The adoption of ARM processor architectures is on the rise in the HPC ecosystem. Fugaku supercomputer is a homogeneous ARM-based machine, and is one among the most powerful machine in the world. In the programming world, dependent task-based programming models are gaining tractions due to their many advantages: dynamic load balancing, implicit expression of communication/computation overlap, early-bird communication posting,...MPI and OpenMP are two widespreads programming standards that make possible task-based programming at a distributed memory level. Despite its many advantages, mixed-use of the standard programming models using dependent tasks is still under-evaluated on large-scale machines. In this paper, we provide an overview on mixing OpenMP dependent tasking model with MPI with the state-of-the-art software stack (GCC-13, Clang17, MPC-OMP). We provide the level of performances to expect by porting applications to such mixed-use of the standard on the Fugaku supercomputers, using two benchmarks (Cholesky, HPCCG) and a proxy-application (LULESH). We show that software stack, resource binding and communication progression mechanisms are factors that have a significant impact on performance. On distributed applications, performances reaches up to 80% of effiency for task-based applications like HPCCG. We also point-out a few areas of improvements in OpenMP runtimes.
X-Aevol is the GPU port of the Aevol model, a bio-inspired genetic algorithm designed to study the evolution of micro-organisms and its effects on their genome structure. This model is used for in-silico experimental evolution that requires the computation of populations of thousands of individuals during tens of millions of generations. As the model is extended with new features and experiments are conducted with larger populations, computational time becomes prohibitive. X-Aevol is a response to the need of more computational power. It was designed to leverage the massive parallelization capabilities of GPU. As Aevol exposes an irregular and dynamic computational pattern, it was not a straightforward process to adapt it for massively parallel architectures. In this paper, we present how we have adapted the Aevol underlying algorithms to GPU architectures. We implement our new algorithms with CUDA programming language and test them on a representative benchmark of Aevol workloads. To conclude, we present our performance evaluation on NVIDIA Tesla V100 and A100. We show how we reach a speed-up of 1,000 over a sequential execution on a CPU and the speed-up gain up to 50% from using the newer Ampere micro-architecture in comparison with Volta one.