Session

DescriptionHigh-energy physics (HEP) experiments have developed millions of lines of code over decades that are optimized to run on traditional x86 CPU systems. However we are seeing a rapidly increasing fraction of floating point computing power in leadership-class computing facilities and traditional data centers coming from new heterogeneous accelerator architectures, such as GPUs. Though the GPU field is currently being led by NVIDIA, other manufacturers such as Intel and AMD are making increasing inroads into this territory, each with their own architecture and compiler languages. Many HEP experiments are also using FPGAs for their frontend detector readouts. Rewriting current CPU based High Energy Physics code for multiple accelerator architectures is not a viable scenario, given the available person power and code maintenance issues. Furthermore, as the number of architectures proliferate, it becomes increasingly onerous to validate the code, and it is vital to ensure that workflows on different hardware produce identical results. Developing portable solutions for HEP codebases to run on multiple heterogeneous architectures is essential. We explore how major HEP experiments are addressing these issues, and also a domain wide investigation to compare the efficacy of the various portability layers such as Kokkos, SYCL, Alpaka, OpenMP, and std::execution::parallel.