Skip to content

[CUDA] Reduce CPU-side stalls due to the CUDA command buffer being full #19042

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ggerganov merged 4 commits into from
Jan 27, 2026
Merged

[CUDA] Reduce CPU-side stalls due to the CUDA command buffer being full #19042

Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
d3298dc
[CUDA] Reduce CPU-side stalls due to the CUDA command buffer being full
gaugarg-nv Jan 23, 2026
29c73ef
Set the env variable in the CUDA backend registry allocation
gaugarg-nv Jan 23, 2026
14de97e
Add link to PR in code comment
gaugarg-nv Jan 23, 2026
ed2e484
Remove warning logs and update documentation
gaugarg-nv Jan 26, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 17 additions & 0 deletions ggml/src/ggml-cuda/ggml-cuda.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4857,6 +4857,23 @@ ggml_backend_reg_t ggml_backend_cuda_reg() {
static std::mutex mutex;
std::lock_guard<std::mutex> lock(mutex);
if (!initialized) {
// Set CUDA_SCALE_LAUNCH_QUEUES before any CUDA API call to improve multi-GPU pipeline parallelism performance
// PR: https://github.com/ggml-org/llama.cpp/pull/19042
if (getenv("CUDA_SCALE_LAUNCH_QUEUES") == nullptr) {
#ifdef _WIN32
_putenv_s("CUDA_SCALE_LAUNCH_QUEUES", "4x");
#else
setenv("CUDA_SCALE_LAUNCH_QUEUES", "4x", 0); // don't overwrite if already set
#endif

@JohannesGaessler JohannesGaessler Jan 24, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
#endif
#endif // _WIN32


GGML_LOG_WARN("\n");
GGML_LOG_WARN("================================================================================\n");
GGML_LOG_WARN(" CUDA_SCALE_LAUNCH_QUEUES=4x has been enabled\n");
GGML_LOG_WARN(" This environment variable improves performance with multiple GPUs\n");
GGML_LOG_WARN("================================================================================\n");
GGML_LOG_WARN("\n");
}

ggml_backend_cuda_reg_context * ctx = new ggml_backend_cuda_reg_context;
const int min_batch_size = getenv("GGML_OP_OFFLOAD_MIN_BATCH") ? atoi(getenv("GGML_OP_OFFLOAD_MIN_BATCH")) : 32;

Expand Down
Loading