Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fully mutating linear solver #48

Open
gdalle opened this issue May 24, 2023 · 4 comments
Open

Fully mutating linear solver #48

gdalle opened this issue May 24, 2023 · 4 comments
Labels
performance Improve performance or stability

Comments

@gdalle
Copy link
Member

gdalle commented May 24, 2023

At the moment, AbstractDifferentiation does not have a fully mutating API, and benchmarks show allocations are non-negligible when multiplying by A and B. Opening this to track JuliaDiff/AbstractDifferentiation.jl#14 and see what we can do

@gdalle gdalle added the feature New feature or request label May 24, 2023
@mohamed82008
Copy link
Collaborator

The problem is that most AD packages don't actually support mutation.

@gdalle
Copy link
Member Author

gdalle commented May 24, 2023

Yeah I know, it would be useless for ChainRules, but not for ForwardDiff?

@gdalle gdalle added performance Improve performance or stability and removed feature New feature or request labels Aug 8, 2023
@gdalle
Copy link
Member Author

gdalle commented Feb 21, 2024

Trying to solve this with https://github.com/gdalle/DifferentiationInterface.jl

@gdalle
Copy link
Member Author

gdalle commented Apr 10, 2024

This is now possible with DifferentiationInterface but

  • it won't work with every outer backend
  • it requires a mutating version of the conditions conditions!(c, x, y)

I hope to make it part of v0.7

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
performance Improve performance or stability
Projects
None yet
Development

No branches or pull requests

2 participants