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

Measure cycles instead of time for benchmarking #632

Merged
merged 36 commits into from
Oct 18, 2024

Conversation

oflatt
Copy link
Member

@oflatt oflatt commented Oct 16, 2024

This PR does several things (sorry):

  • changes the runtime to expose functions that measure cycles on arm and x86
  • changes eggcc to use the updated bril fork that instruments the llvm code to use these
  • changes all the benchmarks to not have loops that run them multiple times
  • changes the benchmarks to have a single print statement before existing to remove I/O overhead
  • copies polybench benchmarks into test folder so that we still test the printed matrices
  • fixes parallelism in nightly so that it exits on first error

@@ -1,30 +1,6 @@
# ARGS: 2550
# ARGS: 38432 25
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

38432 and 25 are coprime so this benchmark is now too easy?

Copy link
Collaborator

@yihozhang yihozhang Oct 17, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nevermind, coprime actually takes longer for gcd to compute

@@ -1,27 +1,10 @@
# ARGS: 9500
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

why's the input smaller now?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It was overflowing or something, but it didn't need to be that big.

Copy link
Collaborator

@yihozhang yihozhang left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

lgtm

@oflatt oflatt merged commit 950ec94 into main Oct 18, 2024
4 checks passed
@oflatt oflatt deleted the oflatt-precise-microbenchmark branch October 18, 2024 17:35
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants