Floating Point Computation Basics

par Dr Vincent LAFAGE (IJCLab)

vendredi 7 avr. 2023, 09:00 → 12:30 Europe/Paris

200/1-101 - Salle 101 (IJCLab)

Pitfalls of floating point computation (float, double): theoretical basis and case studies in high energy physics

1. the quadratic equation in kinematics calculations
2. variance calculations in data analysis
3. calculations with complex numbers
4. accurate summation in large Monte-Carlo calculations
5. precision in matrix and geometry calculations (the interest of factoring)
6. differential equations. Elements of numerical analysis.
7. solution of equations, minimisation
8. scaling: combining accuracy and efficiency
9. function evaluation: series and approximants

Clean code for computation

Prerequisites

• General Unix/Linux shell handling (filesystem, make…).
• Basics of C or C++ or Fortran. Python welcome but not used in the lecture.
• Laptop that can connect to a Linux server over SSH.

Registration

While it is now possible for all lab members to register via Indico, note that this course was initially proposed as a doctoral school training and PhD students registering via ADUM will accordingly get priority for the initial run. If there is too much demand, the course will be run again later in the year.

Precision2_eng.pdf