CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-28 08:31:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

CUDA: Changing the CUDA scheduling strategy to spin (#16585)

Browse Source 
* CUDA set scheduling strategy to spinning for cc121

* Using prop.major and prop.minor, include HIP and MUSA

* Exclude HIP and MUSA

* Remove trailing whitespace

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>

* Remove empty line

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>

---------

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
This commit is contained in:
Julius Tischbein
2025-10-15 13:54:15 +02:00
committed by GitHub
parent 554fd578a5
commit 5acd455460
1 changed files with 9 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
ggml/src/ggml-cuda/ggml-cuda.cu
View File

@@ -273,6 +273,15 @@ static ggml_cuda_device_info ggml_cuda_init() {
} else if (device_name.substr(0, 21) == "NVIDIA GeForce GTX 16") { } else if (device_name.substr(0, 21) == "NVIDIA GeForce GTX 16") {
turing_devices_without_mma.push_back({ id, device_name }); turing_devices_without_mma.push_back({ id, device_name });
} }
// Temporary performance fix:
// Setting device scheduling strategy for iGPUs with cc121 to "spinning" to avoid delays in cuda synchronize calls.
// TODO: Check for future drivers the default scheduling strategy and
// remove this call again when cudaDeviceScheduleSpin is default.
if (prop.major == 12 && prop.minor == 1) {
CUDA_CHECK(cudaSetDeviceFlags(cudaDeviceScheduleSpin));
}
#endif // defined(GGML_USE_HIP) #endif // defined(GGML_USE_HIP)
} }