Menu
▾
▴
x10-announce — X10-related announcements
You can subscribe to this list here.
| 2007 |
Jan
(2) |
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2008 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
(1) |
Dec
(1) |
| 2009 |
Jan
|
Feb
|
Mar
(1) |
Apr
|
May
|
Jun
(1) |
Jul
|
Aug
(1) |
Sep
|
Oct
(1) |
Nov
(1) |
Dec
|
| 2010 |
Jan
(1) |
Feb
|
Mar
(2) |
Apr
(1) |
May
|
Jun
(1) |
Jul
(1) |
Aug
|
Sep
(1) |
Oct
(2) |
Nov
|
Dec
|
| 2011 |
Jan
(3) |
Feb
(2) |
Mar
|
Apr
|
May
(1) |
Jun
(1) |
Jul
|
Aug
|
Sep
(1) |
Oct
|
Nov
|
Dec
|
| 2012 |
Jan
(1) |
Feb
(3) |
Mar
(1) |
Apr
|
May
|
Jun
|
Jul
|
Aug
(1) |
Sep
|
Oct
(1) |
Nov
|
Dec
|
| 2013 |
Jan
(2) |
Feb
|
Mar
(1) |
Apr
(1) |
May
|
Jun
(1) |
Jul
|
Aug
|
Sep
(1) |
Oct
|
Nov
|
Dec
(1) |
| 2014 |
Jan
|
Feb
(2) |
Mar
(1) |
Apr
|
May
(2) |
Jun
|
Jul
|
Aug
|
Sep
|
Oct
(1) |
Nov
(1) |
Dec
(1) |
| 2015 |
Jan
|
Feb
(1) |
Mar
(1) |
Apr
|
May
(2) |
Jun
(1) |
Jul
(1) |
Aug
(2) |
Sep
(1) |
Oct
|
Nov
|
Dec
(1) |
| 2016 |
Jan
|
Feb
(1) |
Mar
|
Apr
|
May
(1) |
Jun
(1) |
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
| 2017 |
Jan
|
Feb
|
Mar
|
Apr
|
May
(1) |
Jun
|
Jul
(1) |
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
|
From: David P G. <gr...@us...> - 2017-07-03 18:40:06
|
We are happy to announce the release of X10 and X10DT 2.6.1. The focus of the X10 2.6.1 release is improvements to Resilient X10. The release is available for download at http://x10-lang.org/releases/x10-release-261 The X10 2.6.1 Release Notes are appended. X10 Release 2.6.1 HIGHLIGHTS OF THIS RELEASE The focus of the X10 2.6.1 release is improvements to Resilient X10. Some highlights include: - Significant improvements to the performance of resilient finish - A new resilient store implemented in X10 - Improved standard library APIs for elasticity and non-shrinking recovery. - Enhancements to the x10.util.resilient.iterative package to exploit elasticity to support non-shrinking recovery. - Updating of all Resilient X10 sample programs and applications to showcase the improvements in Resilient X10. LIMITATIONS OF THIS RELEASE The size and index of arrays must be less than 2^31 with Managed X10. An attempt to allocate a longer array causes IllegalArgumentException. The following features described in the 2.6 language manual do not currently work and may be fixed in the subsequent releases: - Non-static type definitions as class or interface members (static type defs do work) The constraint solver used by X10 typechecker is known to be incomplete for situations in which a constraint implies an infinite number of distinct constraints. Additionally, the following features described in the language manual do not currently work with Native X10. - Non-final generic instance methods - Exception stack traces on Cygwin The generated C++ code requires g++ 4.2 or better to be compiled; we do almost all of our testing against g++ 4.4. + On Power/Linux and BG/Q, you may either use g++ 4.2 or better or xlC 11.1 or better. SUMMARY OF ISSUES RESOLVED IN THIS RELEASE Below is a summary of JIRA issues addressed for the X10 2.6.1 ** New Feature * [XTENLANG-3553] - Add method x10.lang.String.contains(String) * [XTENLANG-3554] - New method x10.Random.nextGaussian * [XTENLANG-3555] - X10 Launcher fails with buffer overflow if X10_HOSTFILE does not exist ** Bug * [XTENLANG-3498] - New place fails to connect to dead places * [XTENLANG-3552] - X10 reading from a file not as expected * [XTENLANG-3556] - x10rt_sockets: blocking_probe doesn't block after a Place is killed For the details of JIRA issues fixed in this release, see https://xtenlang.atlassian.net/secure/ReleaseNote.jspa?version=10600&projectId=10005 Please use the X10 JIRA to report bugs, after ensuring the problem is not already reported: https://xtenlang.atlassian.net/projects/XTENLANG/issues |
|
From: Smith, L. L C. (US) <lau...@ma...> - 2017-05-02 16:26:27
|
CALL FOR PAPERS PAW17: The 2nd Annual PGAS Applications Workshop http://sourceryinstitute.github.io/PAW/ Held in conjunction with SC 17: The International Conference for High Performance Computing, Networking, Storage and Analysis http://sc17.supercomputing.org In cooperation with SIGHPC SUMMARY The race towards Exascale computing is on, and a lot of stress is put on researchers to break the boundaries of productivity and efficiency imposed by traditional programming models. Partitioned Global Address Space (PGAS) languages are an effective alternative, and the most promising path towards sustainable programming environments for exascale machines. Languages such as UPC, Fortran, Chapel, and X10 are now more widely available than ever, thanks to increased support from vendors and open-source communities. PGAS models also take the form of meta-languages and libraries, such as Unified Parallel C++ (UPC++), Co-Array C++, OpenSHMEM, MPI-3 and Global Arrays. These have the benefit of being integrated with existing languages, simplifying the learning curve for existing programmers. Significant improvements in the availability of PGAS compilers and support software have been achieved in the last few years; these open up more opportunities than ever for researchers and developers to test new strategies and port applications to more demanding requirements. Following on the success of PAW16, we invite you to take part in the second PGAS Application Workshop, and to join its vibrant and diverse community of researchers and developers. SCOPE AND AIMS The scope of the PAW workshop is to provide a forum for exhibiting case studies of PGAS programming models in the context of real-world applications as a means of better understanding practical applications of PGAS technologies. We encourage the submission of papers and talks detailing practical PGAS applications, including characterizations of scalability and performance, of expressiveness and programmability, as well as any downsides or areas for improvement in existing PGAS models. In addition to informing other application programmers about the potential that is available through PGAS programming, the workshop is designed to communicate these experiences to compiler vendors, library developers, and system architects in order to achieve broader support for PGAS programming across the community. We also specifically encourage submissions covering big data analytics, deep learning, and other novel and emerging application areas, beyond traditional scientific HPC domains. Topics include, but are not limited to: * Novel application development using the PGAS model. * Real-world examples demonstrating performance, compiler optimization, error checking, and/or reduced software complexity. * Applications from big data analytics, bioinformatics, and other novel areas. * Performance evaluation of applications running under PGAS. * Algorithmic models enabled by PGAS model. * Compiler and runtime environments. * Libraries using/supporting PGAS and applications. * Benefits of hardware abstraction and data locality on algorithm implementation. IMPORTANT DATES: * Submission Deadline: July 31, 2017 * Author Notification: September 1, 2017 * Camera Ready: October 1, 2017 * Workshop Date: November 13, 2017 SUBMISSIONS Submissions are solicited in two categories: a) Full-length papers presenting novel research results: Full-length papers will be published in the workshop proceedings in cooperation with SIGHPC. Submitted papers must be original work that has not appeared in, and is not under consideration for, another conference or a journal. Papers shall not exceed eight (8) pages including text, appendices, and figures. References are not included. b) Extended abstracts summarizing published/preliminary results: Extended abstracts will be evaluated separately and will not be included in the published proceedings; they are intended for timely communications of novel work that is going to be formally submitted elsewhere at a later stage, and/or of already published work that is nonetheless deemed appropriate for dissemination in this venue. Extended abstracts shall not exceed four (4) pages. Submissions shall be submitted through EasyChair (https://easychair.org/conferences/?conf=paw17); they must conform to ACM Guidelines (http://www.acm.org/publications/proceedings-template). Accepted full-length papers will be given longer presentation slots at the workshop; extended abstracts will be given shorter time slots. WORKSHOP CHAIR Karla Morris - Sandia National Laboratories ORGANIZING COMMITTEE * Bradford L. Chamberlain - Cray Inc. * Salvatore Filippone - Cranfield University * Costin Iancu - Lawrence Berkeley National Laboratory * Bill Long - Cray Inc. PROGRAM COMMITTEE CHAIR and CO-CHAIR * Francesco Rizzi - Sandia National Laboratories (Chair) * Bill Long - Cray Inc. (Co-Chair) PROGRAM COMMITTEE * Bradford L. Chamberlain - Cray Inc. * Bert de Jong - Lawrence Berkeley National Laboratory * James Dinan - Intel * Salvatore Filippone - Cranfield University, UK * Jeff Hammond - Intel * Costin Iancu - Lawrence Berkeley National Laboratory * Karla Morris - Sandia National Laboratories * Nicholas Park - U.S. Department of Defense * Anton Shterenlikht - University of Bristol * Min Si - Argonne National Laboratory * Lauren L. Smith - U.S. Department of Defense * Yili Zheng - Google ADVISORY COMMITTEE * Katherine A. Yelick - Lawrence Berkeley National Laboratory * Damian W. I. Rouson - Sourcery Institute In case of questions please email us at: pa...@cr... ----------------------- Lauren L. Smith HPC Computer Science Researcher lau...@ma...<mailto:lau...@ma...> 240-373-6489 |
|
From: David P G. <gr...@us...> - 2016-06-07 19:40:43
|
We are happy to announce the release of X10 and X10DT 2.6.0. The main features of this release are language extensions to enhance support for defining embedded DSLs in X10. In particular, an overloading mechanism for redefining or extending the behavior of control structures was added to X10 and X10 now supports a lightweight trailing closure syntax. The release is available for download at http://x10-lang.org/releases/x10-release-260. |
|
From: Olivier T. <ta...@us...> - 2016-05-06 15:52:26
|
The program for the X10'16 Workshop in Santa Barbara, CA, USA is now available at http://conf.researchr.org/track/pldi-2016/X10-2016-papers#program. The workshop will be held on Tuesday, June 14, 2016, and is colocated with PLDI’16 (http://conf.researchr.org/home/pldi-2016). Registration for the workshop is handled through the PLDI website (http://conf.researchr.org/attending/pldi-2016/Registration). The early registration deadline is May 13. In addition to seven refereed presentations, Dr. Olivier Tardieu from IBM will give a tutorial on X10 programming, and Dr. Sriram Krishnamoorthy from PNNL will give a Keynote presentation entitled “Tracking and Constraining Work Stealing Schedulers.” It will be a great opportunity to connect with the X10 community. Further information can be found at http://x10-lang.org/workshop/workshop16.html. Regards, Claudia Fohry and Olivier Tardieu X10'16 Chairs ---- 9:00-10:00 Session 1: Welcome and Tutorial Introduction to X10 by Olivier Tardieu IBM T.J. Watson Research Center, USA 10:30-12:00 Session 2: Research Papers Control Structure Overloading in X10 by Louis Mandel, Josh Milthorpe, and Olivier Tardieu IBM T.J. Watson Research Center, USA A Memory Model for X10 by Andreas Zwinkau KIT, Germany Cooperation vs. Coordination for Lifeline-Based Global Load Balancing in APGAS by Jonas Posner and Claudia Fohry University of Kassel, Germany 13:30-15:00 Session 3: Keynote and Research Paper Keynote: Tracking and Constraining Work Stealing Schedulers by Sriram Krishnamoorthy Pacific Northwest National Laboratory, USA Resilient X10 over MPI User Level Failure Mitigation by Sara S. Hamouda, Benjamin W. Herta, Josh Milthorpe, David Grove, and Olivier Tardieu Australian National University, Australia; IBM T.J. Watson Research Center, USA 15:30-17:00 Session 4: Research Papers ActorX10: An Actor Library for X10 by Sascha Roloff, Alexander Pöppl, Tobias Schwarzer, Stefan Wildermann, Michael Bader, Michael Glaß, Frank Hannig, and Jürgen Teich University of Erlangen-Nuremberg, Germany; TU Munich, Germany SWE-X10: An Actor-Based and Locally Coordinated Solver for the Shallow Water Equations by Alexander Pöppl and Michael Bader TU Munich, Germany A Case for Distributed Work-Stealing in Regular Applications by Brendan Sheridan and Jeremy Fineman Georgetown University, USA |
|
From: Olivier T. <ta...@us...> - 2016-02-11 19:20:58
|
The 2016 ACM SIGPLAN X10 Workshop (X10'16) co-located with PLDI'16 in Santa Barbara, California, USA Tuesday, June 14, 2016 http://x10-lang.org/workshop/workshop16.html Call for Papers The concurrency and scale-out era is upon us. Application programmers need to confront the architectural challenge of multiples cores and accelerators, clusters and supercomputers. A central need is the development of a usable programming model that can address these challenges -- dealing with thousands of cores and peta-bytes of data. The open-source X10 programming language is designed to address these twin challenges of productivity and performance. It is organized around four basic principles of asynchrony, locality, atomicity and order, developed on a type-safe, class-based, object-oriented foundation. This foundation is robust enough to support fine-grained concurrency, Cilk-style fork-join programming, active messaging, hierarchical and elastic resources, as well as application-level resilience. X10 implementations are available on a wide range of systems ranging from laptops, to clusters, to supercomputers. The X10 Workshop is intended as a forum for X10 programmers, developers, researchers, and educators. We anticipate the program of the workshop to combine keynotes and presentations of selected papers with ample time for discussions. We are soliciting both short papers (4-6 pages) and extended talk abstracts (2 pages). We encourage submissions on all aspects of X10, including theory, design, implementation, practice, curriculum development and experience, applications and tools. This will be a full day workshop. Important Dates Abstracts: Friday, March 11, 2016 (Anywhere on Earth) Submissions: Sunday, March 20, 2016 (Anywhere on Earth) Notification: Friday, April 15, 2016 Final version: Wednesday, April 27, 2016 Workshop: Tuesday, June 14, 2016 Submission Guidelines Papers can be submitted at: https://easychair.org/conferences/?conf=x1016. Submissions should present original research. There are two types of papers: - Short paper: four to six pages in ACM SIGPLAN proceedings style (9-point type, all inclusive), - Extended abstract: two pages in ACM SIGPLAN proceedings style (9-point type, all inclusive). Submissions must be in PDF and printable on US Letter and A4 sized paper. All submissions will be peer-reviewed by the program committee. During the workshop, extended abstracts will receive a shorter presentation and discussion period. Accepted papers will be hosted on the X10 website. Additionally, authors of all accepted papers have the option of including their work in the proceedings that will be published by the ACM up to two weeks before the conference (no further selection process). Topics of interest include, but are not limited to: - Curriculum development using X10 and experience - Applications and experience, X10 programming pearls - High-level frameworks and libraries: map reduce, parallel matrix and - Development and use of APGAS library graph libraries, global load balancing frameworks - Performance analysis, comparison between performance of X10 application in managed environment vs native environment - Foundations: weak-memory models, models of imperative concurrency, reasoning techniques for dynamic concurrency - Extensions: fault-tolerance, dynamic places, hierarchical places - Type systems for concurrency and alias management - Deterministic computation, phased computations -- clock-based concurrency, stream-based computation - Static analyses for atomicity violations, race conditions, deadlock-freedom. - Compilation techniques: code generation, compilation for work-stealing, concurrency and communication optimizations, compilation for scale - Runtime systems, interoperability with Java, MPI - Design and evaluation of JVM extensions for X10 - Distributed GC - Design and experience with development tools (IDEs) for X10 - Performance analysis and monitoring tools - Testing, bug detection and program understanding tools - Debugging frameworks, including large-scale debugging, differential debugging Organizing Committee General Chair: Dr. Olivier Tardieu IBM T.J. Watson Research Center, NY, USA Program Chair: Prof. Dr. Claudia Fohry FG Programmiersprachen/-methodik, Universität Kassel, Germany Program Committee Prof. Dr. Claudia Fohry, Universität Kassel, Germany Prof. Laurie Hendren, McGill University Dr. Vivek Kumar, Rice University, USA Dr. Josh Milthorpe, IBM T.J. Watson Research Center, USA Prof. Nate Nystrom, University of Lugano, Switzerland Dr. Jeeva Paudel, University of Alberta, Canada Dr. Olivier Tardieu, IBM T.J. Watson Research Center, USA Dr. Tomofumi Yuki, INRIA Rhone-Alpes, France |
|
From: David P G. <gr...@us...> - 2015-12-21 20:15:35
|
We are happy to announce the release of X10 and X10DT 2.5.4. The release is available for download at http://x10-lang.org/releases/x10-release-254. X10 Release 2.5.4 HIGHLIGHTS OF THIS RELEASE The main features of this release are improvements in Resilient X10. (a) Significant performance improvements to the implementations of resilient finish to eliminate resiliency overheads of local activity creation. (b) The addition of ULFM-MPI as a network transport for Resilient X10 applications. (c) Enhanced standard library support for writing Resilient X10 applications and frameworks (see x10.util.resilient). LIMITATIONS OF THIS RELEASE The size and index of arrays must be less than 2^31 with Managed X10. An attempt to allocate a longer array causes IllegalArgumentException. The following features described in the 2.5 language manual do not currently work and may be fixed in the subsequent releases: - Non-static type definitions as class or interface members (static type defs do work) The constraint solver used by X10 typechecker is known to be incomplete for situations in which a constraint implies an infinite number of distinct constraints. Additionally, the following features described in the language manual do not currently work with Native X10. - Non-final generic instance methods - Exception stack traces on Cygwin The generated C++ code requires g++ 4.2 or better to be compiled; we do almost all of our testing against g++ 4.4. + On Power/Linux and BG/Q, you may either use g++ 4.2 or better or xlC 11.1 or better. SUMMARY OF ISSUES RESOLVED IN THIS RELEASE Below is a summary of JIRA issues addressed for the X10 2.5.4 Release Notes - X10 - Version X10 2.5.4 ** New Features and Improvements * [XTENLANG-3536] - Indicate cancellation via CancellationException * [XTENLANG-3539] - Update X10DT to Eclipse Luna (4.4) Release * [XTENLANG-3540] - SparsePlaceGroup should support any set of Places; don't require ordering ** Bugs * [XTENLANG-1551] - AST incompleteness -- ClassDecl has no body * [XTENLANG-1834] - Missing comma in parameter list, but compiler error message points to the first valid line of code in the file * [XTENLANG-1841] - Spurious initial error message when method body doens't parse * [XTENLANG-2860] - Compiler crashes on string++ * [XTENLANG-3500] - NativeX10: poor performance of array indexing in loop in Lulesh * [XTENLANG-3501] - Combination of @StackAllocateUninitialized rails and NO_CHECKS yields incorrect results on Lulesh * [XTENLANG-3509] - Problem creating new team from a non-member place * [XTENLANG-3510] - Bug in generated Java code for tuple of generic types * [XTENLANG-3541] - Resilient X10: 'at' unblocked too soon on place failure * [XTENLANG-3545] - Team collectives fail for user-defined types over X10RT=MPI For the details of JIRA issues fixed in this release, see https://xtenlang.atlassian.net/jira/secure/ReleaseNote.jspa?projectId=10005&version=10300 Please use the X10 JIRA to report bugs, after ensuring the problem is not already reported: https://xtenlang.atlassian.net/projects/XTENLANG/issues |
|
From: Vivek K. <vi...@ri...> - 2015-09-10 03:40:53
|
[Apologies if you got multiple copies of this email.] ———————— RESPA 2015 - 1st Workshop on Runtime Systems for Extreme Scale Programming Models and Architectures http://respa15.rice.edu/ <http://respa15.rice.edu/> 16th November 2015 To be in held in conjunction with SC15, Austin ———————— Important Dates Submission deadline: September 11, 2015 (AOE) Notification: October 2, 2015 Final papers for distribution at workshop: October 30, 2015 (AOE) Workshop Date: Monday, November 16, 2015 Call for Papers Extreme-scale and exascale systems impose new requirements on application developers and programming systems to target platforms with hundreds of homogeneous and heterogeneous cores, as well as energy, data movement and resiliency constraints within and across nodes. Runtime systems play a critical role in enabling future programming models, execution models and hardware architectures to address these challenges, and in reducing the widening gap between peak performance and the performance achieved by real applications. The goal of this workshop is to attract leading international researchers to exchange ideas and share their work-in-progress and latest results involving runtime approaches to address these extreme-scale and exascale software challenges. We welcome submissions of original work on all aspects of runtime systems related to parallel programming models and architectures, which includes (but is not limited to) runtime system support for: High-level programming models and domain-specific languages; Scalable intra-node and inter-node scheduling; Scalable coordination and synchronization mechanisms; Memory management across coherence domains and vertical hierarchies of volatile/non-volatile storage; Optimized locality and data movement; Energy management and optimization; Performance tuning and optimization; Resilience and fault-tolerance; Scalable I/O and access to “big data”; Case studies and comparisons of runtime systems. Since the primary goal of this workshop is to foster discussion of cutting edge research related to runtime systems for extreme scale, there will be no formal proceedings for the workshop to ensure that papers presented at the workshop are not disqualified from publication in more formal venues in the future. Instead, all papers will only be posted on the workshop web site, along with the workshop schedule and copies of workshop presentations. Submission Guidelines: We invite two kinds of submissions to this workshop: Full-length research papers (8-page limit) Short papers (4-page limit), which can take the form of position papers, experience reports, or surveys/comparisons of runtime systems All submissions will be peer-reviewed by members of the program committee and external reviewers. We are using HotCRP to manage submissions. The link to submit papers is: http://respa15.cs.rice.edu/ <http://respa15.cs.rice.edu/>. The papers should be formatted according to ACM format (see http://www.acm.org/sigs/publications/proceedings-templates <http://www.acm.org/sigs/publications/proceedings-templates>). For references there is no page limit. Program Committee: Eduard Ayguade (UPC, Spain; PC member) Siegfried Benkner (University of Vienna, Austria; PC member and Workshop Co-Chair) Francois Bodin (IRISA, France; PC member) Ron Brightwell (Sandia National Labs, USA; PC member) Michael Gerndt (TU Munich, Germany; PC member) David Grove (IBM Watson Research Center, USA; PC member) Laxmikant (Sanjay) Kale (University of Illinois at Urbana-Champaign, USA; PC member) Milind Kulkarni (Purdue University, USA; PC member) Vivek Kumar (Rice University, USA; PC member and Publicity Chair) Raymond Namyst (University of Bordeaux, France; PC member) Brian C Van Essen (LLNL, USA; PC member) Vivek Sarkar (Rice University, USA; PC member and Workshop Co-Chair) More information ———————— Further details, as well as a link to the submission site, can be found at the RESPA website: http://respa15.rice.edu/ <http://respa15.rice.edu/> Also, be sure to visit the SC15 site at: http://sc15.supercomputing.org/schedule/event_detail?evid=wksp161 <http://sc15.supercomputing.org/schedule/event_detail?evid=wksp161> |
|
From: Vivek K. <vi...@ri...> - 2015-08-31 15:32:03
|
[Apologies if you got multiple copies of this email.] ———————— RESPA 2015 - 1st Workshop on Runtime Systems for Extreme Scale Programming Models and Architectures http://respa15.rice.edu/ <http://respa15.rice.edu/> 16th November 2015 To be in held in conjunction with SC15, Austin ———————— Important Dates Submission deadline: September 11, 2015 (AOE) Notification: October 2, 2015 Final papers for distribution at workshop: October 30, 2015 (AOE) Workshop Date: Monday, November 16, 2015 Call for Papers Extreme-scale and exascale systems impose new requirements on application developers and programming systems to target platforms with hundreds of homogeneous and heterogeneous cores, as well as energy, data movement and resiliency constraints within and across nodes. Runtime systems play a critical role in enabling future programming models, execution models and hardware architectures to address these challenges, and in reducing the widening gap between peak performance and the performance achieved by real applications. The goal of this workshop is to attract leading international researchers to exchange ideas and share their work-in-progress and latest results involving runtime approaches to address these extreme-scale and exascale software challenges. We welcome submissions of original work on all aspects of runtime systems related to parallel programming models and architectures, which includes (but is not limited to) runtime system support for: High-level programming models and domain-specific languages; Scalable intra-node and inter-node scheduling; Scalable coordination and synchronization mechanisms; Memory management across coherence domains and vertical hierarchies of volatile/non-volatile storage; Optimized locality and data movement; Energy management and optimization; Performance tuning and optimization; Resilience and fault-tolerance; Scalable I/O and access to “big data”; Case studies and comparisons of runtime systems. Since the primary goal of this workshop is to foster discussion of cutting edge research related to runtime systems for extreme scale, there will be no formal proceedings for the workshop to ensure that papers presented at the workshop are not disqualified from publication in more formal venues in the future. Instead, all papers will only be posted on the workshop web site, along with the workshop schedule and copies of workshop presentations. Submission Guidelines: We invite two kinds of submissions to this workshop: Full-length research papers (8-page limit) Short papers (4-page limit), which can take the form of position papers, experience reports, or surveys/comparisons of runtime systems All submissions will be peer-reviewed by members of the program committee and external reviewers. We are using HotCRP to manage submissions. The link to submit papers is: http://respa15.cs.rice.edu/ <http://respa15.cs.rice.edu/>. The papers should be formatted according to ACM format (see http://www.acm.org/sigs/publications/proceedings-templates <http://www.acm.org/sigs/publications/proceedings-templates>). For references there is no page limit. Program Committee: Eduard Ayguade (UPC, Spain; PC member) Siegfried Benkner (University of Vienna, Austria; PC member and Workshop Co-Chair) Francois Bodin (IRISA, France; PC member) Ron Brightwell (Sandia National Labs, USA; PC member) Michael Gerndt (TU Munich, Germany; PC member) David Grove (IBM Watson Research Center, USA; PC member) Laxmikant (Sanjay) Kale (University of Illinois at Urbana-Champaign, USA; PC member) Milind Kulkarni (Purdue University, USA; PC member) Vivek Kumar (Rice University, USA; PC member and Publicity Chair) Raymond Namyst (University of Bordeaux, France; PC member) Brian C Van Essen (LLNL, USA; PC member) Vivek Sarkar (Rice University, USA; PC member and Workshop Co-Chair) More information ———————— Further details, as well as a link to the submission site, can be found at the RESPA website: http://respa15.rice.edu/ <http://respa15.rice.edu/> Also, be sure to visit the SC15 site at: http://sc15.supercomputing.org/schedule/event_detail?evid=wksp161 <http://sc15.supercomputing.org/schedule/event_detail?evid=wksp161> |
|
From: Vivek K. <vi...@ri...> - 2015-08-03 21:11:30
|
[Apologies if you got multiple copies of this email.] ———————— RESPA 2015 - 1st Workshop on Runtime Systems for Extreme Scale Programming Models and Architectures http://respa15.rice.edu/ <http://respa15.rice.edu/> 16th November 2015 To be in held in conjunction with SC15, Austin ———————— Important Dates Submission deadline: September 4, 2015 (AOE) Notification: October 2, 2015 Final papers for distribution at workshop: October 30, 2015 (AOE) Workshop Date: Monday, November 16, 2015 Call for Papers Extreme-scale and exascale systems impose new requirements on application developers and programming systems to target platforms with hundreds of homogeneous and heterogeneous cores, as well as energy, data movement and resiliency constraints within and across nodes. Runtime systems play a critical role in enabling future programming models, execution models and hardware architectures to address these challenges, and in reducing the widening gap between peak performance and the performance achieved by real applications. The goal of this workshop is to attract leading international researchers to exchange ideas and share their work-in-progress and latest results involving runtime approaches to address these extreme-scale and exascale software challenges. We welcome submissions of original work on all aspects of runtime systems related to parallel programming models and architectures, which includes (but is not limited to) runtime system support for: High-level programming models and domain-specific languages; Scalable intra-node and inter-node scheduling; Scalable coordination and synchronization mechanisms; Memory management across coherence domains and vertical hierarchies of volatile/non-volatile storage; Optimized locality and data movement; Energy management and optimization; Performance tuning and optimization; Resilience and fault-tolerance; Scalable I/O and access to “big data”; Case studies and comparisons of runtime systems. Since the primary goal of this workshop is to foster discussion of cutting edge research related to runtime systems for extreme scale, there will be no formal proceedings for the workshop to ensure that papers presented at the workshop are not disqualified from publication in more formal venues in the future. Instead, all papers will only be posted on the workshop web site, along with the workshop schedule and copies of workshop presentations. Submission Guidelines: We invite two kinds of submissions to this workshop: Full-length research papers (8-page limit) Short papers (4-page limit), which can take the form of position papers, experience reports, or surveys/comparisons of runtime systems All submissions will be peer-reviewed by members of the program committee and external reviewers. Program Committee: Eduard Ayguade (UPC, Spain; PC member) Siegfried Benkner (University of Vienna, Austria; PC member and Workshop Co-Chair) Francois Bodin (IRISA, France; PC member) Ron Brightwell (Sandia National Labs, USA; PC member) Michael Gerndt (TU Munich, Germany; PC member) David Grove (IBM Watson Research Center, USA; PC member) Laxmikant (Sanjay) Kale (University of Illinois at Urbana-Champaign, USA; PC member) Milind Kulkarni (Purdue University, USA; PC member) Vivek Kumar (Rice University, USA; PC member and Publicity Chair) Raymond Namyst (University of Bordeaux, France; PC member) Brian C Van Essen (LLNL, USA; PC member) Vivek Sarkar (Rice University, USA; PC member and Workshop Co-Chair) More information ———————— Further details, as well as a link to the submission site (which will be up shortly), can be found at the RESPA website: http://respa15.rice.edu/ <http://respa15.rice.edu/> Also, be sure to visit the SC15 site at: http://sc15.supercomputing.org/schedule/event_detail?evid=wksp161 <http://sc15.supercomputing.org/schedule/event_detail?evid=wksp161> |
|
From: David P G. <gr...@us...> - 2015-07-03 19:23:12
|
[Apologies if you got multiple copies of this email.]
CALL FOR PAPERS
The 9th International Conference on Partitioned
Global Address Space Programming Models (PGAS 2015)
http://pgas.org/conference
September 16-18, 2015
George Washington University
Washington DC, USA
Partitioned Global Address Space (PGAS) programming models offer a shared
address space model that simplifies programming while exposing data thread
locality to enhance performance. This facilitates the development of
programming models that can deliver both productivity and performance. The
PGAS conference is the premier forum to present and discuss ideas and
research developments in the area of: PGAS models, languages, compilers,
runtimes, applications and tools, PGAS architectures and hardware features.
Topics of interest include, but are not limited to:
Architectures: System Architectures, Networks, and Memory Architectures
designed to enhance and enable PGAS programming models.
Applications: New applications that are uniquely enabled by the PGAS
model, existing applications and effective application development
practices for PGAS codes.
Performance: Analysis of application performance over various
programming models.
Developments in Programming Models and Languages: PGAS models, language
extensions, and hybrid models to address emerging architectures, such as
multicore, hybrid, heterogeneous, SIMD and reconfigurable
architectures.
Tools, Compilers, and Implementations: Integrated Development
Environments, performance analysis tools, and debuggers. Compiler
optimizations for PGAS languages, low level libraries, memory
consistency models. Hardware support for PGAS languages, performance
studies and insights, productivity studies, and language
interoperability.
The PGAS Programming Models Conference is dedicated to the presentation and
discussion of research work in this field. Papers should report on original
research, and should include enough background material to make them
accessible to the entire PGAS research community. Papers describing
experiences should indicate how they illustrate general principles; papers
about parallel programming foundations should indicate how they relate to
practice.
Submissions
We are using EasyChair to manage submissions. The link to submit papers is:
https://easychair.org/conferences/?conf=pgas2015.
Full papers should not exceed 10 pages using ACM format with 10pt font.
Each submission must be a single PDF file.
Conference Chairs
Tarek El-Ghazawi, The George Washington University, General Chair
Dhabaleswar K. (DK) Panda, The Ohio State University, Program Chair
Steering Committee
Bill Carlson, Institute for Defense Analyses
Tarek El-Ghazawi, The George Washington University
Lauren Smith, U.S. Government
Kathy Yelick, University of California at Berkeley and LBNL
Program Committee Members
Gheorge Almasi, IBM TJ Watson
Bill Carlson, Institute of Defense Analysis
Brad Chamberlain, Cray Inc.
Pedro Diniz, Univ. of Southern California/Information Science
Institute
Tarek El-Ghazawi, The George Washington University
Salvatore Filippone, Università di Roma "Tor Vergata", Italy
Michael Garland, NVIDIA
Khaled Hamidouche, The Ohio State University
Jeff Hammond, Intel Labs
Paul Hargove, Lawrence Berkeley National Laboratory
Khaled Ibrahim, Lawrence Berkeley National Laboratory
Abdullah Kayi, Intel
Allen Malony, University of Oregon
Steve Poole, United States Department of Defense
Mitsuhisa Sato, University of Tsukuba
Tyler A Simon, Laboratory for Physical Sciences
Lauren Smith, U. S. Government
Sayantan Sur, Intel
Guangming Tan, ICT, Chinese Academy of Sciences, China
Abhinav Vishnu, Pacific Northwest National Laboratory
Michèle Weiland, EPCC, The University of Edinburgh, UK
Yili Zheng, Lawrence Berkeley National Laboratory
Important Dates
· Submission due: July 13th - Extended
· Authors Notification: August 10th
· Camera-ready papers due: August 17th
|
|
From: David P G. <gr...@us...> - 2015-06-23 13:33:03
|
A belated release announcement for X10 and X10DT 2.5.3 which were released right before the X10 workshop last week. We are happy to announce the release of X10 and X10DT 2.5.3. The primary features in this release are improved compiler support for X10DT features. The X10 Editor is now much more resilient to syntax errors: features such as outline, hyperlink, hover help, and content-assist continue to work in the face of compilation errors. The release is available for download at http://x10-lang.org/releases/x10-release-253. The release notes are appended. X10 Release 2.5.3 HIGHLIGHTS OF THIS RELEASE We have switched to using an ANTLR generated parser in the X10 compiler. This change significantly improved compiler support for X10DT features. The X10 Editor is now much more resilient to syntax errors: features such as outline, hyperlink, hover help, and content-assist continue to work in the face of compilation errors. For the content-assist feature to work, the user must first enter a semicolon (';') at the end of the statement to be completed. Managed X10 now supports compression on the network communication links. To enable it, set the system property "X10RT_COMPRESSION" to "snappy" when you launch your program. MINOR LANGUAGE CHANGE To improve the quality of compiler diagnostics for syntax errors and the usability of the X10DT editor, two backwards incompatible syntax changes are being made in X10 2.5.3: (1) There is no implicit return in method definitions (2) The body of a method or constructor cannot be a block introduced by an "=" Therefore, the following method definition is now incorrect: def f (x: long) = { x + 1 } It now should be rewritten to either: def f (x: long) { return x + 1; } or: def f (x: long) = x + 1; A tool converting programs to the new syntax is available: http://x10.sourceforge.net/misc/x10-2.5.3-converter/ LIMITATIONS OF THIS RELEASE The size and index of arrays must be less than 2^31 with Managed X10. An attempt to allocate a longer array causes IllegalArgumentException. The following features described in the 2.5 language manual do not currently work and may be fixed in the subsequent releases: - Non-static type definitions as class or interface members (static type defs do work) The constraint solver used by X10 typechecker is known to be incomplete for situations in which a constraint implies an infinite number of distinct constraints. Additionally, the following features described in the language manual do not currently work with Native X10. - Non-final generic instance methods - Exception stack traces on Cygwin The generated C++ code requires g++ 4.2 or better to be compiled; we do almost all of our testing against g++ 4.4. + On Power/Linux and BG/Q, you may either use g++ 4.2 or better or xlC 11.1 or better. SUMMARY OF ISSUES RESOLVED IN THIS RELEASE Below is a summary of JIRA issues addressed for the X10 2.5.3 Release Notes - X10 - Version X10 2.5.3 ** New Feature * [XTENLANG-3382] - Switch MacOS build of X10DT to 64bit ** Bug * [XTENLANG-3495] - Stack overflow compiling LULESH with x10c -O * [XTENLANG-3496] - Remote reference counting causes DeadPlaceException * [XTENLANG-3503] - Native backend generates invalid C++ code or calls wrong method for function interfaces * [XTENLANG-3507] - Team.WORLD.split causes an error when using openMPI For the details of JIRA issues fixed in this release, see https://xtenlang.atlassian.net/jira/secure/ReleaseNote.jspa?projectId=10005&version=10160 Please use the X10 JIRA to report bugs, after ensuring the problem is not already reported: https://xtenlang.atlassian.net/projects/XTENLANG/issues |
|
From: Ahmad A. <an...@gw...> - 2015-05-19 01:18:08
|
[Apologies if you got multiple copies of this email.] *CALL FOR PAPERS* download pdf <http://hpcl.seas.gwu.edu/pgas15/images/PGAS15_CFP1.pdf> The 9th International Conference on Partitioned Global Address Space Programming Models (PGAS 2015) http://pgas.org/conference September 16-18, 2015 George Washington University Washington DC, USA Partitioned Global Address Space (PGAS) programming models offer a shared address space model that simplifies programming while exposing data thread locality to enhance performance. This facilitates the development of programming models that can deliver both productivity and performance. The PGAS conference is the premier forum to present and discuss ideas and research developments in the area of: PGAS models, languages, compilers, runtimes, applications and tools, PGAS architectures and hardware features. Topics of interest include, but are not limited to: - *Architectures:* System Architectures, Networks, and Memory Architectures designed to enhance and enable PGAS programming models. - *Applications:* New applications that are uniquely enabled by the PGAS model, existing applications and effective application development practices for PGAS codes. - *Performance:* Analysis of application performance over various programming models. - *Developments in Programming Models and Languages:* PGAS models, language extensions, and hybrid models to address emerging architectures, such as multicore, hybrid, heterogeneous, SIMD and reconfigurable architectures. - *Tools, Compilers, and Implementations:* Integrated Development Environments, performance analysis tools, and debuggers. Compiler optimizations for PGAS languages, low level libraries, memory consistency models. Hardware support for PGAS languages, performance studies and insights, productivity studies, and language interoperability. The PGAS Programming Models Conference is dedicated to the presentation and discussion of research work in this field. Papers should report on original research, and should include enough background material to make them accessible to the entire PGAS research community. Papers describing experiences should indicate how they illustrate general principles; papers about parallel programming foundations should indicate how they relate to practice. *Submissions* We are using EasyChair to manage submissions. The link to submit papers will be added to the conference home page in the near future. *Conference Chairs* Tarek El-Ghazawi, The George Washington University, General Chair Dhabaleswar K. Panda, The Ohio State University, Program Chair *Steering Committee* Bill Carlson, Institute for Defense Analyses Tarek El-Ghazawi, The George Washington University Lauren Smith, U.S. Government Kathy Yelick, University of California at Berkeley and LBNL *Important Dates* · Submission due: July 6th · Authors Notification: August 10th · Camera-ready papers due: August 17th |
|
From: Olivier T. <ta...@us...> - 2015-05-13 20:09:35
|
The program for the X10'15 Workshop in Portland, OR is now available at http://conf.researchr.org/track/pldi2015/X10-2015-papers. The workshop will be held on Sunday, June 14, 2015, and is part of FCRC 2015 (http://fcrc.acm.org). Registration for the workshop is handled through the FCRC website (https://www.regonline.com/Register/Checkin.aspx?EventID=1692718). The early registration deadline is May 18. In addition to 8 refereed presentations, Olivier Tardieu will give a tutorial on X10 programming. It will be a great opportunity to connect with the growing X10 community. Further information can be found at http://x10-lang.org/workshop/workshop15.html. Regards, Olivier Tardieu and José Nelson Amaral X10'15 Chairs ---- 9:00-11:00 Session 1: Programming in X10 Opening and Welcome X10 Tutorial by Olivier Tardieu IBM T.J. Watson Research Center The X10 Global Matrix Library: A Resilient Framework for Linear Algebra Applications by Sara Salem Hamouda, Josh Milthorpe, Peter Strazdins and Vijay Saraswat Australian National University IBM T.J. Watson Research Center 11:00-11:20 Coffee break 11:20-12:20 Session 2: Parallel Loops Revisiting Loop Transformations with X10 Clocks by Tomofumi Yuki INRIA / LIP / ENS Lyon Local Parallel Iteration in X10 by Josh Milthorpe IBM T.J. Watson Research Center 12:20-14:00 Lunch 14:00-15:30 Session 3: Compilers and Runtimes Cutting Out the Middleman: OS-Level Support for X10 Activities by Manuel Mohr, Sebastian Buchwald, Andreas Zwinkau, Christoph Erhardt, Benjamin Oechslein, Jens Schedel and Daniel Lohmann Karlsruhe Institute of Technology FAU Erlangen-Nuremberg Optimization of X10 Programs with ROSE Compiler Infrastructure by Michihiro Horie, Mikio Takeuchi, Kiyokuni Kawachiya and David Grove IBM Research - Tokyo IBM T.J. Watson Research Center The APGAS Library: Resilient Parallel and Distributed Programming in Java 8 by Olivier Tardieu IBM T.J. Watson Research Center 15:30-16:00 Coffee break 16:00-17:00 Session 4: Global Load Balancing Towards an Efficient Fault-Tolerance Scheme for GLB by Marco Bungart, Claudia Fohry and Jonas Posner University of Kassel, Germany Scalable Parallel Numerical Constraint Solver Using Global Load Balancing by Daisuke Ishii, Kazuki Yoshizoe and Toyotaro Suzumura Tokyo Institute of Technology Tokyo, Japan JST ERATO Tokyo, Japan IBM Research, JST Dublin, Ireland |
|
From: David P G. <gr...@us...> - 2015-03-09 16:04:14
|
We are happy to announce the release of X10 and X10DT 2.5.2. The primary features of the X10 2.5.2 release are improvements to the x10rt_mpi implementation and class library changes in the x10.util.resilient package and the movement of the implementation class for the X10 Runtime from x10.lang to the new x10.xrx package. The release is available for download at http://x10-lang.org/releases/x10-release-252. The 2.5.2 release notes are appended. HIGHLIGHTS OF THIS RELEASE Although there were no X10 language changes in X10 2.5.2, this release does contain user-visible changes to the structure of the X10 class library. (1) The implementation of the X10 Runtime was moved from the x10.lang package to a new x10.xrx package. Unlike x10.lang, this package is not auto-imported into all programs. Programs that make calls to x10.xrx.Runtime (or other impacted classes) will need to be updated to import the used XRX classes. (2) We continue to improve Resilient/Elastic X10 and have made a number of API improvements to classes in the x10.util.resilient package. (3) We have added built-in support for ghost regions to the x10.regionarray.DistArray class. Some mostly internal changes to the DistArray API were made to enable this new functionality (XTENLANG-3488,3489). This release also contains significant enhancements to the x10rt_mpi implementation, including exploitation of MPI-3 non-blocking collections by X10 Team operations to enable unrestricted mixing of APGAS asynchronous messaging with Team collective operations. This release is also the first time we are bundling the Lulesh proxy application with the X10 benchmarks. We continue to improve our Lulesh code, but believe it has now reached a level of stability and performance that make it generally usable as a X10 benchmark. LIMITATIONS OF THIS RELEASE The size and index of arrays must be less than 2^31 with Managed X10. An attempt to allocate a longer array causes IllegalArgumentException. The following features described in the 2.5 language manual do not currently work and may be fixed in the subsequent releases: - Non-static type definitions as class or interface members (static type defs do work) The constraint solver used by X10 typechecker is known to be incomplete for situations in which a constraint implies an infinite number of distinct constraints. Additionally, the following features described in the language manual do not currently work with Native X10. - Non-final generic instance methods - Exception stack traces on Cygwin The generated C++ code requires g++ 4.2 or better to be compiled; we do almost all of our testing against g++ 4.4. + On Power/Linux and BG/Q, you may either use g++ 4.2 or better or xlC 11.1 or better. SUMMARY OF ISSUES RESOLVED IN THIS RELEASE Below is a summary of JIRA issues addressed for the X10 2.5.2 Release Notes - X10 - Version X10 2.5.2 ** New Features and Improvements * [XTENLANG-1373] - Support for ghost cells in distributed arrays * [XTENLANG-3447] - Splittable PNRG * [XTENLANG-3479] - Add FileReader.offset() * [XTENLANG-3480] - GML: add rank-one update (DGER) * [XTENLANG-3484] - Move XRX runtime to x10.xrx package (out of x10.lang) * [XTENLANG-3488] - Replace Dist.offset with Region.indexOf * [XTENLANG-3490] - x10rt_mpi should default to MPI_THREAD_MULTIPLE ** Bugs * [XTENLANG-3354] - asyncCopy deadlocks on BG/Q * [XTENLANG-3355] - alltoall segfaults on BG/Q * [XTENLANG-3485] - On MacOs the jblas library must have extension .jnilib and not .so * [XTENLANG-3486] - GML: transpose DenseMatrix.T(DenseMatrix) should store to invocation target * [XTENLANG-3489] - Replace x10.regionarray.PeriodicDist with x10.regionarray.PeriodicBoundaryConditions * [XTENLANG-3492] - Team creation fails over SparsePlaceGroup For the details of JIRA issues fixed in this release, see http://jira.codehaus.org/secure/ReleaseNote.jspa?projectId=11812&version=20767 Please use the X10 JIRA to report bugs, after ensuring the problem is not already reported: http://jira.codehaus.org/secure/IssueNavigator.jspa?reset=true&&pid=11812&resolution=-1 |
|
From: Olivier T. <ta...@us...> - 2015-02-05 14:49:56
|
X10 Workshop (X10'15)
co-located with PLDI'15
Portland, Oregon, United States
Sunday, June 14, 2015
http://x10-lang.org/workshop/workshop15.html
Call for Papers
The concurrency and scale-out era is upon us. Application programmers need
to confront the architectural challenge of multiples cores and
accelerators, clusters and supercomputers. A central need is the
development of a usable programming model that can address these
challenges -- dealing with thousands of cores and peta-bytes of data.
The open-source X10 programming language is designed to address these twin
challenges of productivity and performance. It is organized around four
basic principles of asynchrony, locality, atomicity and order, developed
on a type-safe, class-based, object-oriented foundation. This foundation
is robust enough to support fine-grained concurrency, Cilk-style fork-join
programming, GPU programming, SPMD computations, active messaging,
MPI-style communicators and cluster programming. X10 implementations are
available on a wide range of systems ranging from laptops, to clusters, to
supercomputers.
The X10 Workshop is intended as a forum for X10 programmers, developers,
researchers, and educators. We anticipate the program of the workshop to
combine keynotes and presentations of selected papers with ample time for
discussions. We are soliciting both short papers (4-6 pages) and extended
talk abstracts (2 pages). We encourage submissions on all aspects of X10,
including theory, design, implementation, practice, curriculum development
and experience, applications and tools. This will be a full day workshop.
Important Dates
Abstracts: Friday, March 13th, 2015 (Anywhere on Earth)
Submissions: Friday, March 20th, 2015 (Anywhere on Earth)
Notification: Friday, April 17th, 2015
Final version: Friday, May 15th, 2015
Workshop: Sunday, June 14th, 2015
Submission Guidelines
Papers can be submitted at: https://easychair.org/conferences/?conf=x1015.
Submissions may be one of the following:
Short paper: four to six pages in ACM SIGPLAN proceedings style (9-point
type, all inclusive),
Extended abstract: two pages in ACM SIGPLAN proceedings style (9-point
type, all inclusive).
Submissions must be in PDF and printable on US Letter and A4 sized paper.
All submissions will be peer-reviewed by the program committee. During the
workshop, extended abstracts will receive a shorter presentation and
discussion period.
Accepted papers will be hosted on the X10 website. Accepted authors will
have the option of having their paper in the proceedings that will be
published by the ACM.
Topics of interest include, but are not limited to:
- Curriculum development using X10 and experience
- Applications and experience, X10 programming pearls
- High-level frameworks and libraries: map reduce, parallel matrix and
graph libraries, global load balancing frameworks
- Performance analysis, comparison between performance of X10 application
in managed environment vs native environment
- Foundations: weak-memory models, models of imperative concurrency,
reasoning techniques for dynamic concurrency
- Extensions: fault-tolerance, dynamic places, hierarchical places
- Type systems for concurrency and alias management
- Deterministic computation, phased computations -- clock-based
concurrency, stream-based computation
- Static analyses for atomicity violations, race conditions,
deadlock-freedom
- Compilation techniques: code generation, compilation for work-stealing,
concurrency and communication optimizations, compilation for scale
- Runtime systems, interoperability with Java, MPI
- Design and evaluation of JVM extensions for X10
- Distributed GC
- Design and experience with development tools (IDEs) for X10
- Performance analysis and monitoring tools
- Testing, bug detection and program understanding tools
- Debugging frameworks, including large-scale debugging, differential
debugging
Organizing Committee
General Chair: Olivier Tardieu, IBM T.J. Watson Research Center, NY, USA
Program Chair: Prof. José Nelson Amaral, Department of Computing Science,
University of Alberta, AB, Canada
Program Committee
José Nelson Amaral, University of Alberta (Chair)
Tiago Cogumbreiro, Imperial College
Alain Darte, Ecole Normale Supérieure de Lyon
Claudia Fohry, Universität Kassel
Laurie Hendren, McGill University
Jens Palsberg, University of California
Avraham Shinnar, IBM T.J. Watson Research Center
Olivier Tardieu, IBM T.J. Watson Research Center
|
|
From: David P G. <gr...@us...> - 2014-12-04 22:45:44
|
We are happy to announce the release of X10 and X10DT 2.5.1. The primary features in this release are enhancements to Resilient and Elastic X10. The release is available for download at http://x10-lang.org/releases/x10-release-251. X10 Release 2.5.1 HIGHLIGHTS OF THIS RELEASE The primary features of the X10 2.5.1 release are enhancements to Resilient and Elastic X10. Key highlights include: + Full integration of the Hazelcast in-memory data grid with Managed X10. Key features include new X10 APIs to utilize Hazelcast as an implementation of x10.util.resilient.ResilientMap, an implementation of resilient finish using Hazelcast (X10_RESILIENT_MODE=12), and integration of Hazelcast grid initialization into the X10 launcher (bin/x10 script). + Resilient X10 using Place 0 resilient finish (X10_RESILIENT_MODE=1) can now support the same levels of concurrency and distribution as non-resilient X10. This was enabled by new features in the X10 runtime and network layer for message stratification and stronger progress guarantees for explicitly marked short messages (see x10.compiler.Immediate). + The development of additional Resilient and Elastic X10 sample applications. Resiliency support has been added to the Global Matrix Library (GML) and several of its sample applications. Resiliency and Elasticity have also been added to M3RLite (Main Memory Map Reduce Lite) and several new M3RLite sample applications are now available as well. For additional information about Resilient and Elastic X10, please see samples/resiliency and the x10-lang.org website. LIMITATIONS OF THIS RELEASE The size and index of arrays must be less than 2^31 with Managed X10. An attempt to allocate a longer array causes IllegalArgumentException. The following features described in the 2.5 language manual do not currently work and may be fixed in the subsequent releases: - Non-static type definitions as class or interface members (static type defs do work) The constraint solver used by X10 typechecker is known to be incomplete for situations in which a constraint implies an infinite number of distinct constraints. Additionally, the following features described in the language manual do not currently work with Native X10. - Non-final generic instance methods - Exception stack traces on Cygwin The generated C++ code requires g++ 4.2 or better to be compiled; we do almost all of our testing against g++ 4.4. + On Power/Linux and BG/Q, you may either use g++ 4.2 or better or xlC 11.1 or better. SUMMARY OF ISSUES RESOLVED IN THIS RELEASE Below is a summary of JIRA issues addressed for the X10 2.5.1 ** New Features and Improvements * [XTENLANG-3333] - X10 WebSite: Add information on current community projects under community * [XTENLANG-3443] - ArrayList.resize * [XTENLANG-3445] - RailUtils.scan * [XTENLANG-3456] - Create as many BDWGC marker threads as X10 worker threads * [XTENLANG-3457] - Reduce memory usage of X10RT sockets launcher processes * [XTENLANG-3461] - Map / reduce methods on local Matrix and Vector classes * [XTENLANG-3468] - Team reduce, allreduce for user-defined Arithmetic/Bitwise/Ordered struct types * [XTENLANG-3475] - GML: Wrap LAPACK eigensolver DSYEVX instead of DSYEV ** Bugs * [XTENLANG-3304] - On Resilient X10, at (p) { body } may block even if the body finished * [XTENLANG-3343] - Resilient X10 fails with "TOO MANY THREADS" * [XTENLANG-3344] - Resilient X10 fails with "TOO MANY THREADS" (another case) * [XTENLANG-3379] - On Resilient X10, at (p) async is not executed in parallel without X10_BUSY_WAITING * [XTENLANG-3428] - Elastic X10 and Runtime.Workers.multiplace * [XTENLANG-3435] - @NoInline on closure within static @Inline method results in wrong saved_this * [XTENLANG-3437] - Experimental Java MPI launcher causes SEGV with non-IBM Java VM * [XTENLANG-3438] - GC library and build system patch for K computer * [XTENLANG-3442] - Some structs can't be interface's members's return types * [XTENLANG-3462] - Incorrect implementation of instanceof and cast for nested generic types * [XTENLANG-3463] - Resilient X10 requires BUSY_WAITING=true * [XTENLANG-3466] - Template-related compile error in Native X10 * [XTENLANG-3470] - Incorrect generated code for variable shadowing ** Tasks * [XTENLANG-3439] - Update sample programs (online; X10DT help) for 2.5.1 For the details of JIRA issues fixed in this release, see http://jira.codehaus.org/secure/ReleaseNote.jspa?projectId=11812&version=20503 Please use the X10 JIRA to report bugs, after ensuring the problem is not already reported: http://jira.codehaus.org/secure/IssueNavigator.jspa?reset=true&&pid=11812&resolution=-1 |
|
From: Olivier S. <se...@gw...> - 2014-11-13 04:30:03
|
All, Are you working with Partitioned Global Address Space technology? Do you have a PGAS project that you want to give a quick update on? Even better, do you have applications that you want to share that use PGAS programming models? The PGAS BOF at SC14 will have a substantial part devoted to Quick Update talks. Our emphasis this year is on applications, so if you are going to talk about an application of PGAS technology, please submit 2 slides. If you are going to talk about programming languages, PGAS technology, or PGAS tools, please submit 1 slide. We will compile all the slides together and during the BOF, when your slides come up, you stand up and talk about your project and/or application! Please note: If we get an overwhelming number of talks, we may not be able to have all of them -- so we reserve the right to prune some. This has not been a problem in the past, though last year came close! Venue: SC 2014 in New Orleans Subject: PGAS BOF When: Wednesday, November 19 @ 12:15pm Where: Room 273 What: PGAS Quick Update Slides Please send Powerpoint slides to: se...@gw... by 9am, November 18! Thanks, Olivier Serres, Tarek El-Ghazawi, and Lauren Smith |
|
From: David P G. <gr...@us...> - 2014-10-03 17:30:20
|
We are happy to announce the release of X10 and X10DT 2.5.0. This release of X10 includes enhancements to the X10 runtime and changes to the X10 standard library in support of our ongoing Resilient and Elastic X10 work. The release is available for download at http://x10-lang.org/releases/x10-release-250. The X10 2.5.0 Release Notes are appended. X10 Release 2.5.0 HIGHLIGHTS OF THIS RELEASE X10 Release 2.5.0 includes a redesign of several X10 standard library APIs to better support Resilient and Elastic X10. In particular, we have rethought various aspects of Place-related functionality to more naturally support execution over a dynamically varying set of Places. For all the details, please see the X10Doc for x10.lang.Place, x10.lang.PlaceGroup, and x10.lang.PlaceTopology. Some of the key changes are: + The removal of static constants such as PlaceGroup.WORLD, and Place.MAX_PLACES. They are replaced by Place.places() and Place.numPlaces() which return values that represent the current view on the dynamically changeable set of Places available to the computation. + The removal of iteration functionality (next/prev) from Place. This functionality is now provided only through PlaceGroup. + The addition of PlaceTopology to provide a more flexible set of APIs describing the topological relationships of Places. X10 2.5.0 also includes a set of enabling runtime system changes to better support the embedding of X10 and the XRX runtime into larger programs written in other programming languages. These changes are surfaced to the programmer via a new set of APIs on x10.lang.Runtime that support initializing the XRX runtime, submitting jobs for execution by the XRX runtime, job cancellation, and runtime shutdown. A backwards incompatible change to the x10.util.Map interface removes the usage of x10.util.Box on the return value of the put and get methods of Map by requiring that Values stored in Map satisfy the haszero constraint (XTENLANG-573). This may require adjustment in application code that uses implementation of x10.util.Map. For additional information about Resilient and Elastic X10, please see samples/resiliency and the x10-lang.org website. LIMITATIONS OF THIS RELEASE The size and index of arrays must be less than 2^31 with Managed X10. An attempt to allocate a longer array causes IllegalArgumentException. The following features described in the 2.5 language manual do not currently work and may be fixed in the subsequent releases: - Non-static type definitions as class or interface members (static type defs do work) The constraint solver used by X10 typechecker is known to be incomplete for situations in which a constraint implies an infinite number of distinct constraints. Additionally, the following features described in the language manual do not currently work with Native X10. - Non-final generic instance methods - Exception stack traces on Cygwin The generated C++ code requires g++ 4.2 or better to be compiled; we do almost all of our testing against g++ 4.4. + On Power/Linux and BG/Q, you may either use g++ 4.2 or better or xlC 11.1 or better. SUMMARY OF ISSUES RESOLVED IN THIS RELEASE Below is a summary of JIRA issues addressed for the X10 2.5.0 ** New Features and Improvements * [XTENLANG-573] - x10.util.HashMap & Boxing * [XTENLANG-1955] - Calling external MPI Library(PWSMP) on multiple nodes has performance issue * [XTENLANG-2493] - static val members with constant values should be const in generated C++ * [XTENLANG-2866] - Modernize x10.util.Map (and others) to provide operators as shorthands * [XTENLANG-3181] - Ensure graceful failure of X10 runtime in presence of failures * [XTENLANG-3302] - Faster put, remove for MapSet * [XTENLANG-3394] - Upgrade to bdwgc 7.4.x * [XTENLANG-3399] - Rail.uncountedCopy for Managed X10 * [XTENLANG-3401] - X10 implementation of Team reduce, allreduce * [XTENLANG-3411] - Native X10: codegen for string literals * [XTENLANG-3414] - GML arbitrary place group distribution * [XTENLANG-3416] - Need Watcher.await with timeout * [XTENLANG-3423] - Change MultipleExceptions.exceptions to Rail [CheckedThrowable] * [XTENLANG-3429] - Change default value of X10_NTHREADS from 1 to availableProcessors * [XTENLANG-3430] - X10 implementation of Team scatter ** Bugs * [XTENLANG-3051] - Cannot pass "INIT_THREADS" flag to Java runtime * [XTENLANG-3312] - Rail.asyncCopy sometime fails on large input * [XTENLANG-3313] - x10c should detect recursive jar creation * [XTENLANG-3371] - Race condition in Runtime.Workers.multiplace * [XTENLANG-3372] - Intermittent stack overflow in polyglot * [XTENLANG-3390] - Allow X10 values to be serialized via Java serialization * [XTENLANG-3392] - Initial activity does not drop its clocks upon terminating. * [XTENLANG-3395] - Spurious wake-up in x10.util.concurrent.SPMDBarrier causes deadlock * [XTENLANG-3396] - Fail to compile a static method in Native X10 * [XTENLANG-3398] - Evaluation of Team.WORLD hangs on multi places with JavaSockets * [XTENLANG-3400] - PlaceLocalHandles: remove dependency on Place 0 for id assignment * [XTENLANG-3408] - Performance of JavaSockets transport * [XTENLANG-3413] - Native X10: closures in generic methods that do not capture type params should go in .cc not .h * [XTENLANG-3417] - Support generation of a jar file which includes no X10/Java source files * [XTENLANG-3418] - Reader#{mark(Int),available():Int} should use Long rather than Int * [XTENLANG-3421] - C++ post compilation error (missing inclusion of x10/lang/Float.h) * [XTENLANG-3425] - Inconsistencies between Managed and Native implementations of zero-size Rail.asyncCopy, Rail.uncountedCopy * [XTENLANG-3426] - Update online help contents for X10DT 2.5 * [XTENLANG-3431] - Killing second place results in program termination * [XTENLANG-3432] - Fix to experimental Java MPI launcher ** Tasks * [XTENLANG-3201] - Need official release of GML For the details of JIRA issues fixed in this release, see http://jira.codehaus.org/secure/ReleaseNote.jspa?projectId=11812&version=19141 Please use the X10 JIRA to report bugs, after ensuring the problem is not already reported: http://jira.codehaus.org/secure/IssueNavigator.jspa?reset=true&&pid=11812&resolution=-1 |
|
From: David P G. <gr...@us...> - 2014-05-23 19:07:42
|
We're pleased to announce that X10 and X10DT 2.4.3 are now available for download at http://x10-lang.org/releases/x10-release-243. Highlights of the X10 2.4.3 Release Notes are appended: HIGHLIGHTS OF THIS RELEASE A major new feature in X10DT 2.4.3 is support for basic source-level debugging of X10 programs running on Managed X10. With this release debugging operations that rely on source line information (e.g. breakpoints, single stepping) can be done at the X10 source level. Debugging operations like variable inspection can only be done at the level of the generated Java code (due to limitations of JSR-045 SMAP mappings). To try this new feature, simply launch a Managed X10 program using the usual Eclipse debugging procedures. In previous versions of X10, when a Place had no activities to execute the XRX runtime would consume CPU by actively pooling the network for incoming activities. In X10 2.4.3 the default has been changed to instead block the polling thread on the network, thus allowing idle X10 Places to not consume CPU resources. If the previous behavior is desired (active polling), set the environment variable X10_BUSY_WAITING to true. SUMMARY OF ISSUES RESOLVED IN THIS RELEASE Below is a summary of JIRA issues addressed for the X10 2.4.3 ** New Features and Improvements * [XTENLANG-1012] - Busy wait in x10::runtime::Pool::scan() * [XTENLANG-3367] - Handle non-trivial at() expressions in CUDA kernels * [XTENLANG-3370] - generate SMAPI debug info in command line builds * [XTENLANG-3385] - GML: add Manhattan, Chebyshev distances to x10.matrix.Vector * [XTENLANG-3386] - New class x10.util.Date ** Bugs * [XTENLANG-3219] - Exceptions during serialization/deserialization can hang XRX * [XTENLANG-3368] - Resilient X10: failure to reliably raise DPE when System.killHere is called * [XTENLANG-3374] - Team: illegal broadcastFlat within atomic block * [XTENLANG-3376] - bugs in x10.io.File.list() * [XTENLANG-3377] - Team.WORLD.barrier() intermittent hang with X10RT MPI * [XTENLANG-3383] - Elements of Rail[struct] not initialized to zero * [XTENLANG-3384] - Incomplete type error on cast to constrained type in postcompile * [XTENLANG-3387] - Machine architecture unknown for X10RT MPI bindings ** Tasks * [XTENLANG-3332] - Update install instructions on website |
|
From: Mikio T. <mik...@gm...> - 2014-05-02 11:54:57
|
The program for the X10'14 Workshop in Edinburgh, UK is now available (http://x10-lang.org/workshop/workshop14/x10-2014-program.html). The workshop will be held on Thursday, June 12, 2014, and is co-located with PLDI. In addition to 7 refereed presentations, Olivier Tardieu will give a tutorial on X10 programming, and David Cunningham will give a keynote on Resilient X10. It will be a great opportunity to connect with the growing X10 community. Further information can be found at http://x10-lang.org/workshop/workshop14.html Regards, Mikio Takeuchi X10'14 Program Chair ---- 9:00-10:15 Session 1: Introduction X10 Tutorial by Olivier Tardieu IBM T.J. Watson Research Center 10:15-10:45 Coffee break 10:45-12:00 Session 2: X10 applications Porting MPI based HPC Applications to X10 by Hiroki Murata, Michihiro Horie, Koichi Shirahata, Jun Doi, Hideki Tai, Mikio Takeuchi and Kiyokuni Kawachiya IBM Research - Tokyo and Tokyo Institute of Technology Performance Analysis of Lattice QCD Application with APGAS Programming Model by Koichi Shirahata, Jun Doi and Mikio Takeuchi Tokyo Institute of Technology and IBM Research - Tokyo Dynamic X10: Resource-Aware Programming for Higher Efficiency by Matthias Braun, Sebastian Buchwald, Manuel Mohr and Andreas Zwinkau Karlsruhe Institute of Technology 12:00-13:30 Lunch 13:30-14:45 Session 3: Resilient X10 Keynote: Resilient X10 (50 minutes) by David Cunningham Google Inc. Writing Fault-Tolerant Applications Using Resilient X10 by Kiyokuni Kawachiya IBM Research - Tokyo 14:45-15:15 Coffee break 15:15-16:30 Session 4: Tooling and scheduling Toward a profiling tool for visualizing implicit behavior in X10 by Seisei Itahashi, Yoshiki Sato and Shigeru Chiba The University of Tokyo Armus: dynamic deadlock verification for barriers by Tiago Cogumbreiro, Raymond Hu, Francisco Martins and Nobuko Yoshida Imperial College London and University of Lisbon A Case for Cooperative Scheduling in X10's Managed Runtime by Shams Imam and Vivek Sarkar Rice University |
|
From: Mikio T. <mik...@gm...> - 2014-03-13 05:10:34
|
The 2014 X10 Workshop (X10'14) co-located with PLDI'14 in Edinburgh, UK Thursday, June 12, 2014 http://x10-lang.org/workshop/workshop14.html Note: Paper deadline is Extended! Abstract submission deadline: Friday, March 14th, 2014 (Anywhere on Earth) Paper submission deadline: Friday, March 21st, 2014 (Anywhere on Earth) Papers can be submitted at: https://www.easychair.org/conferences/?conf=x102014 Call for Papers The concurrency and scale-out era is upon us. Application programmers need to confront the architectural challenge of multiples cores and accelerators, clusters and supercomputers. A central need is the development of a usable programming model that can address these challenges -- dealing with thousands of cores and peta-bytes of data. The open-source X10 programming language is designed to address these twin challenges of productivity and performance. It is organized around four basic principles of asynchrony, locality, atomicity and order, developed on a type-safe, class-based, object-oriented foundation. This foundation is robust enough to support fine-grained concurrency, Cilk-style fork-join programming, GPU programming, SPMD computations, active messaging, MPI-style communicators and cluster programming. X10 implementations are available on a wide range of systems ranging from laptops, to clusters, to supercomputers. The X10 Workshop is intended as a forum for X10 programmers, developers, researchers, and educators. We anticipate the program of the workshop to combine keynotes and presentations of selected papers with ample time for discussions. We are soliciting both short papers (4-6 pages) and extended talk abstracts (2 pages). We encourage submissions on all aspects of X10, including theory, design, implementation, practice, curriculum development and experience, applications and tools. This will be a full day workshop. Important Dates Abstract submission deadline: Friday, March 14th, 2014 (Anywhere on Earth) Paper submission deadline: Friday, March 21st, 2014 (Anywhere on Earth) Author notification: Friday, April 18th, 2014 Final version deadline: Friday, May 9th, 2014 Workshop: Thursday, June 12th, 2014 Submission Guidelines Papers can be submitted at: https://www.easychair.org/conferences/?conf=x102014 Submissions may be one of the following: - Short paper: four to six pages in ACM SIGPLAN proceedings style (9-point type, all inclusive), - Extended abstract: two pages in ACM SIGPLAN proceedings style (9-point type, all inclusive). Submissions must be in PDF and printable on US Letter and A4 sized paper. All submissions will be peer-reviewed by the program committee. During the workshop, extended abstracts will receive a shorter presentation and discussion period. To encourage the presentation and discussion of on-going work and preliminary results that can subsequently be published as full conference papers, we will not publish papers from the workshop. The revised short papers and extended abstracts from the presenters will be available to the workshop participants and others through the workshop website. Topics of interest include, but are not limited to: - Curriculum development using X10 and experience - Applications and experience, X10 programming pearls - High-level frameworks and libraries: map reduce, parallel matrix and graph libraries, global load balancing frameworks - Performance analysis, comparison between performance of X10 application in managed environment vs native environment - Foundations: weak-memory models, models of imperative concurrency, reasoning techniques for dynamic concurrency - Extensions: fault-tolerance, dynamic places, hierarchical places - Type systems for concurrency and alias management - Deterministic computation, phased computations -- clock-based concurrency, stream-based computation - Static analyses for atomicity violations, race conditions, deadlock-freedom. - Compilation techniques: code generation, compilation for work-stealing, concurrency and communication optimizations, compilation for scale - Runtime systems, interoperability with Java, MPI - Design and evaluation of JVM extensions for X10 - Distributed GC - Design and experience with development tools (IDEs) for X10 - Performance analysis and monitoring tools - Testing, bug detection and program understanding tools - Debugging frameworks, including large-scale debugging, differential debugging Organizing Committee General Chair: Olivier Tardieu, IBM T.J. Watson Research Center, USA Program Chair: Mikio Takeuchi, IBM Research - Tokyo, Japan Program Committee Rajkishore Barik, Intel Labs Vincent Cavé, Rice University Tomio Kamada, Kobe University / RIKEN AICS Manuel Mohr, Karlsruhe Institute of Technology Toyotaro Suzumura, IBM Research - Ireland / University College Dublin Mikio Takeuchi, IBM Research - Tokyo (chair) Olivier Tardieu, IBM T.J. Watson Research Center Mikio Takeuchi IBM Research - Tokyo |
|
From: David P G. <gr...@us...> - 2014-02-13 13:17:59
|
We're pleased to announce that X10 and X10DT 2.4.2 are now available for download at http://x10-lang.org/releases/x10-release-242. X10 Release 2.4.2 HIGHLIGHTS OF THIS RELEASE There were no significant language changes or incompatible class library API changes made in this release of X10. An implementation of the X10 GlobalLoadBalancing framework (GLB) has been incorporated into the x10 standard library in the x10.glb package. Additional sample X10 programs using GLB are available in samples/GLB. We have completed the work on a (single host) pure Java launcher for Managed X10. This allows multi-place execution on Managed X10 on a single Windows host without requiring cygwin. This capability is available from X10DT when launching a Managed X10 application on a local Windows host. Support for the closed source x10rt_pgas implementation and the BlueGene/P platform was removed from X10 starting with this release. Several bugs in the CUDA backend present in X10 2.4.0 and 2.4.1 were corrected in this release. All CUDA sample programs now compile and run successfully using the latest (v5.5) release of the NVIDIA CUDA Toolkit. LIMITATIONS OF THIS RELEASE The size and index of arrays must be less than 2^31 with Managed X10. An attempt to allocate a longer array causes IllegalArgumentException. The following features described in the 2.4 language manual do not currently work and may be fixed in the subsequent releases: - Non-static type definitions as class or interface members (static type defs do work) The constraint solver used by X10 typechecker is known to be incomplete for situations in which a constraint implies an infinite number of distinct constraints. Additionally, the following features described in the language manual do not currently work with Native X10. - Garbage collection on AIX - Non-final generic instance methods - Exception stack traces on Cygwin Although greatly improved from previous releases, the X10 runtime still uses a modified "busy wait" loop in the worker threads that execute asyncs. A consequence of this is that even if a Place has no asyncs to execute, one worker thread in the Place will still consume CPU cycles busy waiting for messages to arrive from the network. The generated C++ code requires g++ 4.2 or better to be compiled; we do almost all of our testing against g++ 4.4. + On AIX and Power/Linux, you may either use g++ 4.2 or better or xlC 11.1 or better. SUMMARY OF ISSUES RESOLVED IN THIS RELEASE Below is a summary of JIRA issues addressed for the X10 2.4.2 ** New Features and Improvement * [XTENLANG-2970] - Editing pass over X10 website updating for X10 2.4 * [XTENLANG-3276] - Incorporate Global Load Balancing Framework (GLB) into X10 standard library * [XTENLANG-3325] - Pure Java launcher for Managed X10 (single node) * [XTENLANG-3335] - Add Synthetic annotation to classes, fields, or methods that don't appear in the source code * [XTENLANG-3336] - Hide synthetic classes, methods, and fields from x10doc * [XTENLANG-3338] - Use statement expressions instead of apply of closure literals in internal expansions of ASTs * [XTENLANG-3339] - Remove x10rt_pgas and BlueGene/P support * [XTENLANG-3345] - File should have a useful toString method * [XTENLANG-3359] - Implement map and reduce for x10.array.DistArray ** Bugs * [XTENLANG-3299] - Compiler NPE when optimizing because of incorrect implementation of alpha renaming * [XTENLANG-3334] - Method/constructor guard is not printed properly * [XTENLANG-3346] - x10c with Java8 dislikes methods returning java.lang.Class * [XTENLANG-3347] - x10doc doesn't work with Java8 * [XTENLANG-3348] - Cannot define an X10 class that implements a java interface that extends java.lang.Comparable * [XTENLANG-3357] - Resilient X10 fails with ClockUseException * [XTENLANG-3362] - Closure captures transient instance field directly instead of 'this' For the details of JIRA issues fixed in this release, see http: http://jira.codehaus.org/secure/ReleaseNote.jspa?version=19827&projectId=11812 Please use the X10 JIRA to report bugs, after ensuring the problem is not already reported: http://jira.codehaus.org/secure/IssueNavigator.jspa?reset=true&&pid=11812&resolution=-1 |
|
From: Mikio T. <mik...@gm...> - 2014-02-13 04:34:21
|
The 2014 X10 Workshop (X10'14) co-located with PLDI'14 in Edinburgh, UK Thursday, June 12, 2014 http://x10-lang.org/workshop/workshop14.html Call for Papers The concurrency and scale-out era is upon us. Application programmers need to confront the architectural challenge of multiples cores and accelerators, clusters and supercomputers. A central need is the development of a usable programming model that can address these challenges -- dealing with thousands of cores and peta-bytes of data. The open-source X10 programming language is designed to address these twin challenges of productivity and performance. It is organized around four basic principles of asynchrony, locality, atomicity and order, developed on a type-safe, class-based, object-oriented foundation. This foundation is robust enough to support fine-grained concurrency, Cilk-style fork-join programming, GPU programming, SPMD computations, active messaging, MPI-style communicators and cluster programming. X10 implementations are available on a wide range of systems ranging from laptops, to clusters, to supercomputers. The X10 Workshop is intended as a forum for X10 programmers, developers, researchers, and educators. We anticipate the program of the workshop to combine keynotes and presentations of selected papers with ample time for discussions. We are soliciting both short papers (4-6 pages) and extended talk abstracts (2 pages). We encourage submissions on all aspects of X10, including theory, design, implementation, practice, curriculum development and experience, applications and tools. This will be a full day workshop. Important Dates Submission: Friday, March 14th, 2014 (Anywhere on Earth) Notification: Friday, April 18th, 2014 Final version: Friday, May 9th, 2014 Workshop: Thursday, June 12th, 2014 Submission Guidelines Submissions instructions will be available at: http://x10-lang.org/workshop/workshop14.html Submissions may be one of the following: - Short paper: four to six pages in ACM SIGPLAN proceedings style (9-point type, all inclusive), - Extended abstract: two pages in ACM SIGPLAN proceedings style (9-point type, all inclusive). Submissions must be in PDF and printable on US Letter and A4 sized paper. All submissions will be peer-reviewed by the program committee. During the workshop, extended abstracts will receive a shorter presentation and discussion period. To encourage the presentation and discussion of on-going work and preliminary results that can subsequently be published as full conference papers, we will not publish papers from the workshop. The revised short papers and extended abstracts from the presenters will be available to the workshop participants and others through the workshop website. Topics of interest include, but are not limited to: - Curriculum development using X10 and experience - Applications and experience, X10 programming pearls - High-level frameworks and libraries: map reduce, parallel matrix and graph libraries, global load balancing frameworks - Performance analysis, comparison between performance of X10 application in managed environment vs native environment - Foundations: weak-memory models, models of imperative concurrency, reasoning techniques for dynamic concurrency - Extensions: fault-tolerance, dynamic places, hierarchical places - Type systems for concurrency and alias management - Deterministic computation, phased computations -- clock-based concurrency, stream-based computation - Static analyses for atomicity violations, race conditions, deadlock-freedom. - Compilation techniques: code generation, compilation for work-stealing, concurrency and communication optimizations, compilation for scale - Runtime systems, interoperability with Java, MPI - Design and evaluation of JVM extensions for X10 - Distributed GC - Design and experience with development tools (IDEs) for X10 - Performance analysis and monitoring tools - Testing, bug detection and program understanding tools - Debugging frameworks, including large-scale debugging, differential debugging Organizing Committee General Chair: Olivier Tardieu, IBM T.J. Watson Research Center, USA Program Chair: Mikio Takeuchi, IBM Research - Tokyo, Japan |
|
From: David P G. <gr...@us...> - 2013-12-19 22:55:50
|
We're pleased to announce that X10 and X10DT 2.4.1 are now available for download at http://x10-lang.org/releases/x10-release-241. This release of X10 contains two major new features: (1) a technology preview of Resilient X10. Resilient X10 enhances the X10 language and implementation with the ability to detect and tolerate Place failures. (2) Support for multi-place execution of Managed X10 on Windows using a new pure-Java implementation of the x10rt network layer. Below is a summary of JIRA issues addressed for the X10 2.4.1 ** New Features and Improvements * [XTENLANG-2151] - Multi-place Managed X10 on windows * [XTENLANG-2356] - x10doc: alphabetize names wherever possible * [XTENLANG-2884] - Support compilation with Intel C++ compiler * [XTENLANG-3281] - java interop: Should have a way to convert x10.lang.String to java.lang.CharSequence * [XTENLANG-3290] - Add output buffering to JavaSockets * [XTENLANG-3291] - Specialize String.operator+ for primitives to avoid boxing * [XTENLANG-3292] - Reader.readLine is 5~10x slower than in Java * [XTENLANG-3297] - Unnecessary casts for receiver of interface method * [XTENLANG-3230] - Use OSGi bundle for loading classes in Managed X10 * [XTENLANG-3314] - X10 needs a standard mechanism to globalize messages * [XTENLANG-3319] - GML: add DGEMV with offsets * [XTENLANG-3322] - Native X10: avoid upcast to Any for argument to equals ** Tasks and Bugs * [XTENLANG-1454] - Confusing error messages for switch * [XTENLANG-1994] - Qualify top-level x10 namespaces (x10 and x10aux) in C++ code to make sure they start in the global namespace (-> ::x10, ::x10aux) * [XTENLANG-2215] - FinallyEliminator pass creates (unreachable) control flow path in which statically there is no return statement * [XTENLANG-2724] - Failure to carry type information for an array of arrays of some parameterized class into method call * [XTENLANG-2849] - Add text to language specification to emphasize that clocks are inter-place * [XTENLANG-3088] - Generic exceptions should not be allowed. * [XTENLANG-3094] - hasZero and isRef do not work with constrained types * [XTENLANG-3234] - Compiler crash on inequality constraint involving type-generic args * [XTENLANG-3235] - ResilientKMeans sometimes hang up when a place is killed * [XTENLANG-3236] - DeadPlaceException is not thrown for "at" * [XTENLANG-3238] - Asynchronous place death during repeating "at" hangs up the app without DeadPlaceException * [XTENLANG-3265] - Classic F-Bounded Polymorphism example fails to typecheck * [XTENLANG-3267] - TestTypeParamShaddowing from FrontEndTests failing with "inconsistient type" * [XTENLANG-3268] - TestTypeParamShadowing from FrontEndTests failing to typecheck * [XTENLANG-3277] - Unable to cast a Rail[T] to Rail[T]{self!=null} in argument to constructor * [XTENLANG-3280] - Double.parse doesn't allow leading and trailing whitespaces in Native X10 * [XTENLANG-3284] - release packging scripts should also create RPMs from linux tarballs * [XTENLANG-3287] - Post-compilation error with a class that implements Comparable interface and inherits compareTo method from super class * [XTENLANG-3288] - Offset and length were ignored with OutputStreamWriter.OutputStream.write(Rail[Byte],Long,Long) for Native X10 * [XTENLANG-3289] - Printer.printf to file is absolutely slow compared to fprintf in C * [XTENLANG-3296] - Team.alltoall doesn't work with Managed X10 * [XTENLANG-3299] - Compiler NPE when optimizing because of incorrect implementation of alpha renaming * [XTENLANG-3308] - Native codegen error for interface with a method that takes a non-primitive struct as argument * [XTENLANG-3309] - isref should be closed under sub typing * [XTENLANG-3320] - Multi-place Native X10 crash on cygwin (during x10rt startup?) * [XTENLANG-3326] - Converge the environment variable to select Resilient Finish For the details of JIRA issues fixed in this release, see http: http://jira.codehaus.org/secure/ReleaseNote.jspa?version=19142&projectId=11812 |
|
From: David P G. <gr...@us...> - 2013-09-28 21:43:05
|
We're very excited to announce that X10 and X10DT 2.4.0 are now available for download at http://x10-lang.org/releases/x10-release-240 ! X10 v2.4 is a major revision of the X10 programming language that significantly improves the ability of the X10 programmer to exploit the expanded memory capabilities of modern computer systems. In particular, X10 v2.4 includes an extensive redesign of arrays and a change of the default type of unqualified integral literals (e.g. 2) from Int to Long. Taken together these two changes enable natural exploitation of large memories via 64-bit addressing and Long-based indexing of arrays and similar data structures. Hints on upgrading programs written in previous versions of X10 to X10 2.4 are available at http://x10-lang.org/porting-to-24 The release notes for X10 2.4.0 are appended below: MAJOR CHANGES X10 v2.4 is not backwards compatible with X10 v2.3. The motivation for making backwards incompatible language changes with this release of X10 is to significantly improve the ability of the X10 programmer to exploit the expanded memory capabilities of modern computer systems. In particular, X10 v2.4 includes an extensive redesign of arrays and a change of the default type of unqualified integral literals (e.g. 2) from Int to Long. Taken together these two changes enable natural exploitation of large memories via 64-bit addressing and Long-based indexing of arrays and similar data structures. Please refer to the ChangeLog section of the X10 2.4.0 language specification for a more detailed list of language and class library changes. X10 2.4.0 is also the first official release of X10 with support for IBM's BlueGene/Q system. LIMITATIONS OF THIS RELEASE The following features described in the 2.4 language manual do not currently work and may be fixed in the subsequent releases: - Non-static type definitions as class or interface members (static type defs do work) The constraint solver used by X10 typechecker is known to be incomplete for situations in which a constraint implies an infinite number of distinct constraints. Additionally, the following features described in the language manual do not currently work with Native X10. - Garbage collection on AIX - Non-final generic instance methods - Exception stack traces on Cygwin Although greatly improved from previous releases, the X10 runtime still uses a modified "busy wait" loop in the worker threads that execute asyncs. A consequence of this is that even if a Place has no asyncs to execute, one worker thread in the Place will still consume CPU cycles busy waiting for messages to arrive from the network. The generated C++ code requires g++ 4.2 or better to be compiled; we do almost all of our testing against g++ 4.4. + On AIX and Power/Linux, you may either use g++ 4.2 or better or xlC 11.1 or better. + On MacOS compiling the generated C++ code may result in either a large number of spurious warnings (Xcode 5) or may not compile at all (some prior versions of Xcode). On versions of Xcode before Xcode 5, you can use the g++ front end by setting CC=gcc and CXX=g++ in your environment. We will reduce the number of warnings generated with Xcode 5 in the next release of X10. SUMMARY OF ISSUES RESOLVED IN THIS RELEASE Below is a summary of JIRA issues addressed for the X10 2.4.0 ** Improvements and New Features * [XTENLANG-1082] - "New X10 Project" wizard should offer other source code samples than "Hello, World" * [XTENLANG-1199] - X10DT needs a "Show the type of the selection" operation. * [XTENLANG-1244] - Compiler support for regression testing * [XTENLANG-1304] - Should provide a place in project properties to specify the number of places * [XTENLANG-1306] - Content assist should add import when inserting new reference to an existing type * [XTENLANG-1548] - Browsable online tutorial for X10 2.3 (or programmer's guide? * [XTENLANG-2833] - upgrade bdwgc tarball to version 7.2d * [XTENLANG-2843] - Add @Unserializable interface * [XTENLANG-2933] - Support cross-project dependencies for remote builds in X10DT * [XTENLANG-3022] - Implement object identity map using true hashmap * [XTENLANG-3025] - Please add a way to make a Team with a PlaceGroup * [XTENLANG-3126] - Allow specification of matrix dimensions to DenseMatrixBLAS.comp[*] * [XTENLANG-3155] - Native X10 should not require Java JRE in sourcepath * [XTENLANG-3158] - Emulated bcast, reduce, allReduce using tree communications * [XTENLANG-3161] - WorkerLocalHandle: support lazy initialization and application of closure or reduction to all instances * [XTENLANG-3176] - Support MPI based X10RT with Managed X10 * [XTENLANG-3182] - Incorrect specification of ateach * [XTENLANG-3194] - Build configuration for FX10 * [XTENLANG-3198] - Implemented methods of File in c++ backend * [XTENLANG-3199] - Support generating property file in a different directory from jar file * [XTENLANG-3210] - Support idiom where abstract property methods are used to specify constraints in abstract methods * [XTENLANG-3224] - PlaceLocalHandle and DistArray collection for Managed X10 * [XTENLANG-3227] - Miscellaneous fixes to GML * [XTENLANG-3228] - Implement SUMMA matrix multiplication for transposed A matrix * [XTENLANG-3243] - GML: support copySubset from Rail to DenseMatrix * [XTENLANG-3249] - Remove variance from runtime type of managed X10 * [XTENLANG-3252] - Update language spec for change in int/long literals * [XTENLANG-3254] - Update printing of int/uint literals by X10 compiler in diagnostic methods to have n/un suffix * [XTENLANG-3263] - Redesign CustomSerialzation protocol to use X10Serializer/X10Deserializer classes * [XTENLANG-3270] - GML: add symmetric rank-K update (DSYRK) and DGEMM with offsets ** Tasks * [XTENLANG-3196] - Update to commons-math3-3.1.1.jar * [XTENLANG-3203] - Rewrite Array chapter of language spec for Rail/SimpleArray/RegionArray changes * [XTENLANG-3213] - Update Introduction to X10 to X10 2.4 ** Bugs * [XTENLANG-1011] - x10c and x10 scripts should check whether JAVA_HOME points to a Java 6 JRE/JDK * [XTENLANG-2055] - x10/io/Marshal.x10 cannot be flattened * [XTENLANG-2126] - Compiling with -nooutput leaves some temporary files * [XTENLANG-2250] - Add RunTestSuite to precommit * [XTENLANG-2887] - X10 For loop performance on c++ backend * [XTENLANG-2928] - Remote Array copy runtime exception running on MPI transport * [XTENLANG-2949] - Guard on superclass isn't entailed when it ought to be * [XTENLANG-2956] - X10DT doesn't seem to find existing classes very well * [XTENLANG-2960] - T isref, T haszero should imply that null can be assigned to a variable of type T * [XTENLANG-2991] - ArrayList[Rail[Int]] doesn't contain Rail[Int]s * [XTENLANG-2995] - Compiler NPE upon assignment of null to a non-null type * [XTENLANG-3038] - DistArray#reduce segfaults after many calls when compiled with MPI backend * [XTENLANG-3053] - Need to clean generated artifacts when packages are moved/deleted * [XTENLANG-3117] - No splash screen on X10DT startup on Windows 7 * [XTENLANG-3164] - Language spec missing explanation of at (x) where x is of type GlobalRef * [XTENLANG-3167] - GML: incorrect row/column used for init function in DistDenseMatrix * [XTENLANG-3168] - GlobalRef creates debug strings even when they are not being used * [XTENLANG-3170] - Incorrect "No output code is generated" errors from X10DT * [XTENLANG-3178] - Linking with gperftools requires --no-as-needed * [XTENLANG-3179] - String.format() doesn't allow percent signs for Native X10 * [XTENLANG-3180] - Not all @NativeRepped java classes are hidden from X10 * [XTENLANG-3183] - Language spec: comments and questions on coercions/casts of numeric values * [XTENLANG-3184] - No propagation of checked exceptions through at expressions * [XTENLANG-3185] - javac error when using checked exceptions in at expressions * [XTENLANG-3188] - Assertion fails with TestBlockColl.x10 * [XTENLANG-3191] - Incorrect codegen for struct with ULong property * [XTENLANG-3192] - Typedef only .x10 files are not included in x10.jar etc. making the resolution of the types not possible * [XTENLANG-3195] - x10c doesn't accept -x10lib option with relative path * [XTENLANG-3206] - Post-compilation error for field assign to an instance of type T * [XTENLANG-3207] - Bad interaction between FieldInitializerMover and Inliner generates missing __fieldInitializers calls * [XTENLANG-3208] - Post-compilation error for field assign to an instance of type C[T] * [XTENLANG-3209] - Managed X10: bridge method is not generated for methods with parameters with constraint based on abstract property * [XTENLANG-3211] - X10DT does not "Clean" up after deleted source files * [XTENLANG-3212] - Constructor splitter should not split constructor of Java types * [XTENLANG-3216] - The combination of T isref and T iszero should indicate that null is a valid value of type T * [XTENLANG-3217] - NPE caused by "-o file.jar" with x10c * [XTENLANG-3218] - NPE in compiler triggering an NPE in error queueing/reporting mechanism * [XTENLANG-3220] - using -o with an absolute filename generates an incorrect .properties file * [XTENLANG-3222] - Java codegen bug related to generic interface method * [XTENLANG-3226] - Use of Clock.advanceAll() throws a concurrent modification exception, under some conditions * [XTENLANG-3229] - Post-compilation error due to name clash between user-defined SumReducer that implements Reducible with Reducible.SumReducer * [XTENLANG-3232] - X10 APIs for (de-)serializing data * [XTENLANG-3233] - X10 optimizer eliminates method invocation with side effect * [XTENLANG-3237] - Missing explicit boxing when T is replaced with primitive type by optimizer * [XTENLANG-3239] - Inlining a method that returns a value from a finally block generates invalid c++ code * [XTENLANG-3240] - Compile error in C++ backend * [XTENLANG-3245] - Incomplete type for PlaceLocalHandle when compiling with debug symbols * [XTENLANG-3248] - NullPointerException in RuntimeType.getVariance * [XTENLANG-3250] - @Profile(prof) causes compile error in Java backend * [XTENLANG-3256] - init is called for unnecessary points in x10.array.DistArray_BlockBlock_2 * [XTENLANG-3258] - Missing stack trace from exception at remote place * [XTENLANG-3259] - java interop: post compilation error for calling java constructor with Comparable parameter * [XTENLANG-3261] - Post compilation error with Boolean.TRUE For the details of JIRA issues fixed in this release, see http: http://jira.codehaus.org/secure/ReleaseNote.jspa?version=17010&projectId=11812 Please use the X10 JIRA to report bugs, after ensuring the problem is not already reported: http://jira.codehaus.org/secure/IssueNavigator.jspa?reset=true&&pid=11812&resolution=-1 |