BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20151119T193000Z
DTEND:20151119T200000Z
LOCATION:18AB
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: Measuring and reporting performance of parallel computers constitutes=0Athe basis for scientific advancement of high-performance computing=0A(HPC). Most scientific reports show performance improvements of new=0Atechniques and are thus obliged to ensure reproducibility or at least=0Ainterpretability. Our investigation of a stratified sample of 120 papers=0Aacross three top conferences in the field shows that the state of the =0Apractice is not sufficient. For example, it is often unclear if reported=0Aimprovements are in the noise or observed by chance. In addition to=0Adistilling best practices from existing work, we propose statistically=0Asound analysis and reporting techniques and simple guidelines for=0Aexperimental design in parallel computing. We aim to improve the =0Astandards of reporting research results and initiate a discussion in the=0AHPC field. A wide adoption of this minimal set of rules will lead to=0Abetter reproducibility and interpretability of performance results and=0Aimprove the scientific culture around HPC.
SUMMARY:Scientific Benchmarking of Parallel Computing Systems
PRIORITY:3
END:VEVENT
END:VCALENDAR