The NAS Parallel Benchmarks for evaluating C++ parallel programming frameworks on shared-memory architectures

The NAS Parallel Benchmarks (NPB), originally implemented mostly in Fortran, is a consolidated suite containing several benchmarks extracted from Computational Fluid Dynamics (CFD) models. benchmark has important characteristics such as intensive memory communications, complex data dependencies, different access patterns, and hardware components/sub-systems overload. programming APIs, libraries, frameworks that are written C++ well new optimizations parallel processing techniques can benefit if NPB made fully available this language. In paper we present NPB-CPP, translated version of consisting all the kernels pseudo-applications developed using OpenMP, Intel TBB, FastFlow for multicores. design NPB-CPP leverages Structured Programming methodology (essentially based on patterns). We show structure each application terms composition few patterns (notably Map MapReduce constructs) provided by selected frameworks. experimental evaluation shows accuracy with respect to original source code. Furthermore, carefully evaluate performance three multi-core systems (Intel, IBM Power, AMD) compilers (gcc, icc, clang) discussing differences order give researchers useful insights choose best framework given type problem. • evaluating programming. TBB FastFlow. Performance analysis implementations multicore platforms.