Skip to content

fix: use default maxBytes and ncclBuffSize for p4 running all-to-all #691

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.

Sign up for GitHub

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

Merged
merged 1 commit into from
Oct 2, 2025
Merged

fix: use default maxBytes and ncclBuffSize for p4 running all-to-all #691

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 7 additions & 2 deletions test/cases/nvidia/mpi_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -70,8 +70,13 @@ func multiNode(testName string) features.Feature {
ncclBuffSize := "4194304"
if slices.Contains(instanceSupportsRdmaRead, *nodeType) {
t.Log("Instance supports RDMA")
maxBytes = "16G"
ncclBuffSize = "8388608"
// TODO: revisit this with some kind of per-instance optimizer, or maybe use the defaults for all instance types unless specified
if testName == "alltoall_perf" && strings.Contains(*nodeType, "p4") {
// Keep default values for P4 running all-to-all
} else {
maxBytes = "16G"
ncclBuffSize = "8388608"
}
}
var err error
renderedMpiJobNcclTestMultiNodeManifest, err = fwext.RenderManifests(mpiJobNcclTestMultiNodeManifest, ncclTestManifestTplVars{
Expand Down
Loading