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

Update to Racket 8.0 #103

Merged
merged 6 commits into from
Aug 18, 2021
Merged

Update to Racket 8.0 #103

merged 6 commits into from
Aug 18, 2021

Conversation

bksaiki
Copy link
Contributor

@bksaiki bksaiki commented Aug 17, 2021

This PR supports Racket 8.0 by borrowing Herbie's single-flonum compatability library. The CI runner has been bumped to "ubuntu-latest" since support for 16.04 is ending. Also added Java and OCaml actions to CI for a cleaner and faster setup.

@pavpanchekha
Copy link
Contributor

This is good stuff. A few questions:

  • You removed the caching action—is that right? Why?
  • There are probably other actions we should use, like to set up SBT. It's like Sollya does not have an action, but perhaps we should factor that out?
  • What is the single-precision support in the FPBench tools about? Perhaps the right move is actually to rip it out? (If it's about the reference interpreter, I see no real reason the reference interpreter should support single-precision if it also supports double.)

@bksaiki
Copy link
Contributor Author

bksaiki commented Aug 17, 2021

I removed caching because its caused more problems than helped. It's extremely messy when changing CI. Making every install an action would probably clean things up even more. Single precision is needed to force rounding when testing single-precision.

@pavpanchekha
Copy link
Contributor

Ok, I'm fine with that reason for removing rounding.

For actions per-project, that seems like 1) a nice thing for FPBench to maintain, and 2) check out my setup-z3 action for a simple template, unfortunately the docs are not quite as good as you'd like here.

What do you mean by "forcing rounding when testing single-precision"? Isn't that just flsingle? And what are we comparing against?

@bksaiki
Copy link
Contributor Author

bksaiki commented Aug 17, 2021

Issue resolved. Did some testing and I found the float32 library is unnecessary since the reference interpreter and the sampler will produce valid single-precision values.

@bksaiki bksaiki merged commit cfb4b3f into master Aug 18, 2021
@pavpanchekha pavpanchekha deleted the racket-8.0 branch December 18, 2021 23:43
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