Posts by rdemaria

1) Message boards : Number crunching : GPU advertised for LHC, but they don't do it? (Message 40258) Posted 24 Oct 2019 by rdemaria Post: Hello, I confirm we plan to enable GPU computing in SixTrack with Boinc support. We are developing a new library SixTrackLib which would enable GPU offloading for a large subset of SixTrack simulations. As some of you argued, we are focusing on using OpenCL 1.2 because it allows to run on AMD, Intel and Nvidia GPUs, although is less advanced than other options (OpenCL 1.2 dates back to 2011). Generally GPUs have very strong 32bit floating point capabilities, but we will not be able to profit from them. The calculations needed by SixTrack requires variables with large dynamic range, which is very different from tasks like machine learning. SixTrack computes the trajectories of the particles as they move around the accelerator. There are 6 main quantities (3 positions and 3 velocities for each plane: horizontal, vertical, longitudinal) that oscillates around an equilibrium trajectory. Each position and velocity is mainly affected by the velocities and positions of the corresponding planes and very weakly from the other planes. This weak interactions are small in the short term, but become very important in the long term (that are the typical scale of our simulations). A weak interaction correspond to the operation of adding a small quantity on a large number, and if we the number does not have enough bits, we loose the information and the simulation would be wrong. Despite this, GPUs also support 64bit floating point calculations but at a lower speed, typically 16/32 times more slowly, however still the computing capabilities they provide are comparable and often exceed the ones of CPUs. Even the latest embedded GPUs in notebooks are fairly capable. There are also GPUs that are strong in 64bit arithmetic such as Nvidia Quadro GV100, Tesla V100/P100, Titan V, AMD Radeon VII, Firepro W9100, W8100 although they are very expensive and less common in workstations (instead more common in supercomputers and the cloud providers). For this reason, we would like to enable GPU offloading in SixTrack because we believe there are substantial additional computing resources that we would be able to access for our simulations. We are still in the process of developing SixTrackLib, validating the physics and tackle the integration into the main SixTrack. Our time scale is still probably a year or two to expose the functionality into LHC@Home, but we already performing simulations using SixTrackLib in GPU for specific problems. Please do not hesitate to ask if you have further questions and my apologies for the slow answer.... Riccardo from the SixTrack team

LHC@home