#version 450 core
#extension GL_KHR_memory_scope_semantics : enable
#extension GL_KHR_cooperative_matrix : enable
#extension GL_NV_cooperative_vector : enable
#extension GL_EXT_shader_explicit_arithmetic_types : enable
#extension GL_EXT_buffer_reference : enable
#extension GL_EXT_shader_64bit_indexing : enable

#pragma shader_64bit_indexing

layout (local_size_x = 64, local_size_y = 1, local_size_z = 1) in;

layout(set = 0, binding = 0, buffer_reference) coherent buffer Block {
    float x[];
} block;

void main()
{
    uint64_t element = 0;
    coopmat<float16_t, gl_ScopeSubgroup, 16, 16, gl_MatrixUseA> m;
    coopMatLoad(m, block.x, element, 128, gl_CooperativeMatrixLayoutRowMajor);
    coopMatStore(m, block.x, element, 128, gl_CooperativeMatrixLayoutRowMajor);

    coopvecNV<float16_t, 2> vec;
    coopVecMatMulAddNV(vec, vec, 0, block.x, element, 0, block.x, element, 0, 2, 2, 0, false, 0);
    coopVecMatMulNV(vec, vec, 0, block.x, element, 0, 2, 2, 0, false, 0);
    coopVecLoadNV(vec, block.x, element);
    coopVecStoreNV(vec, block.x, element);
    coopVecOuterProductAccumulateNV(vec, vec, block.x, element, 0, 0, 0);
    coopVecReduceSumAccumulateNV(vec, block.x, element);

    float y = block.x[element];
    int32_t len = block.x.length();
    int64_t len64 = block.x.length64();
}