Ascent build system based on LLVM clang#103
Conversation
4d65d9d to
11ebb48
Compare
|
There was a failing test on Ascent earlier. It passed when I re-ran, but that raises some flags for me. |
This example is nondeterministic due to algorithm used in cusolverRf library. We need to figure out how to set better pass/fail criteria for that test. Thanks for flagging it. |
This issue was fixed by @kswirydo in #109. @nkoukpaizan, thanks for drawing attention to this. |
pelesh
left a comment
There was a problem hiding this comment.
I've got a couple of minor suggestions to consider, but this looks good.
pelesh
left a comment
There was a problem hiding this comment.
Probably a good idea to rebase wrt develop since the bug that is tripping KLU+Rf+FGMRES test is fixed there. It is also okay to just squash and merge imho.
|
@ryandanehy or @nkoukpaizan or myself should be able to add a separate Ascent testing pipeline with this new compiler configuration. Might require some small re-arch, but should be doable. Can be addressed in another PR if not immediately doable. |
No need for that b/c it is getting out of service soon. |
cameronrutherford
left a comment
There was a problem hiding this comment.
LGTM - no need to test on Ascent since we are losing platform on Monday
|
@nkoukpaizan, could you rebase to |
293c367 to
4fcca01
Compare
This PR adds the
ascent-llvmCMake preset to build with LLVM clang on Ascent.Closes #13