search

CASL / Futility

VERA Fortran Utilities
Other
44 stars 20 forks source link

Fix build errors related to Native solvers w/o PETSc #254

Closed bkochuna closed 4 years ago

bkochuna commented 4 years ago

CASL Ticket # - 5855

bkochuna commented 4 years ago

Verified the build with MPI and no PETSc works.

@djabaay has been looking into some SERIAL_DEBUG test failures. I'm going to start digging into these too. If we identify the issues for that we can include them here in this PR.

aarograh commented 4 years ago

Brendan, thanks for pushing this up. I'll merge once the pipeline has passed. The issues with the timeouts can go in separately. This by itself gets an entire pipeline up and running that isn't right now.

bkochuna commented 4 years ago

@aarograh sounds good. Thanks.

stimpsonsg commented 4 years ago

So, this PR could go in as is (even though nothing in LinearSolvers needs this header explicitly. I'd prefer we fold it into the new PR that's created for round 2 of the Native GMRES stuff.

bkochuna commented 4 years ago

@stimpsonsg Yeah, I think that makes the most sense. Matt is still getting to the bottom of things. Most recent update is: Previously, these tests were being executed w/o a preconditioner, so some logic changed that caused these tests to run with the preconditioner. Agreement w/o the preconditioner was very good although not exact. He's tracking down the reason for this presently. The other thing we wish to investigate and resolve is if the preconditioner has problems with single pin cell cases as appeared to be evident from what was experienced here.

My best guess is that Matt should be able to get to the bottom of these issues today sometime today tomorrow.

@mattkab2 (so he sees this comment)