Skip to content

WIP: Delete file to trigger any "first touch" calls #519

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

Open
wants to merge 1 commit into
base: main
Choose a base branch
from

Conversation

roblatham00
Copy link
Contributor

Some storage systems (DAOS, PVFS, possibly others) do a "library init" the very first time a program calls MPI_File_Open or MPI_File_delete.

In some ways it's entirely fair to include these first touch time but it can also obscure the effects of optimizations or tuning to have one of the iterations consistently take 10x longer.

Since I don't know if this is a good idea or not, I thought I'd fire off this pull request to see if the community had any thoughts

Some storage systems (DAOS, PVFS, possibly others) do a "library init"
the very first time a program calls MPI_File_Open or MPI_File_delete.

In some ways it's entirely fair to include these first touch time
but it can also obscure the effects of optimizations or tuning to have
one of the iterations consistently take 10x longer.
Copy link
Contributor

@adilger adilger left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this should be a configurable option, and IMHO not enabled by default. If the initialization time is so significant that it is affecting the benchmark time, then it is also likely significant enough that it affects application runtime and should be measured and not hidden.

Maybe the initialization should be printed as part of the test output?

@roblatham00
Copy link
Contributor Author

Yeah, definitely should be an option. I should be running each iteration/config of the benchmark for tens of seconds anyway, so a one second initialization shouldn't be significant... Just takes a lot longer to parameter sweep (hours, days?) when doing it the right way.

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 this pull request may close these issues.

2 participants