Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

MPI.has_cuda() fails on Spectrum MPI #357

Closed
vchuravy opened this issue Mar 3, 2020 · 5 comments · Fixed by #358
Closed

MPI.has_cuda() fails on Spectrum MPI #357

vchuravy opened this issue Mar 3, 2020 · 5 comments · Fixed by #358
Assignees

Comments

@vchuravy
Copy link
Member

vchuravy commented Mar 3, 2020

using the test script

using MPI
MPI.Init()

@show MPI.has_cuda()

fails with

ERROR: ERROR: ERROR: ERROR: LoadError: LoadError: LoadError: LoadError: ccall: could not find function MPIX_Query_cuda_support in library /opt/ibm/spectrum_mpi/lib/libmpi_ibm.soccall: could not find function MPIX_Query_cuda_support in library
/opt/ibm/spectrum_mpi/lib/libmpi_ibm.soccall: could not find function MPIX_Query_cuda_support in library /opt/ibm/spectrum_mpi/lib/libmpi_ibm.so
Stacktrace:
Stacktrace:
Stacktrace:ccall: could not find function MPIX_Query_cuda_support in library /opt/ibm/spectrum_mpi/lib/libmpi_ibm.so
Stacktrace:
 [1]
@vchuravy vchuravy self-assigned this Mar 3, 2020
@simonbyrne
Copy link
Member

What does MPI.MPI_LIBRARY_VERSION give?

@simonbyrne
Copy link
Member

Do you need to rebuild MPI.jl?

@vchuravy
Copy link
Member Author

vchuravy commented Mar 3, 2020

MPI.MPI_LIBRARY_VERSION = "Open MPI v10.3.0.00rtm0, package: IBM Spectrum MPI, ident: IBM Spectrum MPI, repo rev: IBM_SPECTRUM_MPI_10.03.00.00_2019.04.04_RTM0, Unreleased developer copy\0"

@vchuravy
Copy link
Member Author

vchuravy commented Mar 3, 2020

I just found this entertaining tibit (from https://hpc.llnl.gov/training/tutorials/using-lcs-sierra-system):

Caveat: Do NOT use the MPIX_Query_cuda_support() routine or the preprocessor constant MPIX_CUDA_AWARE_SUPPORT to determine if Spectrum MPI is CUDA-aware. This routine has either been removed from the IBM implementation, or will always return false (older versions).

and this: https://github.com/lammps/lammps/pull/1582/files#diff-3335adca21d9025c15f1c0ba57812a56R43

@simonbyrne
Copy link
Member

Also, what on earth is Open MPI v10.3.0.00rtm0? Their most recent version is 4.0.2

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 a pull request may close this issue.

2 participants