Rishabh Dave L&S Math & Physical Sciences
Just-in-time GPU Compilation for Liquid Argon Simulations
High energy particle physics simulations increasingly rely on highly parallelized computation to calculate detector readouts. One case where this is particularly relevant is the upcoming DUNE neutrino experiment, where simulations of liquid argon time projection chambers rely heavily on parallel GPU computations. DUNE simulations needs the ability to write data from the GPU kernels back to the arrays, which the awkward Python library is currently unable to support. If the awkward array library was edited to allow for this compatibility it would be possible to use both of these tools simultaneously, not only avoiding the trouble of independently implementing variable length arrays without awkward, but also creating a better combination of tools that can be used more generally in high energy physics and parallel scientific computing more generally.
Message To Sponsor
Thank you so much for this opportunity! it was really challenging to find summer physics programs this summer due to budget cuts, so having a local program that allowed me to continue research was hugely helpful. I also thing that having the workshops and small group meetings helped me be better prepared for grad school applications and more aware of the latest topics being studied both in and out of my field. I thank you for the opportunity to continue research in the summer- because of this I was able to pursue a long term project within my group that I might not otherwise have been able to discuss, which is shaping the direction I may focus on in my PhD.