dbcsr
dbcsr copied to clipboard
cp2k
DBCSR: Distributed Block Compressed Sparse Row matrix library
A few things in the [tensor API](https://github.com/cp2k/dbcsr/blob/develop/src/tensors/dbcsr_tensor_api.F) should be improved: - [ ] document preconditions (e.g. `dbcsr_t_contract` requires that tensors have the same block sizes but `dbcsr_t_copy` does not) -...
File to add contributing-to-DBCSR.md, explanation of the build system (and how to hack it), and structure of the directories.
I ran into this issue when testing cp2k/cp2k#581 which is mainly based on DBCSR tensors. After contraction of large tensors, DBCSR multiplication of much smaller matrices takes much more time...
The ROCm build tests currently uses an image from @shoshijak, we should migrate that to our Docker hub and rebuild/update the image periodically like we do with the others. Note:...
There are two potential issues with DBCSR's init/fini flow wrt ACC interface and LIBSMM: 1. DBCSR calls `acc_init()`, which eventually calls `libsmm_acc_init()` internally (this is a potential issue since the...
Performance regression: old: https://object.cscs.ch/v1/AUTH_40b5d92b316940098ceb15cf46fb815e/dbcsr-artifacts/logs/build-679/gnu.test.out new: https://object.cscs.ch/v1/AUTH_40b5d92b316940098ceb15cf46fb815e/dbcsr-artifacts/logs/build-689/gnu.test.out Need to check side by side on the same job.
until https://github.com/pmodels/mpich/issues/4300 is solved we need to add `-fallow-argument-mismatch` when building our MPI wrapper. The most portable and clean solution would be to `try_compile` a snippet triggering the issue when...
CP2K has a top-leve `COPYRIGHT` file that clearly states the years the CP2K Developers Group assert on the copyright. The standard copyright boilerplate for the DBCSR source files does not...
This might be nitpicking, but I noticed the copyright boilerplate at the top of the source files mentions the "DBCSR developers group" everywhere, while the `README.md` tells users to cite...