Pardiso

[vpuri3]

todo

• finish boilerplate
• test on some realistic sciml usecase
• test on pardiso.jl/examples

close #5

[codecov]

Codecov Report

Merging #37 (2d3822f) into main (e99d699) will decrease coverage by 0.19%.
The diff coverage is 83.72%.

@@            Coverage Diff             @@
##             main      #37      +/-   ##
==========================================
- Coverage   85.50%   85.31%   -0.20%     
==========================================
  Files           5        6       +1     
  Lines         276      320      +44     
==========================================
+ Hits          236      273      +37     
- Misses         40       47       +7     

Impacted Files	Coverage Δ
src/wrappers.jl	91.15% <ø> (ø)
src/pardiso.jl	83.33% <83.33%> (ø)
src/LinearSolve.jl	62.50% <100.00%> (+12.50%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update e99d699...2d3822f. Read the comment docs.

[vpuri3]

@ChrisRackauckas i got some boilerplate code with a little functionality. can you point me to usecases in sciml ecosystem to test against so i can figure what more to add if needed? else i can test this with the examples in https://github.com/JuliaSparse/Pardiso.jl/tree/master/examples

[vpuri3]

this is largely done - just need to figure out the right defaults

[ChrisRackauckas]

doesn't pardiso have a mixed factorized and iterate thing?

[vpuri3]

yeah - ANALYSIS_NUM_FACT_SOLVE_REFINE

[vpuri3]

doing that too

[vpuri3]

@KristofferC , we're writing a Pardiso.jl wrapper for SciML ecosystem. Would like for you to have a look at this. I was also wondering how one could control abstol, and reltol. Would it be possible by supplying appropriate iparm/dparms?? cant find any reference to tolerance in docs

[ChrisRackauckas]

This looks good to merge. Of course, more can be tweaked in further PRs as better defaults are found.