sdgoij/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-03-17 16:44:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

vulkan: tune MMVQ for Intel Windows (#19988)

Browse Source
This commit is contained in:
Ruben Ortlam
2026-03-02 15:58:25 +01:00
committed by GitHub
parent ec88c3ceea
commit feefb92836
1 changed files with 12 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
ggml/src/ggml-vulkan/ggml-vulkan.cpp
View File

@@ -7574,6 +7574,18 @@ static bool ggml_vk_should_use_mmvq(const vk_device& device, uint32_t m, uint32_
return false;
}
if (device->driver_id == vk::DriverId::eIntelProprietaryWindows) {
// Intel Windows proprietary driver tuning
switch (src0_type) {
case GGML_TYPE_MXFP4:
case GGML_TYPE_Q4_K:
case GGML_TYPE_Q5_K:
return false;
default:
return true;
}
}
switch (src0_type) {
// From tests on A770 Linux, may need more tuning
case GGML_TYPE_Q4_0: