rarias/jungle
rarias/jungle
2
0
Fork 2
Code Issues 52 Pull Requests 9 Actions Packages Projects Releases Wiki Activity

Fix mpich cross compilation and clean up tagaspi #213

Open
abonerib wants to merge 7 commits from abonerib/jungle:fix/cross-mpich into master
pull from: abonerib/jungle:fix/cross-mpich
merge into: rarias:master
Conversation 3 Commits 7 Files Changed 6 +50 -16
abonerib commented 2025-10-30 12:21:27 +01:00
Collaborator
Copy Link

mpich seems to work (although we lose fortran support when cross compiling). This also fixed its dependants:

  • cross.riscv64.osumb
  • cross.riscv64.sonar
  • cross.riscv64.tampi

tagaspi gpi-2 for tagaspi fails in configure phase still:

<...>
----------------------
   Checking for MPI
----------------------
checking for /nix/store/pjz8q4a8v10m43xm9d2vvbg48qca7pw9-mpi-all/include/mpi.h... configure: error: cannot check for file existence when cross compiling

Details in weasel:
http://weasel:3001/jobset/bscpkgs/fix-cross

mpich seems to work (although we lose fortran support when cross compiling). This also fixed its dependants: - `cross.riscv64.osumb` - `cross.riscv64.sonar` - `cross.riscv64.tampi` ~tagaspi~ gpi-2 for tagaspi fails in configure phase still: ``` <...> ---------------------- Checking for MPI ---------------------- checking for /nix/store/pjz8q4a8v10m43xm9d2vvbg48qca7pw9-mpi-all/include/mpi.h... configure: error: cannot check for file existence when cross compiling ``` Details in weasel: http://weasel:3001/jobset/bscpkgs/fix-cross
abonerib added the
mpi
 label 2025-10-30 12:21:27 +01:00
abonerib added 2 commits 2025-10-30 12:21:27 +01:00
Fix cross for tagaspi
All checks were successful
CI / build:cross (pull_request) Successful in 5s
Details
CI / build:all (pull_request) Successful in 18m8s
Details
89442e6af8
abonerib added the
cross
 label 2025-10-30 12:25:42 +01:00
rarias commented 2025-10-30 13:17:49 +01:00
Owner
Copy Link

Thanks!

mpich seems to work (although we lose fortran support when cross compiling). This also fixed its dependants:

  • cross.riscv64.osumb
  • cross.riscv64.sonar
  • cross.riscv64.tampi

Can we enable cross = true on those?

tagaspi fails in configure phase still:

<...>
----------------------
   Checking for MPI
----------------------
checking for /nix/store/pjz8q4a8v10m43xm9d2vvbg48qca7pw9-mpi-all/include/mpi.h... configure: error: cannot check for file existence when cross compiling

We may be able to patch it to follow what TAMPI is doing. Not sure if anybody maintains it anymore.

Thanks! > mpich seems to work (although we lose fortran support when cross compiling). This also fixed its dependants: > > - `cross.riscv64.osumb` > - `cross.riscv64.sonar` > - `cross.riscv64.tampi` Can we enable `cross = true` on those? > tagaspi fails in configure phase still: > ``` > <...> > ---------------------- > Checking for MPI > ---------------------- > checking for /nix/store/pjz8q4a8v10m43xm9d2vvbg48qca7pw9-mpi-all/include/mpi.h... configure: error: cannot check for file existence when cross compiling > ``` We may be able to patch it to follow what TAMPI is doing. Not sure if anybody maintains it anymore.
abonerib force-pushed fix/cross-mpich from 89442e6af8 to a17ad0417d 2025-10-30 15:11:34 +01:00 Compare
abonerib force-pushed fix/cross-mpich from a17ad0417d to ee7daa3922 2025-10-30 15:18:36 +01:00 Compare
abonerib force-pushed fix/cross-mpich from ee7daa3922 to a4e0c14e5c 2025-10-30 15:35:49 +01:00 Compare
abonerib changed title from WIP: Fix mpich and tagaspi cross compilation to Fix mpich cross compilation and clean up tagaspi 2025-10-30 16:51:04 +01:00
abonerib force-pushed fix/cross-mpich from a4e0c14e5c to 31302d4a97 2025-10-30 17:00:03 +01:00 Compare
abonerib force-pushed fix/cross-mpich from 31302d4a97 to 81fa5392bc 2025-10-30 17:23:41 +01:00 Compare
abonerib force-pushed fix/cross-mpich from 81fa5392bc to 00e7bafb6b 2025-10-30 17:40:56 +01:00 Compare
abonerib force-pushed fix/cross-mpich from 00e7bafb6b to 8f96735e85 2025-10-30 18:08:54 +01:00 Compare
abonerib commented 2025-10-30 18:22:13 +01:00
Author
Collaborator
Copy Link

osu micro bencharks breaks since it uses:

configureFlags = [
      "CC=mpicc"
      "CXX=mpicxx"
];

With the patch to do CC=${CC:-gcc} mpicc references itself.

EDIT: I'm going to leave it as CC=gcc for now.

osu micro bencharks breaks since it uses: ``` configureFlags = [ "CC=mpicc" "CXX=mpicxx" ]; ``` With the patch to do `CC=${CC:-gcc}` mpicc references itself. EDIT: I'm going to leave it as `CC=gcc` for now.
👍 1
abonerib force-pushed fix/cross-mpich from 8f96735e85 to ad4d3f6719 2025-10-30 18:33:28 +01:00 Compare
abonerib added 2 commits 2025-10-31 11:22:21 +01:00
Fix osu cross-compilation
All checks were successful
CI / build:cross (pull_request) Successful in 13m47s
Details
CI / build:all (pull_request) Successful in 20m1s
Details
4778cb3b5e
abonerib reviewed 2025-10-31 11:26:55 +01:00
pkgs/mpich/default.nix Outdated
@ -35,0 +49,4 @@
sed -i 's:^CC=.*:CC=${targetPackages.stdenv.cc}/bin/${targetPackages.stdenv.cc.targetPrefix}cc:' $out/bin/mpicc
sed -i 's:^CXX=.*:CXX=${targetPackages.stdenv.cc}/bin/${targetPackages.stdenv.cc.targetPrefix}++:' $out/bin/mpicxx
'' + lib.optionalString enableFortran ''
sed -i 's:^FC=.*:FC=${targetPackages.gfortran or gfortran}/bin/${targetPackages.gfortran.targetPrefix or gfortran.targetPrefix}gfortran:' $out/bin/mpifort
abonerib commented 2025-10-31 11:26:55 +01:00
Author
Collaborator
Copy Link

I noticed openmpi nixpkgs they use targetPacages.gfortran or gfortran daf6dc47aa/pkgs/by-name/op/openmpi/package.nix (L189)

Not sure if we should also do that in our openmpi.

I noticed openmpi nixpkgs they use `targetPacages.gfortran or gfortran` https://github.com/NixOS/nixpkgs/blob/daf6dc47aa4b44791372d6139ab7b25269184d55/pkgs/by-name/op/openmpi/package.nix#L189 Not sure if we should also do that in our openmpi.
rarias commented 2025-10-31 13:24:24 +01:00
Owner
Copy Link

Currently we don't use the openmpi package of jungle, so I wouldn't change it.

Currently we don't use the openmpi package of jungle, so I wouldn't change it.
abonerib force-pushed fix/cross-mpich from 4778cb3b5e to 12abc41e48 2025-10-31 11:44:44 +01:00 Compare
abonerib requested review from rarias 2025-10-31 11:45:10 +01:00
abonerib force-pushed fix/cross-mpich from 12abc41e48 to 100e4e7029 2025-10-31 11:54:09 +01:00 Compare
abonerib force-pushed fix/cross-mpich from 100e4e7029 to 4752afae0c 2025-10-31 11:57:18 +01:00 Compare
abonerib force-pushed fix/cross-mpich from 4752afae0c to 32a2b87313 2025-10-31 12:22:54 +01:00 Compare
rarias reviewed 2025-10-31 13:25:39 +01:00
rarias left a comment
Owner
Copy Link

Thanks!

Thanks!
pkgs/gpi-2/default.nix Outdated
@ -60,5 +70,6 @@ stdenv.mkDerivation rec {
maintainers = with lib.maintainers.bsc; [ rarias ];
platforms = lib.platforms.linux;
license = lib.licenses.gpl3Plus;
cross = false;
rarias commented 2025-10-30 17:01:05 +01:00
Owner
Copy Link

Can we add a comment here on why is broken? Maybe also in the --with-mpi=yes above.

Can we add a comment here on why is broken? Maybe also in the `--with-mpi=yes` above.
All checks were successful
CI / build:cross (pull_request) Successful in 12s
Details
CI / build:all (pull_request) Successful in 15s
Details
This pull request can be merged automatically.
You are not authorized to merge this pull request.
View command line instructions.

Checkout

From your project repository, check out a new branch and test the changes.
git fetch -u fix/cross-mpich:abonerib-fix/cross-mpich
git checkout abonerib-fix/cross-mpich
Sign in to join this conversation.
Reviewers
No Reviewers
rarias
Labels
Clear labels
bug
ceph
ci
config
cross
hw
io
kernel
mpi
net
nix
ops
repair
slurm
No Label
cross
mpi
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
abonerib arocanon varcila rarias
No Assignees
2 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: rarias/jungle#213
No description provided.
Delete Branch "abonerib/jungle:fix/cross-mpich"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?