happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

vulkan: disable mmvq on Intel Windows driver (#20672) 

* vulkan: disable mmvq on Intel Windows driver

* improve comment
This commit is contained in:
Ruben Ortlam 2026-03-17 21:51:43 +01:00 committed by GitHub
parent ee4801e5a6
commit 892e3c333a
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 5 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
ggml/src/ggml-vulkan/ggml-vulkan.cpp
View File

@ -7646,20 +7646,14 @@ static bool ggml_vk_should_use_mmvq(const vk_device& device, uint32_t m, uint32_
return true;
}
case VK_VENDOR_ID_INTEL:
if (k < 2048) {
if (device->driver_id == vk::DriverId::eIntelProprietaryWindows) {
// Intel Windows proprietary driver MMVQ performance is worse than fp16, see
// https://github.com/ggml-org/llama.cpp/issues/17628
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;
}
if (k < 2048) {
return false;
}
switch (src0_type) {