[petsc-dev] ResQPASS in Petsc

[Jed Brown]

Matthew Knepley <knepley at gmail.com> writes:

>> We are developing novel subspace methods to deal with inequality bounds on
>> the variables.
>> ResQPASS, short for Residual Quadratic programming active set Subspace
>> method,  behaves asymptotically as a Krylov subspace.  The paper is
>> published in SISC and is available in https://urldefense.us/v3/__https://arxiv.org/abs/2302.13616__;!!G_uCfscf7eWS!YmnwRAx_eHfCG2V2WlrTySK1w_eSwS7AWzqhbyF8TX1KlhOBuYKbPTTnNY1Z7sYdVgQj9V1rS6hZig316hL9$ 
>> <https://urldefense.us/v3/__https://arxiv.org/abs/2302.13616__;!!G_uCfscf7eWS!bnM-6GCLUYnx_bcmKoE5yD6ehhfMLkUeEB7g_LRq0OMhvkWuwolEBu_gw8jdqo-VxP43j8InK6Eb6X6U9g$>
>> .
>>
>> In the past we contributed several pipelined methods in KSP.    But the
>> current methods might also be of value for large scale optimization
>> problems.  We believe that the method will be very useful for contact
>> problems and variational inequalities that model interacting mechanical
>> systems.
>>
>> Who should I contact to discuss the best strategy to add the solver?
>>
>
> That sounds great! I am currently working on variational inequalities for
> frictional contact (for earthquake systems). I would be happy to help you
> integrate these solvers. We are excited to test different formulations (the
> LVPP people are also talking to us). Let me know how I can help.

I agree this is useful. You can create an issue in the repository or make a new topic on the PETSc discord.

https://urldefense.us/v3/__https://gitlab.com/petsc/petsc/-/issues/new__;!!G_uCfscf7eWS!Y26oQ0Kg5KC-FFbyyj95vOMLAKbJ7o9AUhZb_3PtRekzy3HGrleNytFKCkG1C4BrBLiA9z6BdwZYmJiqVX4$