added sparse multiplication and division for triangular matrices. Fix… #28507
Conversation
|
thank you @KlausC impatient n00b here, can this be merged? |
|
@mcognetta , @andreasnoack may I bring the PR back to your mind. |
|
Thanks, this is really nice. I will merge this into #28883. |
…ns' into optimize_binop_structured_matrices Adding sparse triangular sparse array operations from JuliaLang#28507
|
@andreasnoack is there a reason, why this is not merged into master? |
|
I suspect because it's a fair amount of code to review. @andreasnoack has got a few other things I happen to know he's working on somewhat urgently at the moment. Have patience, we'll have more cycles available next week. @Sacha0, any chance you could take a look? |
|
@andreasnoack @Sacha0 it's a little discouraging that important linear algebra improvements like this get forgotten |
|
Regrettably I have few spare cycles to review pull requests of this scale at the moment. Best! |
stdlib/SparseArrays/test/sparse.jl
Outdated
| A = sprand(rng, n, n, 0.01) | ||
| B = ones(n) | ||
| MA = Matrix(A) | ||
| for tr in (identity, adjoint, transpose) |
There was a problem hiding this comment.
It would be great if you could put a testset here that prints the loop variables. E.g. just something like
@testset "tr=$tr, wr=$wr" for
tr in (identity, adjoint, transpose),
wr in (UpperTriangular, LowerTriangular, UnitUpperTriangular, UnitLowerTriangular)helps a lot to identify the problem when a test fails.
The performance from #28451 has been improved for some types of sparse triangular matrices.
All the new performance figures are now in range 1-2ms, means improvements by factors 300-1000. See #28451.