high performance computing with the cell broadband...
TRANSCRIPT
Scientific Programming 17 (2009) 1–2 1DOI 10.3233/SPR-2009-0279IOS Press
Guest Editorial
High Performance Computing with the CellBroadband Engine
The Cell Broadband Engine was conceived to enablethe design of novel and highly efficient systems forcompute-intensive applications. The Cell/B.E. departsfrom prior architectures by adopting a heterogeneouschip multiprocessor architecture with novel accelera-tor cores and an explicitly managed memory hierar-chy. The increased computing density of the design im-proves peak performance as well as power efficiency.
Beyond the Sony PlayStation 3, its first target sys-tem, the Cell Broadband Engine architecture has beenused in systems ranging from compute-intensive em-bedded systems to high-end compute servers, demon-strating the versatility and power of its new concept.In 2008 the IBM RoadRunner supercomputer, incor-porating nearly 13,000 Cell/B.E. processors, becamethe first system in the world to exceed the 1 Petaflopper second performance barrier as measured by theTOP500 list of the world’s most powerful supercom-puters. In the same year, Cell-based systems were alsothe world’s most power-efficient systems as measuredby the Green 500 list.
The Cell/B.E.’s departure from a traditional homo-geneous SMP system design with hardware-managedcaches change algorithms, programming models, com-pilers and run-time systems. The Cell/B.E. architec-ture offers new ways for software to control hardwareresources that challenge software architects and algo-rithm researchers to rethink algorithms and computingparadigms to exploit its greatly increased performancepotential.
Peak performance is obtained when vector opera-tions are fully utilized in each accelerator core (SPE),all cores are simultaneously fully engaged, and byoverlapping communication with computation usingbulk DMA transfers between main memory and theSPEs. This benefits from a contiguous data layout ofarrays as bulk DMA transfers are most efficient oncontiguous data residing in memory. Such a layout
for a 2-D array is square block format. This choiceleads to very high Cell/B.E. performance of kernel rou-tines that operate on these square block sub arrays si-multaneously in the SPEs. In particular, this issue de-scribes nearly perfectly performing matrix multiplica-tion kernel routines. By exploiting these array layoutsand overlapping Cell/B.E. bulk DMA transfers withcomputation to ensure that kernel routines will alwayshave their new sub matrix operands in their SPE localstores ready for their next invocations, the Cell/B.E.usually reaches its full potential on the most importantdense linear algebra algorithms.
The ultimate measure of the success of an architec-ture is the applications that it enables and the scientificbreakthroughs that become possible with those appli-cations. Thus, the spotlight in this issue is on innova-tive algorithms, programming models and applicationarchitectures to exploit this new hardware design.
The papers included in this special issue fall in threecategories, covering performance engineering for keyalgorithms and library functions; programming modelsand runtime systems; and, most importantly, applica-tions that build on these foundations to enable scien-tific discoveries across a wide range of disciplines.
We received 42 submissions from 16 countries in re-sponse to our call for papers. To select the best of thesepapers, we relied on an Editorial Board consisting ofexperts covering all areas of high-performance com-puting as well as on external subject matter experts.Together, they produced 138 reviews, or an average ofover 3 reviews per paper. Based on these reviews, weselected twelve papers for this special issue.
In the performance engineering and library develop-ment category:
– Implementing a Parallel Matrix Factorization Li-brary on the Cell Broadband Engine
– QR Factorization for the CELL Processor
1058-9244/09/$17.00 © 2009 – IOS Press and the authors. All rights reserved
2 Guest Editorial
– Programming the Linpack Benchmark for theIBM PowerXCell 8i Processor
In the programming models and runtime systemscategory:
– Available Task-level Parallelism on the Cell/B.E.– CellSs: scheduling techniques to better exploit
memory hierarchy
In the applications category:
– High Performance Protein Sequence DatabaseScanning on the Cell/B.E. Processor
– Building High-Resolution Sky Images using theCell/B.E.
– Implementation of scientific computing applica-tions on the Cell Broadband Engine
– Efficient SIMDization and Data Management ofthe Lattice QCD Computation on the Cell Broad-band Engine
– Streaming Model Based Volume Ray Casting Im-plementation for Cell Broadband Engine
– 3D Seismic Imaging through Reverse-TimeMigration on Homogeneous and HeterogeneousMulti-core Processors
– Implementation and Performance Modeling ofDeterministic Particle Transport (Sweep3d) onthe IBM Cell/B.E.
We want to thank all authors for their excellent con-tributions, only a few of which we were able to accom-modate in this special issue. We would like to thankthe members of the Editorial Board and external ex-perts who guided us in the selection of these papers fortheir many excellent reviews, expert suggestions andinsights.
Finally, we want to thank the Editor-in-Chief of Sci-entific Programming, Prof. Boleslaw Szymanski, whoguided it us through the entire process of preparing theissue with encouragement and expert advice.
We hope you have as much enjoyment in readingthis issue as we had in preparing it!
Michael GschwindFred Gustavson
Jan F. Prins
December 2008
Submit your manuscripts athttp://www.hindawi.com
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttp://www.hindawi.com
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation http://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Applied Computational Intelligence and Soft Computing
Advances in
Artificial Intelligence
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Electrical and Computer Engineering
Journal of
Journal of
Computer Networks and Communications
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
ArtificialNeural Systems
Advances in
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Modelling & Simulation in EngineeringHindawi Publishing Corporation http://www.hindawi.com Volume 2014
The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014