Christopher Rackauckas

Log transforms can make it numerically easier to stay positive. It's not automated so you'd have to do this by hand, for now, though there's work in MTK to do...

Yeah, this case just isn't handled. Not as much of a bug as a not handled case. If you have a VectorContinuousCallback and you have multiple events hit simultaneously (or...

Can you share an example which has the issue? I still haven't seen an example to work on.

Yes important point, added to the list.

We don't have to be too general, just have an interface which provides everything SciML needs, and update it as SciML needs more.

I talked with @YingboMa and the approach we're going to go with is to prefer canonicalization, i.e. the internals to the solvers won't need any `fmap` or special broadcast because...

It should be more efficient. We plan to have a non-allocating restructure!/destructure!

That's because for example there is no Float64 which equals 3//10 so `Float64(3//10) == 3//10 == false`. We could change https://github.com/SciML/DiffEqCallbacks.jl/blob/master/src/iterative_and_periodic.jl#L101 the condition to be, instead of exact, make it...

Posted to Discourse as well: https://discourse.julialang.org/t/22-seconds-to-3-and-now-more-lets-fix-all-of-the-differentialequations-jl-universe-compile-times/66313

- https://github.com/JuliaLang/julia/pull/41878 - https://github.com/JuliaLang/julia/pull/41877 - https://github.com/JuliaIO/Tar.jl/pull/123 - https://github.com/JuliaLang/Downloads.jl/pull/138 - https://github.com/JuliaLang/Pkg.jl/pull/2695 - https://github.com/JuliaSIMD/LoopVectorization.jl/pull/316 @timholy I think gathering them all here will make it easier to package up in the end.