Public/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-09-19 17:24:57 +02:00
Code Issues Packages Projects Releases Wiki Activity

Fix: default-disable PDL. Enable by setting GGML_CUDA_ENABLE_PDL=1

Browse Source
This commit is contained in:
aendk
2026-05-13 15:27:18 +02:00
parent 98ee686bf7
commit 54483ad5bd
1 changed files with 8 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-cuda/common.cuh
+8 -2
View File
@@ -230,8 +230,14 @@ struct ggml_cuda_pdl_config {
template<typename Kernel, typename... Args>
static __inline__ void ggml_cuda_kernel_launch(Kernel kernel, const ggml_cuda_kernel_launch_params & launch_params, Args&&... args) {
#if defined(GGML_CUDA_USE_PDL)
static const bool disable_pdl = (getenv("GGML_CUDA_PDL") != nullptr);
if (!disable_pdl) {
// PDL is off by default. To enable, set GGML_CUDA_ENABLE_PDL=1
static const bool enable_pdl = []() {
const char * env = getenv("GGML_CUDA_ENABLE_PDL");
return env != nullptr && std::atoi(env) != 0;
}();
if (enable_pdl) {
auto pdl_cfg = ggml_cuda_pdl_config(launch_params);
CUDA_CHECK(cudaLaunchKernelEx(&pdl_cfg.cfg, kernel, std::forward<Args>(args)... ));
return;