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

local:remote in QPX-NPS4 is not reported correctly #377

Open
mabdallah89 opened this issue Jun 12, 2024 · 0 comments
Open

local:remote in QPX-NPS4 is not reported correctly #377

mabdallah89 opened this issue Jun 12, 2024 · 0 comments
Labels
enhancement New feature or request

Comments

@mabdallah89
Copy link

Is your feature request related to a problem? Please describe.
I am running a microbenchmark for QPX+NPS4 mode on MI300X and see the amount of remote:local traffic. However, “TCC_EA0_RDREQ_sum” is always equal “TCC_EA0_RDREQ_DRAM_sum”, no matter what I did in the kernel code . For example, I wrote my code such that each XCD reads all required data from other adjacent XCD’s memory, so all the traffic is remote, however, these the two counters are still the same. It seems what's happening in non-SPX mode on MI300, these requests are still being counted as going "to a DRAM, somewhere".

Describe the solution you'd like
What I expect as in MI200 multi-socket, in this microbenchmark, "TCC_EA0_RDREQ_sum" should be higher than "TCC_EA0_RDREQ_DRAM_sum”
Some further tests would be:
• On MI300X, I still expect accesses to the CPU DRAM to be counted as "remote"
• We need to check how these are counted on MI300A, where everything is HBM.
• We need to check how memory access to a different socket is counted

@mabdallah89 mabdallah89 added the enhancement New feature or request label Jun 12, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request
Projects
None yet
Development

No branches or pull requests

1 participant