arXiv:1507.08101v3 [cs.DC] 15 Feb 2016 Noname manuscript No. (will be inserted by the editor) GHOST: Building Blocks for High Performance Sparse Linear Algebra on Heterogeneous Systems Moritz Kreutzer · Jonas Thies · Melven R¨ ohrig-Z¨ ollner · Andreas Pieper · Faisal Shahzad · Martin Galgon · Achim Basermann · Holger Fehske · Georg Hager · Gerhard Wellein the date of receipt and acceptance should be inserted later Abstract While many of the architectural details of future exascale-class high per- formance computer systems are still a matter of intense research, there appears to be a general consensus that they will be strongly heterogeneous, featuring “standard” as well as “accelerated” resources. Today, such resources are available as multicore processors, graphics processing units (GPUs), and other accelerators such as the Intel Xeon Phi. Any software infrastructure that claims usefulness for such environments must be able to meet their inherent challenges: massive multi-level parallelism, topol- ogy, asynchronicity, and abstraction. The “General, Hybrid, and Optimized Sparse Toolkit” (GHOST) is a collection of building blocks that targets algorithms dealing with sparse matrix representations on current and future large-scale systems. It imple- ments the “MPI+X” paradigm, has a pure C interface, and provides hybrid-parallel numerical kernels, intelligent resource management, and truly heterogeneous paral- lelism for multicore CPUs, Nvidia GPUs, and the Intel Xeon Phi. We describe the details of its design with respect to the challenges posed by modern heterogeneous M. Kreutzer · F. Shahzad · G. Hager Erlangen Regional Computing Center, Friedrich-Alexander-Universit¨ at Erlangen-N¨ urnberg, 91058 Erlan- gen, Germany E-mail: {moritz.kreutzer, faisal.shahzad, georg.hager}@fau.de J. Thies · M. R ¨ ohrig-Z ¨ ollner · A. Basermann German Aerospace Center (DLR), Simulation and Software Technology, 51147 K¨ oln, Germany E-mail: {jonas.thies, melven.roehrig-zoellner, achim.basermann}@dlr.de A. Pieper · H. Fehske Institute of Physics, Ernst-Moritz-Arndt-Universit¨ at Greifswald, 17489 Greifswald, Germany E-mail: {pieper, fehske}@physik.uni-greifswald.de Martin Galgon Bergische Universit¨ at Wuppertal, 42097 Wuppertal, Germany E-mail: galgon@math.uni-wuppertal.de G. Wellein Department of Computer Science, Friedrich-Alexander-Universit¨ at Erlangen-N¨ urnberg, 91058 Erlangen, Germany E-mail: gerhard.wellein@fau.de