William Moses

``` Test Summary: | Pass Total Bessel | 24 24 AbstractType calling convention: Test Failed at /mnt/Data/git/Enzyme.jl/test/runtests.jl:431 Expression: Enzyme.autodiff(dxdt_pred, Active(1.0)) Expected: UndefRefError No exception thrown Stacktrace: [1] macro expansion @...

Wait actually, that is good not to have thrown, one second.

Check also https://github.com/EnzymeAD/Enzyme.jl/issues/293 which was a minimal example not resolved by iirc.

(it also is referenced by several other errors that became variants of that issue, some of which were in fact resolved by the PR)

I'm not able to find a Github reference for this, but this question is very similar to another one we had a while back about taking the minimum of an...

We can also do some other tricks to handle measure zero spaces [determined by a floating point equality check] explicitly (or say insert things automatically), but I thought this a...

One thing that I wanted to add here, just to illustrate how FDM may completely fail here, consider: ``` julia> grad(central_fdm(5, 1), min1, [2, 3, 2, 4, 5.0, 2]) ([0.5000000000050965,...

Sure in this example when composed with fill, numeric differentiation can get the correct derivative of the entire function. The fact that that case works, however, doesn't detract from the...

Current error ```julia ┌ Warning: Returned rooting not fully handled, segfault likely └ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/N98un/src/utils.jl:35 ERROR: LoadError: UndefRefError: access to undefined reference Stacktrace: [1] getproperty @ ./Base.jl:42 [inlined] [2]...

This is now an absence of lapack support. However, julia no longer crashes.