#version 450 core
#extension GL_KHR_memory_scope_semantics : enable
#extension GL_EXT_long_vector : enable
#extension GL_EXT_shader_explicit_arithmetic_types : enable
#extension GL_EXT_buffer_reference : enable
#extension GL_EXT_nonuniform_qualifier : enable

layout (local_size_x = 64, local_size_y = 1, local_size_z = 1) in;

layout(set = 0, binding = 0) buffer MatrixBuf {
    float16_t matrixData[];
} matrixBuf;

layout(set = 0, binding = 0) buffer MatrixBuf2 {
    int8_t matrixData[];
} matrixBuf2;

layout(constant_id = 0) const uint32_t c0 = 0;

void main()
{
    vector<float, c0> v = vector<float, c0>(0.0);

    vec3 x = vec3(1.0), y = vec3(2.0);
    vector<float, c0> v2 = vector<float, c0>(x[0]);

    const vec3 x2 = vec3(1.0), y2 = vec3(2.0);
    vector<float, c0> v3 = vector<float, c0>(x2[0]);

    vec3 x3 = vec3(4, 5, 6), y3 = vec3(7, 8, 9);

    vector<float, c0> v4 = vector<float, c0>(x3.z);

    vector<float16_t, c0> v5 = vector<float16_t, c0>(v4);

    vector<float, c0> v6;

    v6 = v + v2;
    v6 = v - v2;
    v6 = v * v2;
    v6 = v / v2;

    v6 += v;
    v6 -= v;
    v6 *= v;
    v6 /= v;

    float f;
    v6 *= f;
    v6 *= 5.0;

    int len = v6.length();

    v6[0] = f;
    v6[4] = 5.0;

    f += (v += v)[4];
    v6 = v;

    vector<float, 20> v7;
    vector<float16_t, c0> f162 = vector<float16_t, c0>(v7.length());

    float16_t f16;
    vector<float16_t, c0> f161 = vector<float16_t, c0>(f16);

    vector<float16_t, c0> v11, v12, v13;

    v11 = max(v11, v12);
    v11 = min(v11, v12);
    v11 = step(v11, v12);
    v11 = clamp(v11, v12, v13);
    v11 = exp(v11);
    v11 = log(v11);
    v11 = tanh(v11);
    v11 = atan(v11);
    v11 = fma(v11, v12, v13);

    f += (v += v)[len];
}

vector<float16_t, c0> foo(vector<float16_t, c0> x) { return x; }