#version 450 core

#extension GL_EXT_scalar_block_layout : require

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

layout (std430, set = 0, binding = 0) uniform stride_arguments
{
    int pixelCount;
    int dstStartOffset;
};

layout (std430, set = 1, binding = 1) buffer in_s
{
    uint[] in_data;
};

layout (std430, set = 1, binding = 2) buffer out_s
{
    uint[] out_data;
};

void main()
{
    // Determine what slice of the stride copies this invocation will perform.
    int invocations = int(gl_WorkGroupSize.x * gl_NumWorkGroups.x);

    int copiesRequired = pixelCount;

    // Find the copies that this invocation should perform.
    
    // - Copies that all invocations perform.
    int allInvocationCopies = copiesRequired / invocations;

    // - Extra remainder copy that this invocation performs.
    int index = int(gl_GlobalInvocationID.x);
    int extra = (index < (copiesRequired % invocations)) ? 1 : 0;

    int copyCount = allInvocationCopies + extra;

    // Finally, get the starting offset. Make sure to count extra copies.

    int startCopy = allInvocationCopies * index + min(copiesRequired % invocations, index);

    int srcOffset = startCopy * 2;
    int dstOffset = dstStartOffset + startCopy;

    // Perform the conversion for this region.
    for (int i = 0; i < copyCount; i++)
    {
        float depth = uintBitsToFloat(in_data[srcOffset++]);
        uint stencil = in_data[srcOffset++];

        uint rescaledDepth = uint(clamp(depth, 0.0, 1.0) * 16777215.0);

        out_data[dstOffset++] = (rescaledDepth << 8) | (stencil & 0xff);
    }
}