Open babreu-ncsa opened 2 years ago
Were you running this on nodes that were available through the Summer Institute reservation? And by chance do you happen to remember which node? If not, we might be able to track that down if you give me your user name.
We needed to make some modifications at the root level to allow users to access performance counters and run uProf and can check to make sure that paranoid level was set appropriately.
Hi Robert, thanks for your quick reply!
I am XSEDE-ECSS staff, so I have an account attached to project sds164, username is babreu. I was running this from the debug partition with
srun --partition=debug --pty --account=sds164 --ntasks-per-node=4 --mem=8G -t 00:30:00 --wait=0 --export=ALL /bin/bash
Let me know if there is any other information that may be helpful (this was also sent as a ticket via the XSEDE portal, ID 153559).
Hi Bruno - you're not doing anything wrong. We just don't have all the nodes configured correctly yet to run uProf. In fact, we just rebooted the nodes that were reserved for the Summer Institute yesterday morning. I guess the best I can say for now is sit tight and wait until changes have been pushed and all nodes rebooted.
Hi Robert, thank you again for your answer! I will try these commands again in the future. I'll keep this issue open for now in case someone else has the same problem, but as soon as I am able to run uProf with data-access, I will close it.
Data Access Sampling seems to not be working on Expanse. After compiling with
gfortran -march=znver2 -g -O3 -o dmadd_good dmadd_good.f
and running uProf to collect information withAMDuProfCLI-bin collect --config data_access -o dmadd_good-da ./dmadd_good
there is an error: ERROR : For non root users with perf_event_paranoid value greater than -1, the maximum number of events supported together is : 6