0533e7fb38
* subgroup 64 version with subgroup add. 15% faster scalable version tested for subgroup sizes 16-128 * check for subgroup multiple of 16 and greater than 16 * subgroup sizes are always a power of 2 (https://github.com/KhronosGroup/GLSL/issues/45) * force 16 sequential threads per block * make 16 subgroup size a constant |
||
|---|---|---|
| .. | ||
| vulkan-shaders | ||
| CMakeLists.txt | ||
| ggml-vulkan.cpp | ||