high performance computing on graphics processing units: hgpu.org

SYCL is a royalty-free open standard from the Khronos group that enables heterogeneous programming for a broad range of parallel devices, including multicore CPUs, GPUs, and FPGAs. SYCL relies on pure C++, without any additional attributes or pragmas. While SYCL kernels follow a data-parallel model, they are implicitly organized in a task graph built by the runtime from data access specifications. Scheduling, data management, and synchronization between different tasks are handled implicitly by the SYCL runtime. While this simultaneously preserves programmer productivity and allows the SYCL runtime to automatically perform optimizations such as overlapping data transfers and kernel execution, it is not apparent whether a SYCL implementation actually employs such optimizations. Benchmarks are therefore necessary to characterize the performance of programming models such as SYCL that rely heavily on implicit operations. To this end, we present SYCL-Bench, a versatile benchmark suite written in SYCL. The main goal of SYCLBench is to characterize both devices and different SYCL implementations. Therefore, SYCL-Bench not only contains benchmarks to characterize the hardware, but also SYCLspecific benchmarks that present optimization opportunities to the SYCL runtime and test how well a particular SYCL implementation capitalizes on those opportunities.