ConvertD32S8ToD24S8ShaderSource.comp
1 #version 450 core 2 3 #extension GL_EXT_scalar_block_layout : require 4 5 layout (local_size_x = 64, local_size_y = 1, local_size_z = 1) in; 6 7 layout (std430, set = 0, binding = 0) uniform stride_arguments 8 { 9 int pixelCount; 10 int dstStartOffset; 11 }; 12 13 layout (std430, set = 1, binding = 1) buffer in_s 14 { 15 uint[] in_data; 16 }; 17 18 layout (std430, set = 1, binding = 2) buffer out_s 19 { 20 uint[] out_data; 21 }; 22 23 void main() 24 { 25 // Determine what slice of the stride copies this invocation will perform. 26 int invocations = int(gl_WorkGroupSize.x * gl_NumWorkGroups.x); 27 28 int copiesRequired = pixelCount; 29 30 // Find the copies that this invocation should perform. 31 32 // - Copies that all invocations perform. 33 int allInvocationCopies = copiesRequired / invocations; 34 35 // - Extra remainder copy that this invocation performs. 36 int index = int(gl_GlobalInvocationID.x); 37 int extra = (index < (copiesRequired % invocations)) ? 1 : 0; 38 39 int copyCount = allInvocationCopies + extra; 40 41 // Finally, get the starting offset. Make sure to count extra copies. 42 43 int startCopy = allInvocationCopies * index + min(copiesRequired % invocations, index); 44 45 int srcOffset = startCopy * 2; 46 int dstOffset = dstStartOffset + startCopy; 47 48 // Perform the conversion for this region. 49 for (int i = 0; i < copyCount; i++) 50 { 51 float depth = uintBitsToFloat(in_data[srcOffset++]); 52 uint stencil = in_data[srcOffset++]; 53 54 uint rescaledDepth = uint(clamp(depth, 0.0, 1.0) * 16777215.0); 55 56 out_data[dstOffset++] = (rescaledDepth << 8) | (stencil & 0xff); 57 } 58 }