Runtime mechanisms to survive new HPC architectures: A use case in human respiratory simulations

Abstract

Computational fluid and particle dynamics (CFPD) simulations are of paramount importance for studying and improving drug effectiveness. Computational requirements of CFPD codes demand high-performance computing (HPC) resources. For these reasons, we introduce and evaluate in this article system software techniques for improving performance and tolerating load imbalance on a state-of-the-art production CFPD code. We demonstrate benefits of these techniques on Intel-, IBM- and Arm-based HPC technologies ranked in the Top500 supercomputers, showing the importance of using mechanisms applied at runtime to improve the performance independently of the underlying architecture. We run a real CFPD simulation of particle tracking on the human respiratory system, showing performance improvements of up to 2×, across different architectures, while applying runtime techniques and keeping constant the computational resources.

Performance study of HPC applications on an Arm-based cluster using a generic efficiency model

“…Stanisic et al in [4] conduct a performance analysis studies of HPC applications, but focusing on mobile Arm-based CPU. Garcia-Gasulla et al in [5] also evaluate a CFD code on an Arm-based cluster however their focus is on portability of performance across multiple architectures. We complement this work providing an evaluation with a larger set of HPC applications and we provide an analysis method that is mostly architecture independent.…”

“…CP2K is a framework for atomistic simulations written in Fortran and is part of the Unified European Applications Benchmark Suite 5 .…”

On the Benefits of Tasking with OpenMP

“…However, both of these solutions add complexity either to the data structure or to the directive readability, so this is not included in the version evaluated here. We expect support for multidependences [8,17] in OpenMP 5.0 to help with the directive readability issue (we must restrict this effort to OpenMP 4.5 features due to current compiler support). Stencil computations are taskified with an inout dependence on the block they operate on, and therefore depend on the previous communication tasks that write to that block.…”

The Cooperative Parallel: A Discussion About Run-Time Schedulers for Nested Parallelism

“…They can be found in a variety of examples from earth science, where some processes simulate the earth while other the ocean, to biological ones. This section describes the couple run applied to a biological simulation of the human respiratory system [18]. It is composed of the simulation of the air going through the human airways, and the simulation of the transport of particles inhaled through the bronchopulmonary tree.…”