Studying the potential of Graphcore IPUs for applications in Particle Physics
arXiv:2008.09210
Abstract
This paper presents the first study of Graphcore's Intelligence Processing Unit (IPU) in the context of particle physics applications. The IPU is a new type of processor optimised for machine learning. Comparisons are made for neural-network-based event simulation, multiple-scattering correction, and flavour tagging, implemented on IPUs, GPUs and CPUs, using a variety of neural network architectures and hyperparameters. Additionally, a Kálmán filter for track reconstruction is implemented on IPUs and GPUs. The results indicate that IPUs hold considerable promise in addressing the rapidly increasing compute needs in particle physics.
References in corpus (15)
- Self-Attention Generative Adversarial Networks
- Progressive Growing of GANs for Improved Quality, Stability, and Variation
- LHCb Detector Performance
- cuDNN: Efficient Primitives for Deep Learning
- Efficient, reliable and fast high-level triggering using a bonsai boosted decision tree
- How to GAN LHC Events
- Getting High: High Fidelity Simulation of High Granularity Calorimeters with High Speed
- Fast and accurate simulation of particle detectors using generative adversarial networks
- Allen: A high level trigger on GPUs for LHCb
- Dissecting the Graphcore IPU Architecture via Microbenchmarking
- Lund jet images from generative and cycle-consistent adversarial networks
- A new algorithm for identifying the flavour of mesons at LHCb
- Deep Underground Neutrino Experiment (DUNE), Far Detector Technical Design Report, Volume III: DUNE Far Detector Technical Coordination
- GeantV: Results from the prototype of concurrent vector particle transport simulation in HEP
- GPU-Accelerated Event Reconstruction for the COMET Phase-I Experiment