comp-chem/cp2k
1
0
Fork 0
mirror of https://github.com/cp2k/cp2k.git synced 2026-07-28 14:15:19 -04:00
Code Issues Projects Releases Packages Wiki Activity Actions

regtesting: Set HIP_VISIBLE_DEVICES

Browse source
This commit is contained in:
Ole Schütt 2021-07-02 13:25:13 +02:00 • committed by Ole Schütt
parent 22efd666af
commit 5929cfdf3e
1 changed files with 9 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

13
tools/regtesting/do_regtest.py
View file

@ -198,8 +198,11 @@ class Config:
self.debug = args.debug
self.max_errors = args.maxerrors
self.restrictdirs = args.restrictdir if args.restrictdir else [".*"]
cmd = "nvidia-smi --query-gpu=gpu_name --format=csv,noheader | wc -l"
self.num_gpus = int(subprocess.run(cmd, shell=True, capture_output=True).stdout)
nv_cmd = "nvidia-smi --query-gpu=gpu_name --format=csv,noheader | wc -l"
nv_gpus = int(subprocess.run(nv_cmd, shell=True, capture_output=True).stdout)
amd_cmd = "rocm-smi --showid --csv | grep card | wc -l"
amd_gpus = int(subprocess.run(amd_cmd, shell=True, capture_output=True).stdout)
self.num_gpus = nv_gpus + amd_gpus
self.next_gpu = 0 # Used to assign devices round robin to processes.
def launch_exe(
@ -207,10 +210,12 @@ class Config:
) -> Coroutine[Any, Any, Process]:
env = os.environ.copy()
if self.num_gpus > self.mpiranks:
env["CUDA_VISIBLE_DEVICES"] = ""
visible_gpu_devices = []
for _ in range(self.mpiranks): # Utilize all available GPU devices.
self.next_gpu = (self.next_gpu + 1) % self.num_gpus
env["CUDA_VISIBLE_DEVICES"] += f"{self.next_gpu},"
visible_gpu_devices.append(f"{self.next_gpu}")
env["CUDA_VISIBLE_DEVICES"] = ",".join(visible_gpu_devices)
env["HIP_VISIBLE_DEVICES"] = ",".join(visible_gpu_devices)
env["OMP_NUM_THREADS"] = str(self.ompthreads)
exe = self.cp2k_root / "exe" / self.arch / f"{exe_stem}.{self.version}"
cmd = ["mpiexec", f"-np={self.mpiranks}", exe] if self.use_mpi else [exe]