A proposal on adding new TFQMR algorithm
Dear all, I have submitted an python implementation of the sparse linear iterative algorithm "Transpose-Free QMR" (https://github.com/scipy/scipy/pull/14160). This algorithm (cited by 904) is very effective to solve non-Hermitian linear systems because it has a smoother convergence curve than the other algorithms like CGS, Bi-CGSTAB, and it has been widely used in those prominent scientific computing packages such as PETSc etc. The purpose of adding the algorithm is to enrich SciPy's sparse linear library, so that users can have more/better choices when using Krylov subspace methods in SciPy to solve linear systems. I also hope that experts in this field can review the implementation of the algorithm (References are listed in "tfqmr.py"). Hope you like it. Thanks, Gang
Hi Gang, On Wed, Jun 16, 2021 at 4:14 AM 赵刚 <zhaog6@lsec.cc.ac.cn> wrote:
Dear all,
I have submitted an python implementation of the sparse linear iterative algorithm "Transpose-Free QMR" (https://github.com/scipy/scipy/pull/14160). This algorithm (cited by 904) is very effective to solve non-Hermitian linear systems because it has a smoother convergence curve than the other algorithms like CGS, Bi-CGSTAB, and it has been widely used in those prominent scientific computing packages such as PETSc etc.
The purpose of adding the algorithm is to enrich SciPy's sparse linear library, so that users can have more/better choices when using Krylov subspace methods in SciPy to solve linear systems. I also hope that experts in this field can review the implementation of the algorithm (References are listed in "tfqmr.py"). Hope you like it.
This sounds like a great addition to SciPy. It looks like Stefan has already done a detailed review, which is very helpful. So it looks like we are close to merging your PR. Cheers, Ralf
participants (2)
-
Ralf Gommers -
赵刚