-
Notifications
You must be signed in to change notification settings - Fork 45
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Using a custom MPI library #534
Comments
this is required to fix TriBITSPub/TriBITS#534
@bartlettroscoe I might need your help with this too... as you point out it breaks normal Trilinos builds: trilinos/Trilinos#11175 (comment) Is there a way that the TriBITS MPI TPL can be set up to work both with the usual MPI compiler wrappers and with a specified include dir and library? Right now I'm running into this line which erases my
|
actually... maybe that file is different in my copy of SEACAS and latest TriBITS... |
Nope, even with the latest snapshot of TriBITS this doesn't work |
The simple fix would be to change that line to:
Can you try that and see if it fixes your use case? It should be the case that if |
Thanks @bartlettroscoe ! I'm trying this now |
Yes! This is working for my use case! Thank you! |
This reverts commit e49070d29fdd8c1d525bb22c52d7b2745271d7e2. This causes configure failures.
…s:develop' (9dd9187). * trilinos-develop: MueLu: reenable libmuelu-adapters.a small improvements (TriBITSPub/TriBITS#533) Ctest: lightsaber update Sacado: Fix memory leaks in Rad when using DFad nested scalar type. Netcdf modern is only used if HDF5 is modern Tacho : use same max vector length for factor & update Tacho : query max vector / team sizes for factor Tacho : cleaner way to avoid "out-of-bound access" fix modern Netcdf import (TriBITSPub/TriBITS#533) fix modern HDF5 import (TriBITSPub/TriBITS#533) Tacho : to avoid "requested too large team size." Make Netcdf depend on HDF5, fix TriBITS find modules (TriBITSPub/TriBITS#533) Revert "add REQUIRED_HEADERS for TriBITS MPI TPL" (TriBITSPub/TriBITS#534) Don't wipe out TPL_<tplName>_INCLUDE_DIRS if set by user (TriBITSPub/TriBITS#534) Tacho : fix issue with LDLt Add include guards for *_def.hpp file (trilinos#10507) STK: Snapshot 10-21-22 11:03 Netcdf_ENABLE_MODERN mode for TriBITS Netcdf TPL add REQUIRED_HEADERS for TriBITS MPI TPL Avoid double instantiation of CrsMatrix< SCALAR , LO , GO , NODE >::convert< SCALAR >
…s:develop' (9dd9187). * trilinos-develop: (26 commits) updating bug_report.md template instructions Stokhos: Fixes for compiling on HIP MueLu: reenable libmuelu-adapters.a FastILU : use mirror view instead of host view small improvements (TriBITSPub/TriBITS#533) MueLu: Add AT PR script Ctest: lightsaber update Sacado: Fix memory leaks in Rad when using DFad nested scalar type. Netcdf modern is only used if HDF5 is modern Tacho : use same max vector length for factor & update Tacho : query max vector / team sizes for factor Tacho : cleaner way to avoid "out-of-bound access" fix modern Netcdf import (TriBITSPub/TriBITS#533) fix modern HDF5 import (TriBITSPub/TriBITS#533) Tacho : to avoid "requested too large team size." Make Netcdf depend on HDF5, fix TriBITS find modules (TriBITSPub/TriBITS#533) FROSch : remove compiler warnings "unreachable code" FROSch : add an option to call Metis before ILU Revert "add REQUIRED_HEADERS for TriBITS MPI TPL" (TriBITSPub/TriBITS#534) Don't wipe out TPL_<tplName>_INCLUDE_DIRS if set by user (TriBITSPub/TriBITS#534) ...
I've found a solution for this issue that I will outline below, but I thought it was appropriate to describe here as its own GitHub issue. I use my own implementation of MPI for one rank with TriBITS-based projects: https://github.com/sandialabs/fbmpi.
When using TriBITS' MPI TPL, I couldn't specify my own
TPL_MPI_INCLUDE_DIRS
unless the call totribits_tpl_find_include_dirs_and_libraries
had aREQUIRED_HEADERS
argument. We can provide this withmpi.h
, because that is what all MPI codes include. Once that modification is made, I was able to specify the following at configuration time:This issue can form a useful anchor for related branches and pull requests with this change
The text was updated successfully, but these errors were encountered: