https://hgpu.org/?p=12839
NAS Parallel Benchmarks for GPGPUs using a Directive-based Programming Model