Manuel Drehwald

@samuelpmish I recently got a lot of Enzyme issues reported from the Rust side and some included incorrect derivatives. https://github.com/EnzymeAD/rust/issues/196 All of them were simple data types and functions, so...

@tgymnich Do you know why CI still isn't running this?

Doesn't seem to help much, any other ideas? Iirc there were so discussions about only running them on releases, since they were so slow, could that be an issue?

closing this since the current code has too many bugs, in favor of #2035 where I have more fixes

The explorer for llvm-20 seems broken, but 18 runs in an Enzyme error, even though a different one. https://fwd.gymni.ch/2iMsLw Rust's llvm fork is at commit acaea3d2bb8f3, but I guess any...

Now that deadlines are over, ping @wsmoses

Per your chat question, it fails on LLVM 20 and 19 with the same error shown above. On LLVM18, 17, and 16 it fails with (but I obviously don't care...

I'd be very curious if you can come up with an example where Enzyme can't eliminate dead code, do you have a concrete one in mind? Compilers like LLVM have...

Ah, I think I see what you mean. The problem here is that whether it can be NaN will (in more complex cases) at some point be optimization dependent, e.g....

I don't have the capacity to help with Enzyme debugging, but as an FYI on the Rust side we're able to successfully build enzyme on both x86 and aarch MacOS...