CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-12 10:47:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

vulkan: fix mmv subgroup16 selection (#15775)

Browse Source
This commit is contained in:
Ruben Ortlam
2025-09-03 22:55:10 +02:00
committed by GitHub
parent 0fce7a1248
commit dff7551bfd
1 changed files with 1 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
ggml/src/ggml-vulkan/ggml-vulkan.cpp
View File

@@ -2937,9 +2937,7 @@ static void ggml_vk_load_shaders(vk_device& device) {
const bool use_subgroups = device->subgroup_arithmetic && device->architecture != vk_device_architecture::AMD_GCN; const bool use_subgroups = device->subgroup_arithmetic && device->architecture != vk_device_architecture::AMD_GCN;
// Ensure a subgroup size >= 16 is available // Ensure a subgroup size >= 16 is available
const bool use_subgroups16 = use_subgroups && const bool use_subgroups16 = use_subgroups && subgroup_min_size_16;
(!device->subgroup_size_control && device->subgroup_size >= 16 ||
device->subgroup_size_control && device->subgroup_min_size <= 16 && device->subgroup_max_size >= 16);
const uint32_t subgroup_size = (device->vendor_id == VK_VENDOR_ID_INTEL && device->subgroup_size_control && device->subgroup_min_size <= 16 && device->subgroup_max_size >= 16) ? 16 : device->subgroup_size; const uint32_t subgroup_size = (device->vendor_id == VK_VENDOR_ID_INTEL && device->subgroup_size_control && device->subgroup_min_size <= 16 && device->subgroup_max_size >= 16) ? 16 : device->subgroup_size;
const uint32_t subgroup_size16 = std::max(subgroup_size, 16u); const uint32_t subgroup_size16 = std::max(subgroup_size, 16u);