Floating Point Arithmetic for scientists

by Vincenzo Innocente (Fermi National Accelerator Lab. (US))

Friday 26 Jun 2026, 09:00 → 16:00 Europe/Zurich

593/R-010 - Salle 11 (CERN)

Abstract

Floating point arithmetic is at the hearth of any scientific computation. In this course I will introduce the basic concepts of floating point representation and arithmetic and discuss how precision can be preserved or disrupted. I will cover the essentials of elementary function evaluation and conclude discussing how to optimize computations without compromising the required accuracy. Practical examples and exercises will be proposed in both Python and C++. 

Lecturer Biography

Vincenzo Innocente has been responsible for simulation and reconstruction software in three generations of collider experiments at CERN.

In the CMS experiment at LHC he has been Chief Software Architect till year 2006 and then leader of the "Performance Task Force''. He led the Tracking Physics Object group from 2014 to 2018.

Since 2008 he is involved in R&D projects in the CERN Physics Department to adapt HEP software to run efficiently on new multicore computing platforms. HIs current interest is software acceleration and efficient massive parallelization.

He has pioneered machine learning In Particle Physics since 1990.

Source: https://www.linkedin.com/in/vincenzo-innocente-a615921/