Many of today’s applications running on large HPC systems are codes, that have been in development for more than two (or even three) decades, programmed partially in “old fashioned” languages, such as Fortran, and which were not designed for scale and to parallelise for system architectures as we know them today in HPC. We refer to these codes as “legacy codes”.
In the advent of a new generation of HPC systems, the question is whether these codes are still relevant to their owners, and if yes, how these codes can be rewritten to run efficiently on the future HPC systems, and what effort this would require. This document casts a light on the European portfolio of legacy codes through the analysis of a survey that has been distributed amongst European code developers.
Author: G. Colin de Verdière, CEA