Tobias Ribizel

The OpenMP part is not yet parallelized, and the skeleton tree computation is not enabled by default.

Sorry for the delay - not entirely: After fixing the typo in `satisfies`, I got the following error ``` ==> Error: ProcessError: Command exited with status 1: '/storage/apps/opt/spack/opt/spack/linux-rocky9-x86_64_v3/gcc-11.4.1/meson-1.5.1-dbnokczsbh623ysq7xalh6mqekszktz5/bin/meson' 'setup' '/tmp/tribizel/spack-stage/spack-stage-mesa-main-zpguogmpqbttsmxot6z537z2mj2cdaed/spack-src'...

Does it need to happen in a single function call though? Can't we ask the user to call `fill_zero` instead? `fill_mode` introduces a fundamentally new concept we haven't used so...

Why does it need to be done in a single step? Can't it be two steps (construction + fill)? Can you describe a matrix assembly scenario where the number of...

I know the common approach of assembling elements independently (which is what I built sum_duplicates for), but what I'm not familiar with is the case where part of the preallocated...

The question there is: Is the resulting row/column then zero itself? Because that would make the matrix singular, and maybe bring us back to #842 if we want to use...

I see, that makes sense. But that would not work in a more generic grid, where the indexing isn't as simple. But yes, let's not get too much off track....

The take-away from the discussion in today's meeting: * The safety of pre-initialized data should outweigh the performance advantage of not initializing data, so the default case should be initializing...

Thanks for the report! Can you please add some more information about your build environment (ROCm version, rocThrust version, Ginkgo commit hash, CMake configuration and CMakeCache.txt file)? Otherwise we will...

We are using a semi-linear merge workflow, so instead of merging the changes from `develop` into this branch, please rebase onto `develop` so your branch is just a linear sequence...