Skip to content

Conversation

@ronlieb
Copy link
Collaborator

@ronlieb ronlieb commented Dec 12, 2025

…brary (#721)

has fallback to rocm-smi if amd-smi lib not present

gpurun: Process launch utility for GPU applications. This is a wrapper
to execute application binaries including OpenMPI GPU applications.
See help message below (gpurun -h) for more information.

Usage: gpurun [gpurun_options] Program and options
-h --help : display help test
-v : display gpurun command
-vv : display additional debug info
-vvv : display more debug info
-dryrun : do not run bindings
-taskset : use taskset for binding
-numatcl : use numactl for binding [default]
-md : Set number of desired devices for multi-device mode, default=1
-nr : use numactl ROCR_VISIBLE_DEVICES
-nm : use numactl OMPI_COMM_WORLD_LOCAL_RANK
-topo : display the topology and exit
-rocmsmi : force use of rocm-smi rather than amd-smi
--version : Print version of gpurun and exit

Usage Examples:
gpurun true
mpirun -np 4 gpurun env | grep ROCR_VISIBLE_DEVICES

gpurun -topo

GPU[1] : (Topology) Numa Affinity: 6

Topology     Numa: 0   PageSize: [always] madvise never

GPU     Node  Affinity       UUID               Cores
 0        0       0       GPU-b256278bf70405e2    0-23,96-119
 1        1       1       GPU-a33557394e2c744e    24-47,120-143
 2        2       2       GPU-4f78640baf57e5f0    48-71,144-167
 3        3       3       GPU-b66921701d196e10    72-95,168-191

…brary (#721)

has fallback to rocm-smi if amd-smi lib not present

  gpurun: Process launch utility for GPU applications. This is a wrapper
          to execute application binaries including OpenMPI GPU applications.
          See help message below (gpurun -h) for more information.

Usage: gpurun [gpurun_options] Program and options
  -h --help : display help test
  -v        : display gpurun command
  -vv       : display additional debug info
  -vvv      : display more debug info
  -dryrun   : do not run bindings
  -taskset  : use taskset for binding
  -numatcl  : use numactl for binding [default]
  -md       : Set number of desired devices for multi-device mode, default=1
  -nr       : use numactl ROCR_VISIBLE_DEVICES
  -nm       : use numactl OMPI_COMM_WORLD_LOCAL_RANK
  -topo     : display the topology and exit
  -rocmsmi  : force use of rocm-smi rather than amd-smi
  --version : Print version of gpurun and exit

 Usage Examples:
  gpurun true
  mpirun -np  4 gpurun env | grep ROCR_VISIBLE_DEVICES

  gpurun -topo

  GPU[1]          : (Topology) Numa Affinity: 6

    Topology     Numa: 0   PageSize: [always] madvise never

    GPU     Node  Affinity       UUID               Cores
     0        0       0       GPU-b256278bf70405e2    0-23,96-119
     1        1       1       GPU-a33557394e2c744e    24-47,120-143
     2        2       2       GPU-4f78640baf57e5f0    48-71,144-167
     3        3       3       GPU-b66921701d196e10    72-95,168-191
@z1-cciauto
Copy link
Collaborator

@ronlieb
Copy link
Collaborator Author

ronlieb commented Dec 12, 2025

!PSDB

@z1-cciauto
Copy link
Collaborator

@ronlieb ronlieb merged commit 3bb43fc into amd-staging Dec 12, 2025
10 checks passed
@ronlieb ronlieb deleted the amd/dev/rlieberm/gpurunAmdSmi branch December 12, 2025 15:52
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants