# Numerical study of floating point precision within the PEPPER event generator

- Institution: CERN
- Speaker: Yucheng Zhang
- Year: 2026
- Language: English
- Source: [CERN](https://videos.cern.ch/record/3026681)
- Licence: [CERN](https://copyright.web.cern)
- Who can watch: Teens profiles and older (sources give no ratings; placed by Invisible Network from the source's own words)

**Project Description**: This project investigates whether PEPPER can safely use reduced floating-point precision to accelerate parton-level event generation on modern CPUs and GPUs. By comparing FP32 and FP64 calculations with CADNA, it identifies where rounding errors affect matrix elements and event weights, supporting the development of a reliable mixed-precision approach for HL-LHC simulations.

**Supervisor**: Dr. Enrico Bothmann & František Stloukal

00:00:00 Slide 1

00:00:15 Slide 2

00:00:59 Slide 3

00:01:34 Slide 4

00:02:30 Slide 5

00:03:25 Slide 6

00:04:06 Slide 7

00:04:46 Slide 8

00:05:22 Slide 9

00:05:54 Slide 10

00:06:07 Slide 11

00:06:37 Slide 12

## Lectures

1. [Numerical study of floating point precision within the PEPPER event generator](https://invisiblenetwork.ai/learn/cern/3026681/e64be576adfd), 7 min (CERN's player)