Specialise LinearAlgebra.lmul! for LowerTriangular blockdiagonal matrices
#119
Conversation
| x = rand(rng, N1 + N2) | ||
| y = rand(rng, N2 + N4) | ||
|
|
||
| @testset "Lower triangular" begin |
There was a problem hiding this comment.
Do these tests fail on master, or are they just slow? Perhaps we could test for performance as well, see NamedDims.jl which does a few allocation tests, maybe some speed tests as well.
(I'm thinking higher level test, i.e. sampling)
There was a problem hiding this comment.
The test pass on master, just allocate more. So if I add a test for the case in the #116 (and add Distributions to the test deps)?
There was a problem hiding this comment.
I have added an allocation test just on lmul!. I suggest we leave higher level testing to a separate PR and plan it together with a benchmark suite.
Co-authored-by: Miha Zgubic <[email protected]>
|
Oh, also, squashing the commits might be a good idea |
Co-authored-by: Miha Zgubic <[email protected]>
Co-authored-by: Miha Zgubic <[email protected]>
Codecov Report
@@ Coverage Diff @@
## master #119 +/- ##
==========================================
+ Coverage 98.85% 98.88% +0.03%
==========================================
Files 5 5
Lines 348 358 +10
==========================================
+ Hits 344 354 +10
Misses 4 4
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here. |
…invenia/BlockDiagonals.jl into mjp/extend-lmul-lowertriangular
| x = rand(rng, N1 + N2) | ||
| y = rand(rng, N2 + N4) | ||
|
|
||
| @testset "Lower triangular" begin |
|
Actually do need a larger allocation bound. Julia 1.0 on x64 allocates 320. |
closes #116