Operating Systems

• ZeptoOS: ZeptoOS provides a complete software stack for Blue Gene/P, covering both compute and I/O nodes. The compute nodes run ZeptoOS Compute Node Linux, which features big memory (a flat, TLB-miss free memory region) and which provides all the necessary drivers and support to run MPI applications (SMP mode only in the version 2.0 release). Communication between the compute nodes and I/O nodes (file I/O, IP packet forwarding) is handled by the ZeptoOS I/O Daemon (ZOID). The current version 2.0 release is compatible with the IBM Blue Gene/P V1R3M0 software driver. For more information, and to download the release tarball, please consult the project website.

Communication Libraries

• GASNet: GASNet is a language-independent, low-level networking layer that provides network-independent, high-performance communication primitives tailored for implementing parallel global address space languages.

Programming Languages

• Coarray Fortran 2.0: Coarray Fortran (CAF) is a SPMD parallel programming model based on a set of language extensions to Fortran 90. CAF supports access to non-local data using a natural extension to Fortran 90 syntax, lightweight and flexible synchronization primitives, pointers, and dynamic allocation of shared data. Coarray Fortran 2.0 (CAF 2.0), under development at Rice University, differs significantly from the coarray-based extensions being incorporated into Fortran as part of the emerging Fortran 2008 standard. Beyond the basic coarray support in Fortran 2008, CAF 2.0 includes language support for additional features including process subsets, collective communication, asynchronous communication, function shipping, and expressive synchronization.

Computational Math

• PLASMA: The Parallel Linear Algebra for Scalable Multi-core Architectures (PLASMA) library uses out-of-order scheduling as the basis for a scalable and highly efficient software framework for computational linear algebra applications.
• MAGMA: MAGMA is a dense linear algebra library similar to LAPACK but for heterogeneous/hybrid architectures, starting with current "Multicore+GPU" systems.

Performance Analysis

• Tools
  • HPCToolkit performance tools: HPCToolkit is a suite of tools for sampling-based measurement and analysis of application performance. It works on systems of all scales: Linux laptops and desktops, Linux clusters, as well as Cray XT and Blue Gene/P leadership computing platforms.
  • hpcviewer: hpcviewer is a graphical user interface for the HPCToolkit performance tools. It enables a user to interactively explore and analyze performance data from parallel applications in the context of a program's source code.
  • Jumpshot: Jumpshot is a tool for graphical analysis of communication traces of parallel programs. While available separately, it is routinely distributed as part of the MPICH MPI library and the Tau performance tools.
  • hpctraceviewer: hpctraceviewer is a graphical user interface for examining traces of call stack samples collected for parallel programs using the HPCToolkit performance tools.
• Tool Infrastructure
  • libmonitor: libmonitor provides callback functions for begin, end of processes, threads, fork, exec, and provides a layer on which to build process monitoring tools such as profilers. It can be used to enable a profiler to gain control of statically or dynamically linked executables.
  • SymtabAPI: SymtabAPI provides support for parsing object files, including debug information such as line information, variables, and types.
  • InstructionAPI: The InstructionAPI is an emerging open-source infrastructure for parsing, representing, and analyzing machine instructions for multiple architectures. This library is being built as part of the deconstruction of the Dyninst infrastructure.