Closed maltezfaria closed 1 year ago
Merging #22 (6a7553e) into main (0d41571) will decrease coverage by
0.16%
. The diff coverage is73.44%
.
@@ Coverage Diff @@
## main #22 +/- ##
==========================================
- Coverage 71.11% 70.94% -0.17%
==========================================
Files 12 12
Lines 1395 1401 +6
==========================================
+ Hits 992 994 +2
- Misses 403 407 +4
Impacted Files | Coverage Δ | |
---|---|---|
src/HMatrices.jl | 100.00% <ø> (ø) |
|
src/dhmatrix.jl | 0.00% <0.00%> (ø) |
|
src/partitions.jl | 58.82% <53.33%> (+2.03%) |
:arrow_up: |
src/hmatrix.jl | 61.30% <67.45%> (-6.60%) |
:arrow_down: |
src/rkmatrix.jl | 69.69% <68.57%> (+1.80%) |
:arrow_up: |
src/addition.jl | 72.72% <76.31%> (+2.91%) |
:arrow_up: |
src/compressor.jl | 82.02% <78.94%> (+1.54%) |
:arrow_up: |
src/utils.jl | 80.30% <78.94%> (-3.91%) |
:arrow_down: |
src/multiplication.jl | 79.25% <88.59%> (+3.19%) |
:arrow_up: |
src/kernelmatrix.jl | 88.88% <100.00%> (ø) |
|
... and 12 more |
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.
The current
lu
factorization is a bit hacky, and before trying to useDataFlowTasks
to improve the parallel performance of the factorization it makes sense to factor things up a bit. In particular, this branch simplifies some of the logic involved in the algorithm for the hierarchical multiplication based on a dry run, together with theMulLinearOp
structure which aids in assembling the low-rank approximations without allocating intermediate results.