Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Diffrent solvers and Precond. for Hypre #216

Merged
merged 11 commits into from
Jan 31, 2025
Merged

Diffrent solvers and Precond. for Hypre #216

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
552a9ee
[reorg]
mohd-afeef-badri Jan 29, 2025
d6fe7a5
[bug] HYPRE_BoomerAMGCreate called twice
mohd-afeef-badri Jan 29, 2025
7020aff
new parameters for solver and preconditioner
mohd-afeef-badri Jan 29, 2025
7e04965
GMRES
mohd-afeef-badri Jan 29, 2025
a2fae91
[test] gmres test
mohd-afeef-badri Jan 29, 2025
41cee57
fgmres
mohd-afeef-badri Jan 29, 2025
a2309b7
bicgstab and Block Jacobi Precod GPU
mohd-afeef-badri Jan 29, 2025
0f8a489
added doc
mohd-afeef-badri Jan 29, 2025
fd31ad9
pass via enum logic for solver and prcond
mohd-afeef-badri Jan 30, 2025
a8ecc5a
error catching
Jan 31, 2025
1170d0e
hypre-info and hypre-timer
mohd-afeef-badri Jan 31, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
28 changes: 15 additions & 13 deletions doc/hypreParametersCheatSheet.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,20 +1,22 @@
### ArcaneFEM interface to Hypre cheat sheet ###
### ArcaneFEM interface to Hypre cheat sheet

Following abbreviations can be useful for reading the table below.

- [D] - Default
- [I] - Integer type
- [R] - Real type
- [S] - String type

| Parameter | Use | Comment |
| ----------------- | ---------------------------------------------------- | ------------------------------------------------------------ |
| `rtol` | relative convergence tolerance for the Krylov Solver | [R] [D] = 1.0e-7 |
| `atol` | absolute convergence tolerance for the Krylov Solver | [R] [D] = 0.0<br />set `rtol` to 0 if you want to use `arol` |
| `max-iter` | maximum Krylov iterations | [I] [D] = 1000 |
| `verbosity` | verbosity for the log | [I] [D] = 2 |
| `amg-coarsener` | amg parallel coarsening algorithm | [I] [D] = 8<br />0 : CLJP <br />3 : classical RS <br />6 : Falgout <br />8 : PMIS<br />10 : HMIS<br />21 : CGC <br />22 : CGC-E<br />**Note** GPU supported choice 8 |
| `amg-threshold` | amg threshold strength | [R] [D] =0.25 <br /># 2D (rep. 3D) Laplace 0.25 (rep. 0.6-0.6) is a good value<br /># Elasticity, a large threshold, such as 0.9 |
| `amg-interp-type` | amg interpolation operator | [I] [D] = 6<br />0 : classical <br />1 : LS (for use with GSMG) <br />2 : classical (hyperbolic PDEs)<br />3 : direct with separation of weights <br />4 : multipass<br />5 : multipass with separation of weights<br />6 : extended+i<br />7 : extended+i no common C neighbor<br />8 : standard <br />9 : standard with separation of weights<br />10 : classical block (use for nodal systems)<br />11 : 10 with diagonalized diagonal blocks <br />12 : FF interpolation<br />13 : FF1 interpolation<br />14 : extended<br />15 : adaptive weights<br />16 : extended in matrix-matrix form<br />17 : extended+i in matrix-matrix form<br />18 : extended+e in matrix-matrix form<br />**Note** GPU supported choice 3, 6, 14, 15, 18 |
| `amg-smoother` | amg smoother to be used | [I] [D] = 6<br />0 : Jacobi<br/>3 : hybrid GS or SOR, forward solve<br/>4 : hybrid GS or SOR, backward solve<br/>6 : hybrid symmetric GS or SSOR<br/>7 : Jacobi with Matvec<br />8 : $l_1$-scaled hybrid symmetric GS<br/>9 : Gaussian elim. (on coarsest level)<br/>10 : On-proc. direct forward solve for matrices with triangular structure<br/>11 : Two Stage approx. to GS. Uses lower part diagonal matrix<br/>12 : 11 and a second iteration for error approx.<br/>13 : $l_1$ Gauss-Seidel, forward solve<br/>14 : $l_1$ Gauss-Seidel, backward solve<br/>16 : Chebyshev<br/>17 : FCF-Jacobi<br/>18 : $l_1$-scaled jacobi<br/>30 : Kaczmarz<br/>88: 8 with a convergent l1-term<br/>89: Symmetric l1-hybrid GS (i.e., 13 followed by 14)<br/>98 : LU with pivoting<br />**Note** GPU supported choice 3, 4, 6, 7, 18, 11, 12 |
| | | |

| Parameter | Use | Comment |
| ------------------- | ---------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `solver` | apllied Krylov solver | [S] [D] = "cg"<br />"cg" : Conjugate Gradient<br />"gmres" : GMRES<br />"fgmres" : Flexible GMRES<br />"bicgstab" : BiCGSTAB |
| `preconditioner` | applied preconditioner | [S] [D] = "amg"<br />"amg" : BoomerAMG<br />"bjacobi" : Block Jacobi with 0 fill |
| `rtol` | relative convergence tolerance for the Krylov Solver | [R] [D] = 1.0e-7 |
| `atol` | absolute convergence tolerance for the Krylov Solver | [R] [D] = 0.0<br />set `rtol` to 0 if you want to use `arol` |
| `max-iter` | maximum Krylov iterations | [I] [D] = 1000 |
| `verbosity` | verbosity for the log | [I] [D] = 2 |
| `amg-coarsener` | amg parallel coarsening algorithm | [I] [D] = 8<br />0 : CLJP <br />3 : classical RS <br />6 : Falgout <br />8 : PMIS<br />10 : HMIS<br />21 : CGC <br />22 : CGC-E<br />**Note** GPU supported choice 8 |
| `amg-threshold` | amg threshold strength | [R] [D] =0.25<br /># 2D (rep. 3D) Laplace 0.25 (rep. 0.6-0.6) is a good value<br /># Elasticity, a large threshold, such as 0.9 |
| `amg-interp-type` | amg interpolation operator | [I] [D] = 6<br />0 : classical <br />1 : LS (for use with GSMG) <br />2 : classical (hyperbolic PDEs)<br />3 : direct with separation of weights <br />4 : multipass<br />5 : multipass with separation of weights<br />6 : extended+i<br />7 : extended+i no common C neighbor<br />8 : standard <br />9 : standard with separation of weights<br />10 : classical block (use for nodal systems)<br />11 : 10 with diagonalized diagonal blocks <br />12 : FF interpolation<br />13 : FF1 interpolation<br />14 : extended<br />15 : adaptive weights<br />16 : extended in matrix-matrix form<br />17 : extended+i in matrix-matrix form<br />18 : extended+e in matrix-matrix form<br />**Note** GPU supported choice 3, 6, 14, 15, 18 |
| `amg-smoother` | amg smoother to be used | [I] [D] = 6<br />0 : Jacobi ``3 : hybrid GS or SOR, forward solve``4 : hybrid GS or SOR, backward solve ``6 : hybrid symmetric GS or SSOR``7 : Jacobi with Matvec<br />8 : $l_1$-scaled hybrid symmetric GS ``9 : Gaussian elim. (on coarsest level)``10 : On-proc. direct forward solve for matrices with triangular structure ``11 : Two Stage approx. to GS. Uses lower part diagonal matrix``12 : 11 and a second iteration for error approx.``13 : $l_1$ Gauss-Seidel, forward solve``14 : $l_1$ Gauss-Seidel, backward solve ``16 : Chebyshev``17 : FCF-Jacobi ``18 : $l_1$-scaled jacobi``30 : Kaczmarz ``88: 8 with a convergent l1-term``89: Symmetric l1-hybrid GS (i.e., 13 followed by 14)``98 : LU with pivoting<br />**Note** GPU supported choice 3, 4, 6, 7, 18, 11, 12 |
| | | |
Loading