William Moses

What variant of MPI are you using? cc @vchuravy

The code as is doesn't reproduce locally due to an undefined variable. Moreover, can you reduce this to just be a case of calling enzyme.autodiff calling a crash rather than...

Reduced to remove all of sciml. From a minimal test, I believe this error is due to the type instable `TensorProductBasis` (and you'll eventually hit a no reverse pass defined...

```julia using Enzyme abstract type TensorProductBasis

Having not yet fully reduced (nor yet investigated), my guess is what's happening is that the dynamism creates the use of split mode internally -- which triggers a bad interaction...

Never mind, seems to be something in the calling convention.

Yeah basically when enzyme-inline is run Enzyme itself inlines a bunch of stuff for the reverse pass it generates. This results in many different calls to ptls states, sometimes in...

As of master this error longer occurs [presumably from the returnstwice fix]. However, the actual error with exceptional control flow is below: ``` mod: ; ModuleID = 'text' source_filename =...

In theory, setjmp could be handled if, as in this case, the return of nonzero always results in an unreachable.

Doing a test differentiation by marking various things as inactive yields the following internal error, which perhaps you can shed some light on: ``` mod = ; ModuleID = 'text'...