Community for Programming Models for Scalable Parallel Computing ...

1803 days ago, 743 views
PowerPoint PPT Presentation

Presentation Transcript

Slide 1

Libraries, Languages, and Execution Models for Terascale Applications William D. Gropp Argonne National Laboratory Center for Programming Models for Scalable Parallel Computing: Project Meeting Report

Slide 2

Participants Coordinating Principal Investigator: Ewing Lusk – Argonne National Laboratory Co-Principal Investigators (Laboratories): William Gropp – Argonne National Laboratory Ricky Kendall – Ames Laboratory Jarek Nieplocha – Pacific Northwest National Laboratory Co-Principal Investigators (Universities): Barbara Chapman – University of Houston Guang Gao – University of Delaware John Mellor-Crummey – Rice University Robert Numrich – University of Minnesota Dhabaleswar Panda – Ohio State University Thomas Sterling – California Institute of Technology Marianne Winslett – University of Illinois Katherine Yelick – University of California, Berkeley

Slide 3

Problem Statement Problem: Current programming models have empowered improvement of adaptable applications on current expansive scale PCs, yet the application advancement handle itself stays unpredictable, long, and costly, blocking progress in logical application advancement. Arrangement: Facilitate application improvement by giving standard libraries, advantageous parallel programming dialects, and petaflops-focused on cutting edge programming models. Objectives: A variety of appealing alternatives for advantageous, effective, improvement of adaptable, proficient logical applications for terascale PCs

Slide 4

A Three-Pronged Approach to Next-Generation Programming Models Extensions to existing library-based models MPI (- 2; expansions) Global Arrays and augmentations Portable SHMEM Robust executions of dialect based models UPC Co-Array Fortran Titanium OpenMP enhancements Advanced models for cutting edge structures Multithreaded, PIM-based machines, Gilgamesh, and so forth

Slide 5

Application Programming Models Message Passing Remote Memory Shared Memory Mixed Models Language Extensions New Models Model Instances MPI-2 GA GPSHMEM OpenMP + MPI CAF UPC Titanium EARTH Implementation Substrate Common Runtime ADI-3 ARMCI Panda Parallel I/O CAF Packages/Modules Open64 Compiler HDF-5 Communication Firmware MPP Switches VIA Myrinet Infiniband Relationships Among the Parts

Slide 6

Libraries for the remote memory get to display MPI and MPI-2 Global Arrays GA consolidate more elevated amount show with productivity for application accommodation GP-SHMEM Popular Cray T3E demonstrate made compact Co-Array Fortran library Object-based logical library, written in CAF

Slide 7

Languages Three dialects giving a product worldwide address space (appropriate for disseminated memory) and parallelism CAF (Co-Array Fortran) UPC (Unified Parallel C) Titanium (parallel Java) One dialect for shared memory Scalable OpenMP The Open64 compiler foundation Industrial quality compiler for C, Fortran 9x, C++ Used in the above ventures One commitment to the group

Slide 8

Cross-Project Infrastructure Runtime correspondence approaches Exploiting NICs in support of parallel programming models ARMCI GASNet I/O Active buffering in Panda MPI-IO and parallel record frameworks Integrating dynamic buffering into ROMIO usage of MPI-IO Scalable I/O for parallel dialects UPC CAF I/O

Slide 9

New Programming Models Defining another execution demonstrate Semantics first Define for execution Must give the tremendous advantage Bill Camp specified Define to bolster best calculations in support of utilizations Define for likely HPC equipment, including Many (zillions) processors Deep memory order Some equipment bolster for programming model Likely to have some sort of decisively casual memory consistency display Common component of the greater part of the elite libraries and dialects in the venture (even OpenMP) Experiments with new ideas, for example, permeation (move program to information rather than information to program)

Slide 10

Connections With Other Programs Applications from SciDAC, NSF/PACI, and so on. DARPA HPCS Program John Mellor-Crummey (Rice) for HP Bob Numrich (UMN) for SGI Thomas Sterling (JPL/Caltech) for Cray Kathy Yelick (Berkeley) for SUN Guang Gao (U Delaware) IBM ANL an individual from Cray Affiliates program Open64 Community OpenMP (U Houston shaped an organization to join ARB, since no one but organizations can be individuals ) IBM Blue Gene/L and QCDoC More…