.mA)[1].x) < 0.0500000007450580596923828125;
        bool _519;
        if (_534)
        {
            _519 = abs((-4.0) - float2x3(s1.a.mA.mA.mA)[1].y) < 0.0500000007450580596923828125;
        }
        else
        {
            _519 = _534;
        }
        bool _527;
        if (_519)
        {
            _527 = abs((-5.0) - float2x3(s1.a.mA.mA.mA)[1].z) < 0.0500000007450580596923828125;
        }
        else
        {
            _527 = _519;
        }
        _448 = _527;
    }
    else
    {
        _448 = _477;
    }
    bool _346;
    if (_448)
    {
        bool _593 = abs(9.0 - float2x2(s1.a.mA.mB.mA)[0].x) < 0.0500000007450580596923828125;
        bool _586;
        if (_593)
        {
            _586 = abs((-4.0) - float2x2(s1.a.mA.mB.mA)[0].y) < 0.0500000007450580596923828125;
        }
        else
        {
            _586 = _593;
        }
        bool _567;
        if (_586)
        {
            bool _626 = abs((-6.0) - float2x2(s1.a.mA.mB.mA)[1].x) < 0.0500000007450580596923828125;
            bool _619;
            if (_626)
            {
                _619 = abs((-1.0) - float2x2(s1.a.mA.mB.mA)[1].y) < 0.0500000007450580596923828125;
            }
            else
            {
                _619 = _626;
            }
            _567 = _619;
        }
        else
        {
            _567 = _586;
        }
        _346 = _567;
    }
    else
    {
        _346 = _448;
    }
    bool _355;
    if (_346)
    {
        bool _688 = abs((-1.0) - float3x2(s1.a.mA.mB.mB)[0].x) < 0.0500000007450580596923828125;
        bool _681;
        if (_688)
        {
            _681 = abs((-2.0) - float3x2(s1.a.mA.mB.mB)[0].y) < 0.0500000007450580596923828125;
        }
        else
        {
            _681 = _688;
        }
        bool _654;
        if (_681)
        {
            bool _721 = abs(1.0 - float3x2(s1.a.mA.mB.mB)[1].x) < 0.0500000007450580596923828125;
            bool _714;
            if (_721)
            {
                _714 = abs(6.0 - float3x2(s1.a.mA.mB.mB)[1].y) < 0.0500000007450580596923828125;
            }
            else
            {
                _714 = _721;
            }
            _654 = _714;
        }
        else
        {
            _654 = _681;
        }
        bool _662;
        if (_654)
        {
            bool _754 = abs(5.0 - float3x2(s1.a.mA.mB.mB)[2].x) < 0.0500000007450580596923828125;
            bool _747;
            if (_754)
            {
                _747 = abs(7.0 - float3x2(s1.a.mA.mB.mB)[2].y) < 0.0500000007450580596923828125;
            }
            else
            {
                _747 = _754;
            }
            _662 = _747;
        }
        else
        {
            _662 = _654;
        }
        _355 = _662;
    }
    else
    {
        _355 = _346;
    }
    bool _364;
    if (_355)
    {
        _364 = all(uint3(3u, 1u, 5u) == s1.a.mA.mB.mC);
    }
    else
    {
        _364 = _355;
    }
    bool _373;
    if (_364)
    {
        bool _822 = abs(8.0 - float3x2(s1.b.mA.mA.mA)[0].x) < 0.0500000007450580596923828125;
        bool _815;
        if (_822)
        {
            _815 = abs(3.0 - float3x2(s1.b.mA.mA.mA)[0].y) < 0.0500000007450580596923828125;
        }
        else
        {
            _815 = _822;
        }
        bool _788;
        if (_815)
        {
            bool _855 = abs(-float3x2(s1.b.mA.mA.mA)[1].x) < 0.0500000007450580596923828125;
            bool _848;
            if (_855)
            {
                _848 = abs(2.0 - float3x2(s1.b.mA.mA.mA)[1].y) < 0.0500000007450580596923828125;
            }
            else
            {
                _848 = _855;
            }
            _788 = _848;
        }
        else
        {
            _788 = _815;
        }
        bool _796;
        if (_788)
        {
            bool _888 = abs(1.0 - float3x2(s1.b.mA.mA.mA)[2].x) < 0.0500000007450580596923828125;
            bool _881;
            if (_888)
            {
                _881 = abs(8.0 - float3x2(s1.b.mA.mA.mA)[2].y) < 0.0500000007450580596923828125;
            }
            else
            {
                _881 = _888;
            }
            _796 = _881;
        }
        else
        {
            _796 = _788;
        }
        _373 = _796;
    }
    else
    {
        _373 = _364;
    }
    bool _382;
    if (_373)
    {
        bool _970 = abs(-float4x3(s1.b.mA.mA.mB)[0].x) < 0.0500000007450580596923828125;
        bool _955;
        if (_970)
        {
            _955 = abs(9.0 - float4x3(s1.b.mA.mA.mB)[0].y) < 0.0500000007450580596923828125;
        }
        else
        {
            _955 = _970;
        }
        bool _963;
        if (_955)
        {
            _963 = abs((-1.0) - float4x3(s1.b.mA.mA.mB)[0].z) < 0.0500000007450580596923828125;
        }
        else
        {
            _963 = _955;
        }
        bool _918;
        if (_963)
        {
            bool _1020 = abs((-1.0) - float4x3(s1.b.mA.mA.mB)[1].x) < 0.0500000007450580596923828125;
            bool _1005;
            if (_1020)
            {
                _1005 = abs((-7.0) - float4x3(s1.b.mA.mA.mB)[1].y) < 0.0500000007450580596923828125;
            }
            else
            {
                _1005 = _1020;
            }
            bool _1013;
            if (_1005)
            {
                _1013 = abs(7.0 - float4x3(s1.b.mA.mA.mB)[1].z) < 0.0500000007450580596923828125;
            }
            else
            {
                _1013 = _1005;
            }
            _918 = _1013;
        }
        else
        {
            _918 = _963;
        }
        bool _926;
        if (_918)
        {
            bool _1070 = abs((-4.0) - float4x3(s1.b.mA.mA.mB)[2].x) < 0.0500000007450580596923828125;
            bool _1055;
            if (_1070)
            {
                _1055 = abs((-3.0) - float4x3(s1.b.mA.mA.mB)[2].y) < 0.0500000007450580596923828125;
            }
            else
            {
                _1055 = _1070;
            }
            bool _1063;
            if (_1055)
            {
                _1063 = abs(1.0 - float4x3(s1.b.mA.mA.mB)[2].z) < 0.0500000007450580596923828125;
            }
            else
            {
                _1063 = _1055;
            }
            _926 = _1063;
        }
        else
        {
            _926 = _918;
        }
        bool _934;
        if (_926)
        {
            bool _1120 = abs((-4.0) - float4x3(s1.b.mA.mA.mB)[3].x) < 0.0500000007450580596923828125;
            bool _1105;
            if (_1120)
            {
                _1105 = abs((-9.0) - float4x3(s1.b.mA.mA.mB)[3].y) < 0.0500000007450580596923828125;
            }
            else
            {
                _1105 = _1120;
            }
            bool _1113;
            if (_1105)
            {
                _1113 = abs(1.0 - float4x3(s1.b.mA.mA.mB)[3].z) < 0.0500000007450580596923828125;
            }
            else
            {
                _1113 = _1105;
            }
            _934 = _1113;
        }
        else
        {
            _934 = _926;
        }
        _382 = _934;
    }
    else
    {
        _382 = _373;
    }
    bool _391;
    if (_382)
    {
        _391 = all(bool3(true, false, false) == bool3(s1.c[0].mA[0]));
    }
    else
    {
        _391 = _382;
    }
    bool _400;
    if (_391)
    {
        _400 = all(bool3(true, false, false) == bool3(s1.c[0].mA[1]));
    }
    else
    {
        _400 = _391;
    }
    bool _409;
    if (_400)
    {
        _409 = all(bool3(false) == bool3(s1.c[1].mA[0]));
    }
    else
    {
        _409 = _400;
    }
    bool _418;
    if (_409)
    {
        _418 = all(bool3(false) == bool3(s1.c[1].mA[1]));
    }
    else
    {
        _418 = _409;
    }
    if (_418)
    {
        _424.passed++;
    }
}