As part of the UKRI ‘Living Benchmarks’ project that is exploring requirements for future High-Performance Computing infrastructure in the UK, a tutorial and hackathon for large-scale agent-based simulation using GPUs will be held at The James Hutton Institute from 27-29 October.
The tutorial will use the FLAME GPU software and no experience of GPU programming will be required. Why FLAME GPU, you may ask? Well, it is pretty much the only GPU-enabled toolkit specifically aimed at running agent-based models and can achieve extremely high performance, enabling very large simulations or large ensembles of simulations. Future HPC infrastructure will increasingly be GPU-based so this will provide an opportunity for agent-based modellers to future proof their work.
We have a small number (3-5) places available. Attendance is free, but you will need to pay for your own travel, accommodation and subsistence while here.
You will not need a GPU-enabled laptop (but if you have one, feel free to bring it). You should feel confident programming in Python.
Agenda:
- 27 October to lunchtime on 28 October: Tutorial in FLAME GPU
- afternoon of 28 October to 29 October: Hackathon converting models to FLAME GPU
You can attend for just the first of the two sessions, but priority will be given to those who want to attend the whole thing. Otherwise, first-come, first-served. To attend the whole thing, you would need to arrive on the evening of 26 October and depart the morning of 30 October.
To apply to attend, please email Gary Polhill at the James Hutton Institute.