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

Mark lazy spillables as allowSpillable in during gatherer construction #8989

Merged
merged 1 commit into from
Aug 11, 2023

Conversation

abellina
Copy link
Collaborator

Closes #8942

This fixes a memory pressure issue when creating gatherers in GpuHashJoin. At 100K we saw query14_part1 and query14_part2 fail with OOM right around this part of the code. It looks like we were keeping the build and stream data batches non-spillable as we created the gather map. This change makes it so we can spill the data sides during this operation.

I do not see regressions with this change at NDS 3TB. This change allows query14_part1 and part2 to succeed on every test I have tried.

I really want to remove LazySpillable, but I'll save that for a different PR.

@jlowe
Copy link
Member

jlowe commented Aug 10, 2023

build

@jlowe jlowe merged commit b53c803 into NVIDIA:branch-23.10 Aug 11, 2023
28 checks passed
@sameerz sameerz added the reliability Features to improve reliability or bugs that severly impact the reliability of the plugin label Aug 11, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
reliability Features to improve reliability or bugs that severly impact the reliability of the plugin
Projects
None yet
Development

Successfully merging this pull request may close these issues.

[BUG] NDS query 14 parts 1 and 2 both fail at SF100K
4 participants