Workshop on Hybrid Multi-core Computing

Held in conjunction with HiPC 2011

Bangalore, December 18, 2011

(Click here to for a report on WHMC-2010 and program details)

Program

  • 0830 - Welcome Address

  • 0850 - Programmable GPU Pipelines for Graphics and Beyond (slides) by John Owens, UC Davis

    Abstract: The computational power of GPUs, coupled with increasing programmability, is making the GPU a compelling platform for high-performance computing. GPUs excel at regular, structured computation, but irregular computation -- where processors consume an irregular, runtime-dependent amount of input or produce an irregular, runtime-dependent amount of output -- is a challenging problem in a parallel computing environment. These characteristics are common in today's real-time graphics pipelines but are typically handled in hardware. We anticipate these problems will become more relevant as we move toward the next generation of graphics systems that have at their core *programmable real-time graphics pipelines* and must instead support these workloads in more general-purpose ways. Fortunately, the research projects that are beginning to address these challenges can also address other complex, demanding application domains.

  • 0940 - Invited Talk 2 (Jatin Chhugani, Intel)

  • 1030 - Break

  • 1100 - AMD Fusion and the Future of Heterogeneous Computing by Javed Absar, AMD

    Abstract: This talk will present the AMD vision of heterogeneous computing through the Fusion System Architecture. Fusion makes the unprecedented processing capability of the APU (accelerated processing unit) accessible to the programmers with the same ease as CPU does today. The talk will discuss the APU today and its programming environment and present the future through the Fusion System Architecture. It will take the audience through the roadmap and evolution of heterogeneous computing – from proprietary drivers era to the AMD vision of architected era. It will present and get into the important differentiators of FSAIL (FSA Intermediate Layer). It will also present a comparison of FSA with OpenCL and get into the interesting details of task queuing runtime of FSA platform. The audience will not only appreciate what AMD is doing in this domain, but also at the end get a perspective of where the industry is headed.

  • 1130 - Paper Session 1

    1. Erich Marth and Siegfried Benkner. Language Support for Pipelined Applications on Heterogeneous Many-Core Architectures (slides)

    2. Jens Breitbart. Programming hybrid systems with implicit memory based synchronization (slides)

  • 1210 - GPU Computing Challenges: Past, Present, Future (slides) by Mark Harris, Nvidia

    Abstract: Ten years ago, general-purpose computation on GPUs was largely the domain of computer graphics Ph.D. students.  Programming GPUs for applications other than graphics was too difficult for the mainstream, and the few applications produced were generally not suitable for more than research prototypes.  But the potential of GPU computing was clear.  Designing GPUs and GPU software to directly address those early challenges has enabled rapid adoption and success of heterogeneous GPU computing in real science and engineering applications, from desktop to supercomputing.  Heterogeneous computing is becoming ubiquitous because it makes writing efficient high-performance software feasible, but it brings its own challenges for software developers. Addressing these will be essential on the path to exascale computing and beyond.

  • 1300 - Lunch

  • 1400 - Role of Adaptive Runtime Systems in flattening and simplifying manycore and multicore programming (slides1, slides2) by Laxmikant Kale, UIUC

    Abstract: It is clear that near future will see an exponential increase in the number of cores on a chip, as well as in heterogeneity of cores available for a computation. Memory and processing capabilities are organized in complex hierarchies in such systems. Strong NUMA factors further complicate programming. It has been argued that we need a multi-level programming methodology to match this. I argue that addition of a new degree of freedom, an adaptive runtime system, significantly simplifies the parallel programming problem. I wil draw upon our research to illustrate how one can free the programmer from much of the the burden of dealing with the heterogeneity and hierarchy by employing a relatively flat model consisting of interacting objects, aided by a runtime system that can move objects and computations around, among nodes, and among cores and accelerators within a node. Examples  will include Computational astronomy, biomolecular simulations, and KD-trees.

  • 1440 - Paper Session - 2

    1. Matthieu Ospici, Dimitri Komatitsch, Jean-François Méhaut and Thierry Deutsch. SGPU 2: a runtime system for using large applications on clusters of hybrid nodes

    2. Yuhong Zhang, Md. Mostofa Ali Patwary, Sanchit Misra, Ankit Agrawal, Wei-Keng Liao and Alok Choudhary. Enhancing Parallelism of Pairwise Statistical Significance Estimation for Local Sequence Alignment (slides)

  • 1530 - HPC in India: Academia

    • 4 @ 5 minutes each: R Govindarajan (IISc), Srinivas Aluru (IITB),

    Subodh Kumar (IITD), Kishore Kothapalli (IIIT-H).

  • 1600 - Break

  • 1630 - HPC in India: Industry

    6 @ 5 minutes each: Infosys, CRL, IBM IRL, GE, MSR, CDAC

    Santonu Sarkar (Infosys), Vipin Chaudhary (CRL), Jyothish Soman (IRL),

    Srihari Narasimhan (GE), Sarat Chandra Babu (CDAC), Kaushik Rajan (MSR).

  • 1700 - Panel Discussion: "Multi and Manycore computing in the Future"

    Vipin Chaudhary, R Govindarajan, John Owens, Ming Lin (UNC)

    Moderator: P J Narayanan

  • 1800 - Wrapup and Thanks followed by posters in parallel with a social event with food hosted by Nvidia India

    • List of Posters

    1. Tavish Vaidya and Prof. Sharat Chandran (IIT-B)
    2. IIIT-H
    3. ........
  • 1830 - End

Accepted Papers

  • Matthieu Ospici, Dimitri Komatitsch, Jean-François Méhaut and Thierry Deutsch. SGPU 2: a runtime system for using large applications on clusters of hybrid nodes
  • Yuhong Zhang, Md. Mostofa Ali Patwary, Sanchit Misra, Ankit Agrawal, Wei-Keng Liao and Alok Choudhary. Enhancing Parallelism of Pairwise Statistical Significance Estimation for Local Sequence Alignment
  • Erich Marth and Siegfried Benkner. Language Support for Pipelined Applications on Heterogeneous Many-Core Architectures
  • Jens Breitbart. Programming hybrid systems with implicit memory based synchronization

CALL FOR PAPERS 

The second Workshop on Hybrid Multi-core Computing will be held in conjunction with the 18th International Conference on High Performance Computing (HiPC) being held at Bangalore, India. The workshop will focus on algorithmic, applied, and analytical issues in hybrid multi-core computing. The workshop aims at bringing together researchers and practitioners in the rapidly evolving area of hybrid multi-core computing. Hybrid multi-core computing uses a mix of CPUs and special purpose accelerators to arrive at highly efficient, practical, and scalable solutions. This raises several important issues including the mechanism(s) to execute concurrently on a hybrid computing platform.

The workshop will focus on issues emerging out of hybrid multi-core computing. The workshop shall feature invited talk(s) and also contributed research papers. Research papers addressing any of the topics listed below, and possibly beyond, in connection to hybrid multi-core computing are solicited. It is expected that papers are original, correct, and are of high technical quality. Papers should not be currently in submission at any workshop/conference/journal with a published proceedings. The topics of interest include, but are not limited to, the following.

  • Hybrid computing algorithms and applications
  • Case studies and benchmark suites for hybrid programs
  • Programming models for hybrid multi-core computing
  • Issues in hybrid multi-core computing such as load balancing, synchronization
  • Scalability studies, auto-tuning, or frameworks for hybrid programs
  • Paradigms for hybrid multi-core computing
  • Performance issues and models for hybrid multi-core computing
  • Programming languages and tools for hybrid multi-core computing

 

SUBMISSION GUIDELINES

 

The workshop seeks original, correct, and novel submissions of high quality that are not currently under review at any other venue. Submissions should not exceed eight (8) pages in the two column IEEE conference format including references, tables, and all other material. An appendix can be included but there is no guarantee that the appendix shall also be read by the reviewers. Submissions that deviate significantly from the above guidelines may not be considered. All submitted papers shall be reviewed by at least three independent reviewers before arriving at a decision.

Click here to submit your paper using EasyChair.

 

IMPORTANT DATES

 

  • Submission deadline (Extended): October 9th, 2011
  • Acceptance notification: November 5th, 2011
  • Camera Ready deadline: November 12th, 2011
  • Workshop Date: Dec 18th, 2011

ORGANIZATION

 

The workshop is jointly organized by IIIT  Hyderabad and IBM, India.  

General Chairs

  • P. J. Narayanan, IIIT Hyderabad, India..
  • Dinesh Manocha, UNC, USA

Program Chairs

  • David Kirk, Nvidia
  • Kishore Kothapalli, IIIT Hyderabad, India

Organizing Chairs

  • Suresh Purini, IIIT Hyderabad, India
  • Subodh Kumar, IIT Delhi, India

Program Committee

  • John Owens, UC Davis, USA
  • Gagan Agrawal, Ohio State, USA
  • R. Govindarajan, IISc, Bangalore, India
  • Mainak Chaudhari, IIT Kanpur, India
  • David Bader, Georgia Tech, USA
  • Bo Hong, Georgia Institue of Technology, USA
  • Weng-Fai Wong, NUS, Singapore
  • Philippas Tsigas, Chalmers Univeristy, Sweeden