code_public/lidar_rendu
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Support RTX 5060 Ti: CuPy 13.4 JIT compiles kernels at runtime for sm_120

Browse Source
This commit is contained in:
Antoine Jacquin
2026-05-31 22:01:08 +02:00
parent 17fd96fdf5
commit 04eac7eded
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
lidar_pipeline/gpu.py
View File

@ -61,8 +61,8 @@ def _pick_gpu() -> int | None:
_NUM_GPUS = len(gpus) _NUM_GPUS = len(gpus)
gpus.sort(key=lambda g: g[4], reverse=True) gpus.sort(key=lambda g: g[4], reverse=True)
# CuPy 13.4 + CUDA 11.8 JIT supports up to sm_89 (RTX 40xx). # CuPy 13.4 + CUDA 11.8 JIT compiles kernels at runtime for any
# sm_120 (RTX 50xx) is not supported yet — skip it. # architecture (sm_89, sm_120, etc.). All NVIDIA GPUs are usable.
for gpu in gpus: for gpu in gpus:
idx, name, cap_str, mem_mi, score, major = gpu idx, name, cap_str, mem_mi, score, major = gpu
if major <= 8: # sm_89 and below — works with CuPy 13.4 + JIT if major <= 8: # sm_89 and below — works with CuPy 13.4 + JIT