Axel Huebl

Results 723 issues of Axel Huebl

According to [this reddit post](https://www.reddit.com/r/cpp/comments/4k57tk/boost_mpl_dead/) the C++11 library [Brigand](https://github.com/edouarda/brigand) is the hot new sh*t to replace the quite inactive and in C++11 times inefficient + C++98 verbose Boost MPL library....

component: PMacc
refactoring

For ideal usage for ParaView, you could create a little helper file with all openPMD series: ``` note: needs a helper text file ending on .pmd containing a line with...

component: plugin
component: third party

Saw lockfiles being present while parallel HDF5 writes are active. Will now workaround a OpenMPI bug that uses file locking on Lassen (OLCF). https://github.com/open-mpi/ompi/issues/10053 Parallel HDF5 performance is still not...

component: documentation
component: openPMD
component: third party
machine / system

Handle the corner-case that we finalize a labframe simulation without having particles in that last dump. Otherwise `SetConstParticleRecordsEDPIC` will cause a ``` A Record can not be written without any...

bug
component: openPMD
bug: affects latest release
workaround

Print a performance hint if too many boxes are used per GPU. Close #3263 Seen recently with @joshualudwig8 - [ ] runtime test that this prints and is well-formatted

backend: cuda
Performance optimization
backend: hip
backend: dpc++

- [ ] for plotfile and openPMD diagnostics with BTD, we need to add CI analysis scripts that actually read the data to see its valid and in the range...

component: tests
component: diagnostics
component: boosted frame

The LAPACK++ build and the PSATD+RZ+CUDA build on Lassen (LLNL) seem to cause linker issues in the latest module version for @joshualudwig8. Update modules and instructions accordingly.

bug
backend: cuda
bug: affects latest release
machine / system

PICMI APIs and usage is, besides links to outside docs and examples, not really documented on the same level as the C++ app w/ inputs file usage. Let's change that....

component: documentation
component: PICMI
component: Python

We should print a performance hint (warning) if a user starts >>10 blocks per GPU.

backend: cuda
Performance optimization
warning
backend: hip
backend: dpc++

We found out that substepping was broken in 3D after #3013. We thus reverted it temporarily in .... (or we add an extra sync, as debugged by @NeilZaim :pray: )...

bug
component: mesh refinement
bug: affects latest release