ICL Research Profile

Overview

The High Performance LINPACK (HPL) benchmark solves a dense linear system in double precision (64-bit arithmetic on distributed-memory computers. HPL is written in a portable ANSI C and requires an MPI implementation and either BLAS or Vector Signal and Image Processing Library (VSIPL). HPL is often the first program to run on large HPC machines. Carefully optimized versions of HPL are available from major HPC hardware vendors.

The primary focus of HPL 2.3, released in 2018, was to improve the accuracy of reported benchmark results and ensure easier configuration and building on modern HPC platforms. HPL is now hosted on GitHub and features more detailed reporting of the solution’s scaled residual and the achieved performance number. Another addition is a software configuration tool based on GNU Autotools and the removal of deprecated MPI functions. The LINPACK app for iOS reached over 8 Gflop/s on the iPhone X. For the November 2021 TOP500 list, an optimized version of the HPL code achieved over 440 Pflop/s on the Fugaku supercomputer at RIKEN, Japan and in June 2022 achieved 1.1 Eflop/s on Frontier at Oak Ridge National Laboratory.