Skip to content

Parallelise computing of trip pattern geometries - #3766

Merged
leonardehrenfried merged 2 commits into
opentripplanner:dev-2.xfrom
leonardehrenfried:parallel-geometries
Dec 9, 2021
Merged

Parallelise computing of trip pattern geometries#3766
leonardehrenfried merged 2 commits into
opentripplanner:dev-2.xfrom
leonardehrenfried:parallel-geometries

Conversation

@leonardehrenfried

@leonardehrenfried leonardehrenfried commented Dec 6, 2021

Copy link
Copy Markdown
Member

Summary

I noticed that when building a Norway graph 7 out of 20 minutes is taken up computing the trip pattern geometries. This process is run serially when it's quite parallelisable.

With these minor modifications I managed to reduce the graph build time from 20 to 13 minutes on my 16 core test machine.

Unit tests

n/a

Code style

yes

Documentation

n/a

@leonardehrenfried
leonardehrenfried requested a review from a team as a code owner December 6, 2021 12:40
@leonardehrenfried

Copy link
Copy Markdown
Member Author

@vesameskanen @optionsome If you're concerned about recent graph build slowdowns, you could give this a try.

@t2gran t2gran added this to the 2.1 milestone Dec 7, 2021
@t2gran t2gran added the !Optimization The feature is to improve performance. label Dec 7, 2021
Co-authored-by: Thomas Gran <t2gran@gmail.com>
@leonardehrenfried

Copy link
Copy Markdown
Member Author

@abyrd BTW, we talked about a 16 core machine leading to a 7-fold increase. I just rechecked the build machine and it only has 12 cores so it's not as bad as I originally thought.

@leonardehrenfried
leonardehrenfried merged commit f8998b2 into opentripplanner:dev-2.x Dec 9, 2021
@leonardehrenfried
leonardehrenfried deleted the parallel-geometries branch December 9, 2021 14:56
t2gran pushed a commit that referenced this pull request Dec 9, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

!Optimization The feature is to improve performance.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants