yum-mirror/glslang

git clone https://git.yummers.dev/yum-mirror/glslang

Jeff BolzImplement GL_NV_cooperative_vector1b65bd60

main
branchesmainall refs →
1.1 KiB30 linesraw
1#version 450 core
2#extension GL_KHR_memory_scope_semantics : enable
3#extension GL_NV_cooperative_vector : enable
4#extension GL_EXT_shader_explicit_arithmetic_types : enable
5#extension GL_EXT_buffer_reference : enable
6#extension GL_EXT_nonuniform_qualifier : enable
7
8layout (local_size_x = 64, local_size_y = 1, local_size_z = 1) in;
9
10layout(set = 0, binding = 0) buffer MatrixBuf {
11    float16_t matrixData[];
12} matrixBuf;
13
14layout(set = 0, binding = 0) buffer MatrixBuf2 {
15    int8_t matrixData[];
16} matrixBuf2;
17
18layout(constant_id = 0) const int matrixLayout = gl_CooperativeVectorMatrixLayoutInferencingOptimalNV;
19
20void main()
21{
22    coopvecNV<float16_t, 8> v0;
23    coopvecNV<float16_t, 5> v1;
24
25    coopVecOuterProductAccumulateNV(v0, v1, matrixBuf.matrixData, 100, 128, matrixLayout, gl_ComponentTypeFloat16NV);
26    coopVecOuterProductAccumulateNV(v0, v1, matrixBuf2.matrixData, 100, 128, gl_CooperativeVectorMatrixLayoutInferencingOptimalNV, gl_ComponentTypeFloat16NV);
27
28    coopVecReduceSumAccumulateNV(v0, matrixBuf.matrixData, 100);
29    coopVecReduceSumAccumulateNV(v0, matrixBuf2.matrixData, 100);
30}