You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hopefully this is a quick one: although xGESV is broken out in Solve right now, there's a very similar pair of functions in the issue name, documented e.g. on he very top of this page, that (xGETRF) factor a matrix into a form conducive to (xGETRS) using it to solve for many different "B" vectors. If "A" is large and "B" changes quite a bit more frequently than "A", this gives some tangible performance improvements.
This technically also provides you with an LU factorization, but I think that's an implementation detail that shouldn't necessarily be relied upon.
If you don't have the time, let me know; I could be motivated to turn this issue into a PR for my current use case.
Thanks!
The text was updated successfully, but these errors were encountered:
Ah, the Decomposition class wasn't linked from the main documentation--it appears the xGETRF half is already implemented. I still don't see the xGETRS, unfortunately, nor do I see any other fast way to get quick solutions for (two) triangular matrices. Given that the necessary substitution is fairly trivial, it might not actually be worth the JNI boundary cost, but I'd still consider adding it for completeness. At the very least, this will make for a working prototype :)
Hello!
Hopefully this is a quick one: although xGESV is broken out in
Solve
right now, there's a very similar pair of functions in the issue name, documented e.g. on he very top of this page, that (xGETRF) factor a matrix into a form conducive to (xGETRS) using it to solve for many different "B" vectors. If "A" is large and "B" changes quite a bit more frequently than "A", this gives some tangible performance improvements.This technically also provides you with an LU factorization, but I think that's an implementation detail that shouldn't necessarily be relied upon.
If you don't have the time, let me know; I could be motivated to turn this issue into a PR for my current use case.
Thanks!
The text was updated successfully, but these errors were encountered: