Skip to content

Draft: Claude Code implementation of a replacement for jkg_utils that runs faster#7

Open
BrendanKKrueger wants to merge 45 commits intomainfrom
bkk_commutator
Open

Draft: Claude Code implementation of a replacement for jkg_utils that runs faster#7
BrendanKKrueger wants to merge 45 commits intomainfrom
bkk_commutator

Conversation

@BrendanKKrueger
Copy link
Copy Markdown
Collaborator

This was a test of using Claude Code, because actually mucking around with jkg_utils is a really low priority right now.

  • Claude Code's own implementation fails its own correctness check. Look at both the implementation and the tests to fix that.
  • Look at the performance check to verify that.
  • Run the performance check to compare.
  • Consider if additional performance enhancements are possible and worth spending time on.

@BrendanKKrueger BrendanKKrueger self-assigned this Apr 17, 2026
…aster. Claimed 20-200x speedup, but I've not tested or read through the code.
…written by Claude) now passes, but I've not verified anything. The benchmark shows a whopping 1x speedup (in other words: no change). Let's see if Claude can do better.
…a time-limited Monte Carlo routine, and that the better measure is not wall time but throughput.
…e the equations, and verify the equations from jkg_utils.
… by the step counts not varying correctly with the total timestep. It misdiagnosed the cause, but I was able to figure out what went wrong once Claude Code pointed to the existence of a problem.
…Claude wanted to put it in a different location than I thought was appropriate, so I'm making the change manually.
… few million samples, now close to a billion.
…document because we don't need that in the main branch.
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