Benchmarks - lattice/quda GitHub Wiki
Here we give (hopefully) up-to date performance data from using QUDA (through a host application). Given the nature of performance optimization and algorithmic innovation, this is an evolving document. One has to take benchmarks between different machines that are significantly separated in time with a large pinch of salt. E.g., the multigrid algorithm we were running in 2016 is not the same as the one running in 2021.
Feel free to suggest any particular benchmark that we can add to this page.