Discussion:
[OMPI users] cuda at Open MPI runtime
Shrader, David Lee
2017-12-21 22:03:46 UTC
Permalink
Hello All,


I'm trying to get a handle on providing Open MPI with cuda capability for my users. I've been through the FAQ items having to do with cuda, but I still have some questions that I hope those with more experience in this area will be able to answer. I'm currently working with Open MPI version 2.1.2.


First, under what circumstances does Open MPI load libcuda.so at runtime? With my non-cuda tests, it seems that Open MPI doesn't touch it. Does it only load it under certain circumstances?


Second, when Open MPI does load libcuda.so, how does it find it? Does it just pull from the environment or is it looking on a specific path?


Third, does Open MPI need anything else beyond libcuda.so? Like anything from the cudatoolkit libraries?


Thank you all in advance for the help!

David

--

HPC Environments

Los Alamos National Lab
Sylvain Jeaugey
2017-12-21 22:14:43 UTC
Permalink
Hi David,


Open MPI will try to load libcuda during MPI_Init when you are using
either the openib BTL or the smcuda BTL (and CUDA support has been
compiled in).


It is looking in /usr/lib64 and in LD_LIBRARY_PATH as well I guess.


You should not need anything else than libcuda for the Open MPI CUDA
support, but very likely, the CUDA application using Open MPI will need
the CUDA toolkit.


Sylvain
Post by Shrader, David Lee
Hello All,
I'm trying to get a handle on providing Open MPI with cuda capability
for my users. I've been through the FAQ items having to do with cuda,
but I still have some questions that I hope those with more experience
in this area will be able to answer. I'm currently working with Open
MPI version 2.1.2.
First, under what circumstances does Open MPI load libcuda.so at
runtime? With my non-cuda tests, it seems that Open MPI doesn't touch
it. Does it only load it under certain circumstances?
Second, when Open MPI does load libcuda.so, how does it find it? Does
it just pull from the environment or is it looking on a specific path?
Third, does Open MPI need anything else beyond libcuda.so? Like
anything from the cudatoolkit libraries?
Thank you all in advance for the help!
David
--
HPC Environments
Los Alamos National Lab
_______________________________________________
users mailing list
https://lists.open-mpi.org/mailman/listinfo/users
-----------------------------------------------------------------------------------
This email message is for the sole use of the intended recipient(s) and may contain
confidential information. Any unauthorized review, use, disclosure or distribution
is prohibited. If you are not the intended recipient, please contact the sender by
reply email and destroy all copies of the original message.
-----------------------------------------------------------------------------------
Loading...