Fix to issue 235. If max_gpuprocs is specified, use that value even … #236
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…f it results in a crash. If not, attempt to find the number of GPUs and set max_gpuprocs equal to that instead of the number of ranks. When using MPI, cupy.cuda.runtime.getDeviceCount() will not work if gpu-bind is used so we first attempt to check /proc and if that fails we gather the device PCI bus ids to count devices and issue a warning because this is slow (~2s for 64 ranks). Looking at /proc should work on all Linux systems.
|
Thanks. I think this wins the award for the longest DESI branch name yet. Thanks for all the off-list discussion and checks about what is both robust and performance optimal. I checked that the final branch works on both GPU and non-GPU machines with default params. Looks good. Merging. |
sbailey
deleted the
235-rrdesi_mpi-gpu-out-of-memory-errors-unless-specifying-max-gpuprocs-4
branch
|
Haha! I thought the same - I clicked the button on this page to fork a new branch for this issue and it created that super long name for it :) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
…if it
results in a crash. If not, attempt to find the number of GPUs and set max_gpuprocs equal to that instead of the number of ranks. When using MPI, cupy.cuda.runtime.getDeviceCount() will not work if gpu-bind is used so we first attempt to check /proc and if that fails we gather the device PCI bus ids to count devices and issue a warning because this is slow (~2s for 64 ranks). Looking at /proc should work on all Linux systems.