200 likes | 330 Views
Update on TASCS Component Technology Initiatives. Lois McInnes, Focus Lead Presented by Jarek Nieplocha, HPC Initiative Lead. Support for Emerging HPC Hardware and Software Paradigms. Overall Goal : Enable effective use of CCA on emerging HPC architectures
E N D
Update on TASCS Component Technology Initiatives Lois McInnes, Focus Lead Presented by Jarek Nieplocha, HPC Initiative Lead
Support for Emerging HPC Hardware and Software Paradigms • Overall Goal: Enable effective use of CCA on emerging HPC architectures • Multiple-Component-Multiple-Data (MCMD) CCA Technology enables componentization of apps based on multiple levels of parallelism with focus on massively parallel systems. • CCA on Heterogenous Architectures helps with portability and management of software modules executing on hardware accelerators (FPGA, IBM Cell, NVDIA Tesla etc). • Ongoing work: mass spectrometer applications: code transformations to make new component • CCA in Fault Tolerant Environments focuses on better support for massively parallel architectures where frequent hardware faults are expected. • ORNL represents us on ASCR CIFTS project. • Tried to harmonize CCA Event Service with their API for the Fault Tolerance Backplane (currently frozen to support SC’07 demos) • A component-based FT demonstration using the FTB has been delayed due to platform access and application portability issues.
MCMD • Recent activities focus on development of specifications for CCA-based processor groups teams • BOFs held during CCA meetings in April and July, 2007 • Mini-Workshop held January 24, 2007 • Use cases documented and analyzed • Wiki webpage and mailing list: https://www.cca-forum.org/wiki/tiki-index.php?page=MCMD-WG • Specifications document version 0.3 • Telecon held Sept 28, 2007 • Some people sent good comments by email • Issues about threads, fault tolerant environment, MPI-centric narrative and examples, ID representation • Plans • Complete and present the document at the next meeting • Prototype implementation and some application evaluation
Software Quality and Verification • Overall Goal: Extend component interface definitions with semantic annotations that can be verified/enforced at composition or run time (contracts) • Recent Activities • Publications • S. Parker et al., “Enabling Advanced Scientific Software,” submitted to Cyberinfrastructure Technology Watch (http://www.ctwatch.org) for special 4th quarter “DOE SciDAC Software Enabling Technologies for Petascale Science” issue. • Presentations • T. Dahlgren, “Performance-Driven Interface Contract Enforcement for Scientific Components,” in Proceedings of the 10th International ACM SIGSOFT Symposium on Component-Based Software Engineering (CBSE-10), Medford, MA, July 9-11, 2007. Also available as Lawrence Livermore National Laboratory Technical Report UCRL-PRES-232517. • Progress integrating experimental enforcement features into Babel repository • Refactored runtime library features to better encapsulate internal use-only methods. • Revised assertion regression test (Vector Utils) – more robust implementation and client side cleanup (including comments with new enforcement-related calls). • Preliminary integration of new features into command line, user options, DTD, XML parser, and symbol table. • University/education interactions: T. Dahlgren submitted to UC Davis committee the latest version of Ph.D. thesis partially based on this work • Plans • October-December: Complete integration of experimental enforcement features into the Babel compiler (i.e., symbol table, C back end, and IOR) … capabilities to be used by CQoS working group.
Computational Quality of Service • Overall Goal: Develop infrastructure for dynamic component adaptivity, i.e., composing, substituting, and reconfiguring running component applications in response to changing conditions • Performance, accuracy, mathematical consistency, reliability, etc. • CQoS Wiki and mailing list: https://www.cca-forum.org/wiki/tiki-index.php?page=Computational+Quality+of+Service • Recent Activities • CQoS session (pre-CCA meeting): July 18, 2007, Silver Spring, MD • Updates from everyone: See details/slides: https://www.cca-forum.org/wiki/tiki-index.php?page=Summer+2007+Agenda • Joe Kenny (SNL): Automatic selection of integral evaluators in quantum chemistry • Jaideep Ray (SNL): A first evaluation of rule-based dynamic configuration of load-balancers for AMR simulations of flows • Masha Sosonkina (Ames): Adaptivity features of sparse iterative linear algebra kernels represented as components • Alan Morris (Univ of Oregon): CQoS infrastructure in TAU • Andrea Berger (Clarion Univ), Lois Curfman McInnes and Boyana Norris (ANL): Simple linear solver components using SciPy • Li Li (ANL): CQoS infrastructure for scientific applications • Rob Armstrong (SNL): A data component for the CQoS testbed • Group discussion: Applications teams' highest priority for CQoS infrastructure: tools for performance database query and management • Decided to establish monthly group telecons to sync on overall progress • Thus far have been very useful and effective
Computational Quality of Service • Recent Activities (cont.) • CQoS telecon: Aug 27, 2007 • Details: https://www.cca-forum.org/wiki/tiki-index.php?page=CQoS:8-27-2007 • Joe Kenny (SNL): plans for quantum chemistry simulations using CQoS tools • Jaideep Ray (SNL): plans for combustion simulations using CQoS tools • Li Li (ANL): plans for adaptive linear solvers using matrix metadata via Anamod; proposed performance database store and query API • CQoS telecon: Sept 24, 2007 • Details: https://www.cca-forum.org/wiki/tiki-index.php?page=CQoS:9-24-2007 • Jaideep Ray (SNL): CQoS database usage in combustion • Joe Kenny (SNL): CQoS database usage in quantum chemistry • Meng-Shiou Wu (Ames): chemistry performance database and CQoS • Li Li, Boyana Norris, and Lois McInnes (ANL): CQoS database usage in adaptive linear solvers • CQoS BOF:to be held tomorrow (Oct 12, 2007) • Review new performance metadata formats that were coordinated by PERI (Boyana Norris) • Establish specific requirements for new performance database capabilities • Work on defining component interfaces for database query and management (C++ and SIDL)
EnvironmentYear 1 Deliverables Gary Kumfert Key = Done = On Track | Visible progress = Need More Next Meeting = Unknown ?
PORTING to NLCF Machines XT-3 Babel (ORNL-lead) Wael started this in Indiana Mtg. Ccaffeine (SNL-lead) Tutorial (?) Toolkit (?) Blue Gene Babel (LLNL-lead) Ccaffeine (SNL-lead) Tutorial (?) Toolkit (?) Set up Helpdesk (LLNL-lead) https://cca-forum.org/bugs/help help@cca-forum.org Documentation CCA Specification (LLNL-lead) Core Tools: BenInstitutions: SNL, LLNL, UU, ORNL, ANL Babel ported to BG/L (thanks Tom) C,C++,F77 independently confirmedat ANL (thanks Boyana) Postpone? Tutorial & Toolkit are being rewritten w/ Bocca for SC|07 Feedbackto GNU
Inter-framework Prototypes CCA-Kepler (UU-lead) XCAT-BabelRMI (BU-lead) CCA Services CommandLineService (???) MPIService (SNL-lead) EventService (UU-lead) Waiting on CCARB to be finalized GuiBuilderService (ANL-lead) Waiting on EventService Enhancements: TomInstitutions: LLNL, BU, UU, SNL, ORNL ? Funding delays Bonus Items • New CCA Specification Process • Passed • Awaiting CCARB to be named & approved • CCA 1.0 Requirements Doc • (Modelled after Babel’s activity) • Will start up as a WG once CCARB is seated • Babel Structs (Y2 deliverable) • C/C++ bindings are in SVN • More details in Babel Roadmap talk later
Draft CCA-Lite Spec & Fwk (LANL-lead) Component Test Harness (ORNL-lead) Document Multilanguage & Component Debugging Techniques (SNL or ANL) Matt Usability: CraigInstitutions: LANL, SNL, ANL, ORNL, Tech-X, PNNL, LLNL ? This section is being refocused. Matt will lead a separate discussion about usability.
Major New Activity • Started collaboration with CScADS, ITAPS, and Chemistry Apps to look for optimization techniques to improve performance of Babel. • John Mellor-Crumney is looking to apply compiler optimization technologies to the multi-lingual environment Babel presents. • New mailing list: cca-fast@cca-forum.org • We will meet again at SC|07 (exact time/place TBD)
Toolkit Rob Armstrong Focus Lead
Toolkit News • Sandia is looking at componentizing Chambo. • Bocca is to be released. • Toolkit components can be easily created from existing code • See other presentations today • A toolkit component porting camp?
User and ApplicationsOutreach and Support David E. Bernholdt Focus Lead
Potential Collaborations • Govindaraju–ORNL on HPC checkpointing • not CCA yet, but natural connection with HPC initiative • Chiu-BU cardiac modeling • CCA potential • Bernholdt, Shasharina–ITER Integrated Modeling • Common data model, interfaces, possibly frameworks • September meeting in France, second with this group • Norris–Community Surface Dynamics Modeling System @ U Colorado (NSF) • Larson–A Data Domain to Model Domain Conversion Package (DMCP) for Sparse Climate Related Process Measurements (SciDAC SAP)
Publications & Presentations • Fang Liu and Randall Bramley, CCA-LISI: On Designing A CCA Parallel Sparse Linear Solver Interface, Proc. of the IEEE International Parallel and Distributed Processing Symposium, 2007 • Steven G. Parker, Kostadin Damevski, Ayla Khan, Ashwin Swaminathan, Chris R. Johnson. The SCIJump Framework for Parallel and Distributed Scientific Computing. In Advanced Computational Infrastructures for Parallel/Distributed Adaptive Applications. Edited by Manish Parashar, Xiaolin Li, and Sumir Chandra, Wiley Press, 2007 (to appear) • K.Damevski, A. Swaminathan, S. Parker. "Highly Scalable Distributed Component Framework for Scientific Computing". In Proceedings of the 3rd International Conference on High Performance Computing and Communication (HPCC 2007), Houston, Texas • K. Damevski, K. Zhang, S. Parker "Practical Parallel Remote Method Invocation for the Babel Compiler". In Proceedings of the HPC-GECO/CompFrame'07 Workshop, Montreal, Canada (accepted) • Jarek Nieplocha keynote at 7th International Conference on Parallel Processing and Applied Mathematics on Sept. 10 in Gdansk, Poland • D. Dechow and L. McInnes, Components for Beam Dynamics, All-Hands kickoff meeting for the Community Petascale Project for Accelerator Science and Simulation (COMPASS), Fermilab, Sept 17, 2007.
Publications & Presentations • D.R. Dechow, B. Norris, and J. Amundson, The Common Component Architecture for Particle Accelerator Simulations, Proceedings of HPC-GECO/CompFrame'07, October 21-22, 2007, Montreal, Quebec, Canada, ACM, 2007. • J.W. Larson and B. Norris, Component Specification for Parallel Coupling Infrastructure, in / Proceedings of the International Conference on Computational Science and its Applications (ICCSA 2007)/, O. Gervasi and M.L. Gavrilova (eds) Lecture Notes in Computer Science Volume 4707, Springer-Verlag, 55-68 (2007). • CTWatch paper • David E. Bernholdt, The Role of Component Software Technology in Meeting the Challenge of Petascale Scientific Simulation, seminar, PetroBras CENPES (Research Center), 2007. • David E. Bernholdt, The Common Component Architecture: Building Frameworks for Computational Science, keynote talk, International Conference on Modeling and Simulation in the Petroleum Industry, 2007. People & Places • Fang (Cherry) Liu (IU) @ Tech-X (intern) • Samantha Foley (IU) @ ORNL (intern) • Damian Rouson @ SNL • New hire, but past Forum participant
Recent Activities/Near-Term Plans • October • Kumfert to represent TASCS at SciDAC Centers and Institutes Workshop • Shasharina invited talk at ICALEPCS07 • HPC-GECO/CompFrame, 21-22 Oct, Montreal • November • SC2007: Babel, CCA, Global Arrays tutorials • December • e-Science 2007, 10-13 December, Bangalore, India (contact Ken Chiu) • April • High Performance Computing and Simulation Symposium (HPCS 2008), Ottawa, Canada (contact Masha Sosonkina)