Preliminary Performance Evaluation of the Fujitsu A64FX Using HPC Applications

RIKEN Center for Computational Science has been installing the supercomputer Fugaku. The Fujitsu A64FX, based on the Armv8.2-A+SVE architecture, is used in the system. In this paper, we evaluated the seven HPC applications and benchmarks on the A64FX. In a performance comparison with Marvell (Cavium) ThunderX2 processor and Intel Xeon Skylake processor, the A64FX achieved higher performance in a memory bandwidth-intensive application thanks to its high memory bandwidth. However, we confirmed that the performance of the A64FX decreased from a lack of out-of-order resources. To mitigate this problem, the “loop fission” function of the Fujitsu compiler was used to improve the performance.

I. Introduction

The supercomputer Fugaku has been under installation at RIKEN Center for Computational Science as the successor system of the supercomputer K, and it is equipped with the A64FX processor [1] developed through a co-design between Fujitsu and RIKEN. The A64FX is a general-purpose processor based on the Armv8.2-A architecture and has a Scalable Vector Extension (SVE) and High Bandwidth Memory 2 (HBM2). At ISC20, the Fugaku system with 396 racks achieved 415.53 PFLOPS in TOP500, and the one with 360 racks achieved 13366.40 TFLOPS in HPCG.