Skip to content
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

Bug in computation of operational intensity with likwid-mpiirun #530

Open
christiealappatt opened this issue Jun 2, 2023 · 1 comment
Open
Labels

Comments

@christiealappatt
Copy link
Contributor

Description
The operational intensity reported with likwid-mpirun seems to have a bug.
The problem is that the operational intensity is only counted for cores which count the uncore memory traffic. The rest of the cores have 0 operational intensity.
Therefore, at the end in the summary table the sum value is way lower that the actually expected operational intensity.
Note that this error does not happen with likwid-perfctr.

To Reproduce
To observe the bug try executing a code which does roughly same work on all the available processes. We expect the operational intensity to be roughly the same for different number of processes. But with likwid-mpirun the operational intensity of the following two cases different.

  • likwid-mpirun -np 4-g MEM_DP ./a.out
  • likwid-mpirun -np 1-g MEM_DP ./a.out

I used LIKWID version 5.2.

@TomTheBear
Copy link
Member

The operational intensity is a special metric as it combines hwthread counts (FP rate) with socket/memoryDomain counts (memory bandwidth) and there is specific code for this metric. likwid-mpirun does not contain this special support.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

No branches or pull requests

2 participants