Kristoffer Carlsson
Kristoffer Carlsson
@nanosoldier `runtests()`
Another thing that can be used is GFlops.jl (requires the https://github.com/charleskawczynski/GFlops.jl/tree/ck/julia1.10 branch). ```julia julia> tree = BallTree(rand(3,10^5)); julia> v = rand(3) 3-element Vector{Float64}: 0.38248598898980024 0.6378296367207168 0.9255061078525778 julia> @count_ops knn(tree, v,...
On dev it is at least: https://docs.julialang.org/en/v1.10-dev/base/base/#__init__ (link gets corrupted)
ForwardDiff requires that code is written generically. The Bunchkaufman code only works for normal floating point numbers so that's why it errors.
If you can compute the derivative analytically you can add an overload for `baunchkaufman(v::Vector{
ForwardDiff has a custom number type `D` and claims that `d::D + x` should be evaluated as `d + D(x)`. Symbolics has a custom number type `N` and claims that...
Yes, see https://juliadiff.org/ForwardDiff.jl/stable/user/advanced/#Retrieving-Lower-Order-Results.
I just want to point out that this is still happening and causes ~18s load time increase to any package that recursively ends up getting Enzyme as a dependency.
- Someone adds a method that invalidates some function called in `Enzyme.__init__`. - `Enzyme.__init__` in invalidated and has to be recompiled at load time. - Compiling `Enzyme.__init__` takes about 18...
One would usually use SnoopCompile. I can try (later) to see if I can find what actually ends up invalidating it.