Closed kostrzewa closed 7 years ago
Hi @kostrzewa ! Thanks for doing this. Are you sure you want to merge to 'master'? I thought we were still on tweaking 'devel'. As you see Travis builds are broken ATM -- I think this is related to Python/Jinja. I have tried to debug this but have been unsuccessful.
On the issue about other ND stuff, I have been working on the part about having QPhiX reconstruct the full solutions from the preconditioned ones. This made me put some new functions into linearOp. I have added stubs for these into the twisted mass ops too (tho they do nothing). Also, probably these work only for the solvers that solve with M rather than M^dag M where the source prep/solution may involve a multiply with M^\dagger or M. Still you may want to get those changes....
Yes, I wanted to merge this into master since it was an oversight on my part to leave Peter and Martin out of the list.
I will be able to do cleanup and prepare the merge of the ND stuff into devel in the coming two weeks as other things have quieted down somewhat now.
With regard to reconstruction Peter and Martin had also worked on something. This took the form of abstract operator classes which provided source preparation and solution reconstruction (independent of whether the operator did or did not have a clover term). From our side, this has now been somewhat superceded by code within tmLQCD which does this instead, since we figured out all the even-odd packing routines.
Added Peter, Martin and Bartek to contributors. Some of the things here will only be relevant once I have merged all the ND stuff, but since I had to touch the file anyway to add Peter I also included those contributions.
@martin-ueding I didn't know if you wanted your e-mail here.