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.

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

Jump to bottom

[BUG] CBO causes less efficient plan for NDS q84 #2739

Closed
andygrove opened this issue Jun 18, 2021 · 1 comment · Fixed by #2794
Closed

[BUG] CBO causes less efficient plan for NDS q84 #2739

andygrove opened this issue Jun 18, 2021 · 1 comment · Fixed by #2794
Assignees
@andygrove
Labels
bug Something isn't working performance A performance related task/issue
Milestone
June 21 - July 2

Comments

@andygrove
Copy link
Contributor

andygrove commented Jun 18, 2021 •
edited
Loading

Describe the bug
With the cost-based optimizer enabled, there is a CPU projection between two GPU BroadcastHashJoin operators, which does not make sense.

cpu-project-between-gpu-bhj

Steps/Code to reproduce bug
Decimals enable, AQE on, CBO on. SF=3k.

Expected behavior
Projection should stay on GPU.

Additional context
None
@andygrove andygrove added bug Something isn't working ? - Needs Triage Need team to review and classify performance A performance related task/issue labels Jun 18, 2021
@andygrove andygrove added this to the June 7 - June 18 milestone Jun 18, 2021
@andygrove andygrove self-assigned this Jun 18, 2021
@sameerz sameerz removed the ? - Needs Triage Need team to review and classify label Jun 22, 2021
@andygrove
Copy link
Contributor Author

I tracked the issue down and it was an integer overflow when estimating the size of data that an operator would produce. The estimated row counts were extremely high, which is why I could only reproduce this on a large data set.

@andygrove andygrove mentioned this issue Jun 23, 2021
Merged
@andygrove andygrove changed the title [BUG] CBO causes less effecient plan for NDS q84 [BUG] CBO causes less efficient plan for NDS q84 Jun 23, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@andygrove andygrove

Labels
bug Something isn't working performance A performance related task/issue
Projects
None yet
Milestone
June 21 - July 2
Development

Successfully merging a pull request may close this issue.

Prevent integer overflow when estimating data sizes in cost-based optimizer andygrove/spark-rapids
2 participants
@andygrove @sameerz