Update sparspak #253
Update sparspak #253
Conversation
* Base sparspak interface on sparspaklu * Introduce tests for MultiFloats and ForwardDiff
|
In fact precompilation does not help much here - there are so many different generic FP types (and ForwardDiff breeds them in masses with diffent tags anyway) that probably a hint for users to care about precompilation themselves in the docs would make sense. |
|
This looks good. Agreed on the precompilation. |
|
Precompiling for Float64, which is by far the most common case, is probably useful though. |
src/factorization.jl
Outdated
| Sparspak.Problem.infullrhs!(p, b) | ||
| s = Sparspak.SparseSolver.SparseSolver(p) | ||
| return s | ||
| return sparspaklu(A) |
There was a problem hiding this comment.
this performs an extra lu-factorization though? Can we avoid this and just build a blank/empty LU object to fill later?
There was a problem hiding this comment.
I see. Assuming I have all info on A this is possible but not in the API. I think I could add a kwarg to sparspaklu.
There was a problem hiding this comment.
I think a kwarg would be a nice solution.
There was a problem hiding this comment.
This will be in PetrKryslUCSD/Sparspak.jl#16 . Once this is done I will add a corresponding commit here.
Codecov Report
@@ Coverage Diff @@
## main #253 +/- ##
==========================================
+ Coverage 65.99% 66.31% +0.31%
==========================================
Files 12 12
Lines 744 754 +10
==========================================
+ Hits 491 500 +9
- Misses 253 254 +1
📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
|
I'll run the formatter, thanks! |
No description provided.