Method to Identify Slow GPUs + Output bugfix #339
Merged
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Outdated suggestions cannot be applied.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.
Identifying Slow GPUs
Added a section to
OneTime::end()
that checks if the rank took more than 10% longer than average. If it did then it prints the node ID and GPU PCIe Bus ID to std::cerr. This is not enabled for all timers by default and enabling it requires passing the proper argument toOneTime::end()
; it's currently enabled only for theHydro_Integrator
timerThis should let us figure out if the weird performance issues on Frontier are Cholla or a small number of slow GPUs.
Output Bugfix
If a maximum number of timesteps was set with
N_STEPS_LIMIT
then the final time step would still output even if theOUTPUT
macro was not provided. Added ifdef statement to fix this.