Alfio Lazzaro

I have no clue... I can suggest to re-read my suggestions at https://github.com/cp2k/cp2k/issues/546#issuecomment-612525643 , especially to print how many sub-communications we are creating. Another solution is to try another MPI...

The error is at line 775 when it is freeing the communicators? I wonder if there is a cumulative limit on that... On the other hand, maybe we can think...

> I'm currently working with matrices where >99% of the values of 0, and it's very unlikely most of the blocks will be larger than 1x1 in size. However, the...

> Anyway, I think that answers the question in this issue: libsmm_acc does not optimize 1x1 blocks, nor does any other part of the library, since that's outside the scope...

@hfp for my understanding: - We don't see the issue since we NDEBUG is defined? But we really don't define it in CP2K tests... - We really never call printf...

(BTW, trying to recover the Daint-CI output...)

CSCS CI seems broken on their side: ``` + sbatch --wait --time=0:15:00 --account=g90 --job-name=DBCSR.gnu.build --output=sbatch.jenkins-g90-DBCSR-1116.gnu.build.out .ci/daint.cscs.ch/gnu.build.sh sbatch: error: Batch job submission failed: Invalid account or account/partition combination specified java.nio.file.NoSuchFileException: /users/jenkg90/workspace/g90/DBCSR/sbatch.jenkins-g90-DBCSR-1116.gnu.build.out...

ACK.

Sorry for the slow response, I will see if I have time to review it in June... In any case, DBCSR is tested in CP2K, where the machinery is already...

@LecrisUT sorry, I was not clear in my message... As I said, I will take care of your changes. We do not need to follow any CP2K repo rule here...