| الوصف: |
The goal of this deliverable is to report the status of work related to the efficient deployment of the SCALABLE codes on the peta-scale and the pre-exascale systems. The supercomputing centers IT4I and FZJ are continuously deploying the newly developed versions of WaLBerla and ProLB codes on their peta-scale systems throughout the lifetime of the project. They also provide support for optimal code and tool compilation including environment setup for (i) particular CPU architectures, (ii) GPU accelerators (as of now only Nvidia GPUs), (iii) network topology and communication libraries (MPI, UCX, GPU-Direct, etc.). This work is mainly needed by both WP2 and WP4. In addition, in collaboration with local infrastructure support teams, the IT4I team enables the hardware parameter tuning and energy consumption monitoring on their both major production systems (Barbora and newly also Karolina) which is essential for work in WP2. On the Karolina system, we are now able to tune both CPU and GPU parameters as well as measure their energy consumption. These features will be used in D2.3. As of now, WP6 secured access to 4 peta-scale systems, that are used by all WPs dealing with code development and optimization. These systems are: Barbora (IT4I), Karolina (IT4I), JUWELS Cluster (FZJ) and JUWELS Booster (FZJ). We have used several mechanisms including IT4I Open-access, PRACE Preparatory access, EuroHPC code development access and GCS Compute Access. Finally, there is only one EuroHPC pre-exascale machine that is partially available to the general public at the time of submission of this report. This is the LUMI machine and, the nonaccelerated LUMI-C partition to which access has been secured as well. |
