dbcsr issues

Document cuda profiling feature

2

Distribution defaults

I open this issue to discuss how the distribution of DBCSR matrices/tensors should be handled. If possible, DBCSR should provide a default distribution that gives reasonably good load balance for...

pseewald

Hit acc_devmem_setzero: failed in dbcsr/acc/dbcsr_acc_devmem.F:445

40

**Describe the bug** Hit acc_devmem_setzero: failed in dbcsr/acc/dbcsr_acc_devmem.F:445 **To Reproduce** Steps to reproduce the behavior: git clone --recursive https://github.com/cp2k/dbcsr.git commit id: f35f901e4460980aa06757294463a1e6308f8dc9 cd dbcsr; mkdir build; cd build; cmake -DTEST_MPI_RANKS=2...

vitesselin

bug

Explicitly reset the active GPU device

3

Currently, DBCSR assumes that the active GPU device never changes after `dbcsr_init_lib()` has been called. With the arrival of more GPU accelerated libraries in CP2K this assumption will likely break....

oschuett

GPU test coverage

It would be good to consider coverage of GPU tests too (currently we do only on travis). Assuming that we can do on daint, I see that 1. ninja and...

alazzaro

enhancement

discussion needed

Thread number has changed

21

When building and running CP2K as POPT variant (any workload), DBCSR complains: ```text ******************************************************************************* * ___ * * / \ * * [ABORT] * * \___/ Thread number has changed...

hfp

OpenMP/offload backend

2

OpenMP 4.0 introduced offloading to attached devices and made extensions in OpenMP 4.5 (future versions of OpenMP will likely only evolve the offload capability with more advanced features on top...

hfp