BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Making Gravity SWIFTer: GPU offloads for gravity in the SWIFT cosm
 ology code - Sarah Johnston - Durham University
DTSTART:20260212T130000Z
DTEND:20260212T140000Z
UID:TALK242527@talks.cam.ac.uk
CONTACT:Jack Atkinson
DESCRIPTION:To fully utilise modern heterogeneous HPC systems and improve 
 power efficiency\, large astronomy codes must become GPU compatible. SWIFT
  is a versatile open-source cosmology code which is used to model a variet
 y of astrophysical scenarios. It utilises task-based parallelism\, dividin
 g the workload into independent tasks managed by a scheduler for efficient
  CPU utilisation\, and is highly optimised for use on memory-intensive\, C
 PU-only clusters. A significant portion of SWIFT’s runtime is dedicated 
 to gravity calculations. However\, the repetitive and non-interdependent n
 ature of these interactions makes them ideal candidates for GPU accelerati
 on.\n\nIn this talk I will explain how we have built on the existing SWIFT
  code by replacing CPU-based gravity with new GPU kernels\, while minimisi
 ng changes to the rest of the code. The GPU-accelerated kernels achieve hi
 gh accuracy\, with &lt\;1% deviation from the CPU results. However\, we ar
 e limited by a memory transfer bottleneck in moving data between the CPU a
 nd GPU. To mitigate this\, we have employed a novel ‘task-bundling’ sy
 stem which allows us to group tasks in the scheduler to give higher occupa
 ncy on the GPU. This provides more work\, reducing the overhead from the C
 PU-GPU memory transfers. To further exploit the GPU\, we're exploring a re
 distribution of the gravity calculations meaning more interactions can be 
 carried out using direct particle-particle summations\, rather than multip
 ole-based approximations\, which gives us more accurate results and provid
 es more work for the GPU. Future work will focus on optimising these new G
 PU implementations to provide an end-to-end speedup in SWIFT on heterogene
 ous systems.
LOCATION:Room C\, West Hub
END:VEVENT
END:VCALENDAR