Blog
On the technical debt of high-performance scientific software
High-performance scientific software must overcome two specific challenges: scientific validation, and performance on bleeding-edge and short-lived hardware. Success in each requires time, expertise, and cumulative experience over many failed attempts. Thus, engineering software developers are mostly experts in physical modelling or in high-performance computing (HPC) and rarely experts in the management of technical debt.
Key points and take-outs from the FORTRAN webinar organised by EXCELLERAT
The EXCELLERAT Centre of Excellence held a Webinar with Wadud Miah on the topic “Fortran for High Performance Computing” on the 4th of May 2021. Following this event, CERFACS created a COOP blog article with the objective to further share the answers to the questions discussed during the webinar with the HPC ecosystem.
EXCELLERAT’s Fourth All-Hands Meeting
From 30 November to 1 December 2021, our project meeting was once again hosted as a digital event due to the continuing pandemic situation. About 30 partner representatives per day presented the EXCELLERAT project tasks’ good progress, some minor 2021 challenges, and the requirements for the project’s extension and great final results.
EXCELLERAT Exploitation Strategy Seminar with Horizon Results Booster for Vistle, Alya and Uqit
In collaboration with experts from Horizon Results Booster (HRB), EXCELLERAT has completed its first internal Exploitation Strategy Seminar for three selected Key Exploitable Results (KERs): Vistle, Alya and UQit. After an introduction by Dr. Péter Mogyorósi, the HRB expert, partners presented and discussed each KER and its exploitation strategy with HRB experts.
How can SMEs benefit from using High Performance Computing (HPC) Cloud
European growth is unthinkable without its 23 million small and medium-sized enterprises (SMEs). Representing 99.8 percent of European companies, they form the foundation for innovation, competition, and jobs. It’s important to support these SMEs in sustainable development through innovation, processes optimisation, and increased competitiveness: e.g. through the software and data processing support of HPC.
Road and Scouts: a versioning strategy for large Research software.
Versioning of large research software applications for industry can be sloppy because audiences have different needs. End-users require a clear official version, while research asks for many exploratory versions.
Born in the 1990s, AVBP is combustion software used and developed today by dozens of industrial users and researchers simultaneously. This post will consider whether it is suitable for a semantic versioning strategy.
EXCELLERAT’s Third Virtual All-Hands Meeting
From 8 to 9 June 2021, we held our 3rd virtual project meeting of the continuing pandemic. Over 30 partner representatives on each day presented the EXCELLERAT project tasks’ good progress, some early 2021 challenges, and the needs for a continuous evolution to master the final project months (including its three month extension).
EXCELLERAT Training Workshop: Using Machine Learning To Analyse Engineering Data
EXCELLERAT’s latest workshop dealt with data analysis for engineering data, particularly simulation data. Partners from Fraunhofer SCAI conducted this one-day online training course with sessions dedicated to the topics of Clustering, Dimensionality Reduction, Pre-processing of Simulation Data, and two hands-on tutorials with exemplary use-cases from car crash analysis and aeroacoustics.
White Paper – Monitoring High-Performance Computing at scale
There is still a long way to go before using Tier-0 High Performance Computers for engineering design every day. Moving from a handful of demonstration runs to the desired mass production will require a special pair of glasses to get insights on how end-users actually use each software on each machine.
The use of in-situ analysis tools by EXCELLERAT
Paraview-Catalyst is widely used to perform in-situ visualisation, as in the case of OpenFOAM and Code-Saturne. However, one needs An appropriate version of Paraview-Catalyst on the computer system, the code must be instrumented to transfer its data towards Catalyst, and one usually needs a Python script automatically produced by Paraview.