Euro-Par2020-Accelerate.zip (387.48 kB)
Artifact for Euro-Par 2020 paper Accelerating Nested Data Parallelism: Preserving Regularity
software
posted on 2020-07-03, 14:58 authored by Lars van den Haak, Trevor McDonell, Gabriele Keller, Ivo Gabe de WolffThis artifact is concerned with Section 5 (Evaluation) of the paper "Accelerating Nested Data Parallelism: Preserving Regularity". It runs benchmarks of a nested quicksort and a nested fourier transformation in Accelerate comparing the results with Futhark. Both Accelerate and Futhark are data parallel languages. To run the benchmarks, a Nvidia GPU is needed, it is made to run on Ubuntu. The benchmarks can be run with Docker or the dependency can be manually installed and runs bash scripts.