Seth Axen

Sounds good! Happy to do it.

Here's an initial repo with a failing test: https://github.com/sethaxen/Buffers.jl. `Buffer`'s adjoints require `Zygote.grad_mut`, which is dependent on `Zygote.Context` and `Zygote.cache`, not just `ZygoteRules.AContext` I don't really understand how `grad_mut` works....

Sorry it took a while to get back to this. I've finished porting the `Buffer`-related code and tests to https://github.com/sethaxen/Buffers.jl, and the tests pass. However, after loading Buffers, `Buffer` seems...

> [These methods](https://github.com/sethaxen/Buffers.jl/blob/master/src/lib.jl) are all type-piratical. Not sure if that's the direct cause but would be worth fixing. Those are all created in Zygote, but this package only depends on...

Yes, this is one of several dependency-related issues, see #44

Hm, I seem to recall it taking longer to load in the past. I wonder if invalidations were recently fixed in a dependency.

I agree, a benchmarking GitHub Action, either run automatically or triggered via a label, would be good. Since you and I have been noting opposite trends comparing the performance of...

Note that the test suite doesn't currently pass because due to some bugs on main. I fixed one in this PR, but not the other.

On my machine, a non-exhaustive list of load times of the heaviest dependencies in a clean environment: - LoopVectorization: 2.7s - AxisKeys: 0.86s - MCMCDiagnosticTools: 0.80s - Distributions: 0.77s -...

> Another way to cut load times might be switching from FFTESS to another ESS method so we can drop FFTW, but I have no idea how they compare performance/accuracy-wise....