UPCBLAS: a library for parallel matrix computations in Unified Parallel C

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 such as multicore clusters. This paper describes UP-CBLAS, a parallel numerical library for dense matrix computations using the PGAS Unified Parallel C (UPC) language. The routines developed in UPCBLAS are built on top of sequential BLAS functions and exploit the particularities of the PGAS paradigm, taking into account data locality in order to achieve a good performance. Furthermore, the routines implement other optimization techniques, several of them by automatically taking into account the hardware characteristics of the underlying systems on which they are executed. The library has been experimentally evaluated on a multicore supercomputer and compared to a message-passing based parallel numerical library, demonstrating good scalability and efficiency.

parSRA: A framework for the parallel execution of short read aligners on compute clusters

“…PGAS languages (such as UPC, Co-Array Fortran [27] or Titanium [28]) are often easier to use than their message passing counterparts [29,30] and can also obtain better performance by using efficient one-sided communication [31][32][33]. UPC++ combines these advantages of the PGAS model with object oriented programming.…”

Parallel Differential Evolution in Unified Parallel C

“…One or more candidate vectors will be affine to each UPC thread (see fig. 4 for illustration of the memory layout); 2 Evaluate an objective function ranking the vectors in the population in parallel; 3 while Termination criteria not satisfied do 4 Each thread T in parallel 5 for i = {1, . .…”

“…On the other hand, wrong communication patterns in which processes unnecessarily access non-local regions of shared memory can be obtained easily due to the simplicity of the model. The PGAS is a popular model for high performance computing (HPC) [3] implemented by a number of programming languages. Most used PGAS languages include Coarray Fortran and Unified Parallel C (UPC) [2], [4], [5].…”

Parallel and Scalable Short-Read Alignment on Multi-Core Clusters Using UPC++

“…Our parallel implementation overcomes the scalability issues of pMap thanks to an efficient use of UPC++ [ 18 ], an extension of C++ for parallel computing which has evolved from Unified Parallel C (UPC) [ 19 ]. PGAS (Partitioned Global Address Space) languages (such as UPC, Co-Array Fortran [ 20 ] or Titanium [ 21 ]) are often easier to use than message passing counterparts [ 22 , 23 ] and can also obtain better performance than them thanks to efficient one-sided communication [ 24 – 26 ]. UPC++ combines these advantages of the PGAS model and object oriented programming.…”