XPRESS: eXascale PRogramming Environment and System Software Sandia National Laboratories PI: Ron Brightwell Thomas Sterling, Project Chief Scientist, Indiana University Barbara Chapman, University of Houston Hartmut Kaiser, Louisiana State University Alice Koniges, Lawrence Berkley National Laboratory Andrew Lumsdaine, Indiana University Allen Maloney, Oregon University Alan Porterfield, University of North Carolina Chris Baker, Oak Ridge National Laboratory
1
Scope of the Project New Model Applications
Legacy Applications
OpenMP
Domain Specific Language Domain Specific Active Library
MPI
Metaprogramming Framework
2. LXK lightweight OS for O(N) scalability, performance portability, no OS noise and minimum abstraction layer to runtime software and billion-way hardware
1. Guided by ParalleX execution model & SLOW performance model for codesign of dynamic adaptive resource management and task scheduling system software & programming model with introspection
3. HPX runtime software for efficiency and scalability of science application by adapting workload to resource availability and discovery of datadirected parallelism 4. XPI & RIOS interface protocols for cross-platform portability
Scope of the Project v Establish logical and organizational abstractions for foundations of general-purpose exascale computing architectures and programming models v Enable scalability to exascale by dynamic parallelism discovery and low overhead for fine grain threads v Achieve best time to solution and low energy by up to 10X efficiency for real apps through low overhead and adaptive latency and contention management v Full-scale deployment of prototype XPRESS system software for general-purpose DOE applications v Legacy codes and methods migration path v Partnering with other DOD, DOE, NSF programs
Milestones & Deliverables
• Milestones & deliverables 1. 2. 3. 4. 5. 6.
XPI intermediate form and low level programming interface RIOS interface protocol between OS and run?me system LXK lightweight kernel scalable opera?ng system HPX-‐4 run?me system based on ParalleX execu?on model MPI transla?on to XPI for legacy code migra?on ParalleX, introspec?on framework and policies, translators Milestones (Level 1) • XPI & RIOS Specifications • Merger of Kitten & HPX-3 • Introspective framework • MPI to XPI Port • HPX-4 Developed with LXK • Deploy, Test, Evaluate with apps
Scheduled FEB 2013 AUG 2013 FEB 2014 AUG 2014 FEB 2015 Aug 2015
Dependencies • Needs from other projects? – – – – – –
Which projects do you think can provide what you need 3 co-‐design centers for applica?on drivers Quinlan project for DSL OCR for high level programming interface Intel project for possible target architecture MIT for control policies
• What we offer to other projects? – – – – –
Lightweight kernel opera?ng system alterna?ve to Linux Run?me system that is dynamically adap?ve and scalable Low level intermediate form and programming interface Protocol specifica?on for interface between OS and Run?me ParalleX execu?on model abstrac?on for seman?c founda?on
Gaps • Gaps in your project – Interface to co-‐design centers; Post-‐doc model may not work – Where is the parallelism going to come from to drive billion-‐wary concurrency – Performance models for exascale system – Well founded technology roadmap to base 2020 extrapola?ons
• An?cipated needs in the future – – – –
Testbed plaWorms for proof-‐of-‐concept test and cer?fica?on Base-‐level benchmarks and metrics XPI versions of key applica?on kernels with data sets I/O models including mass storage for big data
EEC-2012-10-Brightwell.pdf
Page 1 of 6. Sandia National Laboratories. PI: Ron Brightwell. Thomas Sterling, Project Chief Scientist, Indiana University. XPRESS: eXascale PRogramming. Environment and System Software. 1. Barbara Chapman, University of Houston. Hartmut Kaiser, Louisiana State University. Alice Koniges, Lawrence Berkley ...