UPCBLAS : a numerical library for unified parallel C with architecture-aware optimizations

[Abstract] The popularity of Partitioned Global Address Space (PGAS) languages has increased during the last years thanks to their high programmability and performance through an efficient exploitation of data locality, especially on hierarchical architectures like multicore clusters. This PhD Thesis describes UPCBLAS, a parallel library for numerical computation using the PGAS Unified Parallel C (UPC) language. The routines are built on top of sequential BLAS and SparseBLAS functions and exploit the particularities of the PGAS paradigm, taking into account data locality in order to achieve a good performance. However, the growing complexity in computer system hierarchies due to the increase in the number of cores per processor, levels of cache (some of them shared) and the number of processors per node, as well as the high-speed interconnects, demands the use of new optimization techniques and libraries that take advantage of their features. For this reason, this Thesis also presents Servet, a suite of benchmarks focused on detecting a set of parameters with high in uence on the overall performance of multicore systems. UPCBLAS routines use the hardware parameters provided by Servet to implement optimization techniques that improve their performance. The performance of the library has been experimentally evaluated on several multicore supercomputers and compared to message-passing-based parallel numerical libraries, demonstrating good scalability and efficiency. UPCBLAS has also been used to develop more complex numerical codes in order to demonstrate that it is a good alternative to MPI-based libraries for increasing the productivity of numerical application developers