Michel Schanen

Okay. Finally got that second-order. Sorry for removing previous posts. ```julia using Enzyme # Rosenbrock @inline function f(x...) (1 - x[1])^2 + 100 * (x[2] - x[1]^2)^2 end @inline function...

> @michel2323 I think the rules are now fixed. I hope you did not mind me reformatting your tests. @wsmoses if @michel2323 does not respond, we might need another reviewer....

Since we're using the memory, I added the batched mode to get the Hessian in one go.

https://github.com/EnzymeAD/Enzyme.jl/issues/585 @vchuravy This should be merged I guess. Just checking again.

@vchuravy @wsmoses The MPI test crashes now. With 1.9 we have the finalizer issue, and with 1.8 the logs and Manifest.toml are attached. [log.tar.gz](https://github.com/EnzymeAD/Enzyme.jl/files/11804637/log.tar.gz)

@wsmoses @vchuravy I marked the test here as broken so we can merge the MPI tests and mark it unbroken with the fix. Issue opened https://github.com/EnzymeAD/Enzyme.jl/issues/1138.

I get this now on the most recent Enzyme build: ``` ERROR: LoadError: ERROR: LoadError: Enzyme execution failed. Enzyme: jl_call calling convention not implemented in aug_forward for %33 = call...

Since the error is in the compilation, it can be triggered via `include("test/mpi.jl")`—no need to run with two ranks.

@wsmoses This still breaks the same way on `KA#main` and `Enzyme#main` (self-compiled `[email protected]`). No `CUDA` needed. Removed the `synchronize` in the kernel to not be dependent on the KA GPU...

This works with arrays. I've checked now with KA using `CPU()` and `CUDABackend()`. What is weird is that with `CUDABackend()` the rule gets triggered, however with `CPU()` not. Also, in...