Added the posibility to modify the control paramteres of UMFPACK alg #482
Conversation
|
Needs docs. |
src/factorization.jl
Outdated
| @@ -739,6 +739,7 @@ patterns with “more structure”. | |||
| Base.@kwdef struct UMFPACKFactorization <: AbstractFactorization | |||
| reuse_symbolic::Bool = true | |||
| check_pattern::Bool = true # Check factorization re-use | |||
| control::Vector{Float64}=SparseArrays.UMFPACK.get_umfpack_control(Float64,Int64) # Control Parameters of UMFPACK | |||
There was a problem hiding this comment.
This will heap allocate each time in the default algorithm. Instead, make a const of the default, set the default here to nothing, and then if nothing use the const global.
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #482 +/- ##
===========================================
- Coverage 64.22% 22.49% -41.74%
===========================================
Files 28 28
Lines 2200 2165 -35
===========================================
- Hits 1413 487 -926
- Misses 787 1678 +891 ☔ View full report in Codecov by Sentry. |
…qual to nothing; in the solver if no control has been given use the default values
src/factorization.jl
Outdated
| @@ -736,9 +736,12 @@ patterns with “more structure”. | |||
| `A` has the same sparsity pattern as the previous `A`. If this algorithm is to | |||
| be used in a context where that assumption does not hold, set `reuse_symbolic=false`. | |||
| """ | |||
| const default_control::Array{FLoat64}=[1.0, 0.2, 0.2, 0.1, 32.0, 0.0, 0.7, 0.0, 1.0, 0.3, 1.0, 1.0, 0.9, 0.0, 10.0, 0.001, 1.0, 0.5, 0.0, 1.0] | |||
There was a problem hiding this comment.
This is blocking the docstring above from going to the right thing.
Co-authored-by: Christopher Rackauckas <[email protected]>
Co-authored-by: Christopher Rackauckas <[email protected]>
| preserved. Defaults to true. | ||
| * `check_pattern`: Whether the sparsity pattern is checked for changes to allow for symbolic factorization caching. | ||
| The default is true. | ||
| * `control`: A control vector for more options to pass to UMFPACK. See the UMFPACK documentation for more details. |
There was a problem hiding this comment.
We should fill in the details here, since it has been unchanged for 20 years. Is there something simple that can be pasted, quoted, or linked to?
There was a problem hiding this comment.
eventualli we could somehow insert the table present in the doc of UFMPACK
There was a problem hiding this comment.
do u think would be appropriate?
Co-authored-by: Christopher Rackauckas <[email protected]>
#383
I have added a vector of type Float 64 in the UMFPACKFactorization structure in order to give the possibility to the user to plug in their customized control vector int the UMFPACK algorithm;
This control vector is used in the solve function and is passed to SparseArrays through the lu function.
Checklist
contributor guidelines, in particular the SciML Style Guide and
COLPRAC.
Additional context
For a deeper understing on the control vector i sugest to read subsection 5.10 at page 23 in the following pdf
https://github.com/DrTimothyAldenDavis/SuiteSparse/blob/dev/UMFPACK/Doc/UMFPACK_UserGuide.pdf