apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 00:17:25 +08:00
Code Issues Actions 19 Packages Projects Releases Wiki Activity

cuda13.0, implement changes to CCCL (#6751)

Browse Source
This commit is contained in:
wangyifei
2026-03-10 16:47:02 +08:00
committed by GitHub
parent 54581b8653
commit b57c960837
13 changed files with 211 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files
custom_ops/setup_ops.py
+6
View File
@@ -368,6 +368,12 @@ elif paddle.is_compiled_with_cuda():
nvcc_version = get_nvcc_version()
print(f"nvcc_version = {nvcc_version}")
# CUDA 13.0+ (CCCL 3.0) changes the default -static-global-template-stub behavior
# Restore old linking behavior to allow kernel symbols to be visible in shared libraries
if nvcc_version >= 13.0:
nvcc_compile_args += ["-static-global-template-stub=false"]
if nvcc_version >= 12.0:
sources += ["gpu_ops/sample_kernels/air_top_p_sampling.cu"]
cc = max(get_sm_version(archs))