search

LLNL / libROM

Model reduction library with an emphasis on large scale parallelism and linear subspace methods
https://www.librom.net
Other
198 stars 36 forks source link

Building libROM for TOSS4 #202

Closed debog closed 1 year ago

debog commented 1 year ago

Added the toolchain to build libROM on TOSS4 (LC/Quartz, etc). Also, needed to fix the script pulling MKL's version info since there is now column-aligning whitespace between #define vars and values in mkl_version.h

chldkdtn commented 1 year ago

@debog Thank you so much for doing this. By the way, would you mind changing the default toolchains in scripts/compile.sh to toss4, i.e., default-toss_4_x86_64_ib-librom-dev.cmake ? I tried to push this change directly to this branch, but it is complaining the authentication all of a sudden. Should I set some sort of personal access token?

debog commented 1 year ago

@chldkdtn Done! Re. authentication problem, after the TOSS4 update, I needed to delete the github line in my .ssh/known_hosts. Maybe you are having the same issue?

@debog Thank you so much for doing this. By the way, would you mind changing the default toolchains in scripts/compile.sh to toss4, i.e., default-toss_4_x86_64_ib-librom-dev.cmake ? I tried to push this change directly to this branch, but it is complaining the authentication all of a sudden. Should I set some sort of personal access token?

debog commented 1 year ago

Given the current state of the branch, the following should happen on Quartz:

chldkdtn commented 1 year ago

This works for me. The only thing is that all the regression tests fail, but this is expected. After this PR is merged to master branch, I expect all the regression tests will be successful again.

debog commented 1 year ago

@chldkdtn Is there a more rigorous way to make sure everything is okay? All my other codes are also facing this issue after the TOSS4 update. Of course, compiling with a new set of compilers (and especially, HDF5 library) introduces small differences, but it's nice to be sure.

Also, in the automated tests on github, I see that the CI/mac is failing, and the message says something about not finding a FORTRAN compiler. Any idea what's happening?

This works for me. The only thing is that all the regression tests fail, but this is expected. After this PR is merged to master branch, I expect all the regression tests will be successful again.

dylan-copeland commented 1 year ago

@debog I am manually running the regression tests on a non-toss4 machine. This should verify the code is correct, but as @chldkdtn mentioned, on toss4 we may just have to see what happens after merging this PR. The Mac CI has been broken for a long time, and we need someone to look into it.

chldkdtn commented 1 year ago

@debog I am manually running the regression tests on a non-toss4 machine. This should verify the code is correct, but as @chldkdtn mentioned, on toss4 we may just have to see what happens after merging this PR. The Mac CI has been broken for a long time, and we need someone to look into it.

@dylan-copeland after this PR is merged to the master branch, we should also check if LaghosROM compiles fine with TOSS4 update.

chldkdtn commented 1 year ago

@chldkdtn Is there a more rigorous way to make sure everything is okay? All my other codes are also facing this issue after the TOSS4 update. Of course, compiling with a new set of compilers (and especially, HDF5 library) introduces small differences, but it's nice to be sure.

Also, in the automated tests on github, I see that the CI/mac is failing, and the message says something about not finding a FORTRAN compiler. Any idea what's happening?

This works for me. The only thing is that all the regression tests fail, but this is expected. After this PR is merged to master branch, I expect all the regression tests will be successful again.

@debog if you can check your Hyper DMD results with this libROM update, that would be wonderful.

debog commented 1 year ago

@debog if you can check your Hyper DMD results with this libROM update, that would be wonderful.

@chldkdtn yes, I need to check the HyPar results. Right now, I am getting major regression test fails for HyPar (tests with and without libROM), so it will take me a couple of days to work my way through that.