 | Level: Introductory John Easton (JKJ@uk.ibm.com), Infrastructure Architect, Emerging Technologies, IBM
Ingo Meents (MEENTS@de.ibm.com), Architect for Cell Solutions, Advanced Planning, Simulation, and Optimization, IBM
Olaf Stephan (STEPHANO@de.ibm.com), Server Specialist, DB2, Warehousing BI Solutions, IBM
Horst Zisgen (horst_zisgen@de.ibm.com), Program Manager Simulation/Operations Research, IBM
Sei Kato (SEIKATO@jp.ibm.com), Research Staff Member, IBM
06 Nov 2007 The seven quick-read parts of this "Porting workshop" series take
you on a real-world trip from strategy and planning through workload execution,
performance tweaking, optimization, and a solid conclusion. The series describes how to
most effectively port compute-intensive applications to the Cell Broadband Engine
platform. In part seven, the authors evaluate the performance data to date.
This seven-part, quick-read workshop series is taken from the real-world case
study whitepaper,
"Porting Financial Markets Applications to the Cell Broadband Engine Architecture"
(written by John Easton, Ingo Meents, Olaf Stephan, Horst Zisgen, and Sei Kato,
IBM Systems and Technology Group, June 2007; see Resources). You can probably spend less than 10
minutes reading each installment and come out at the end with a strong basic
knowledge of the requirements for effectively porting a compute-intensive
application (in this case, a financial market application) to the Cell/B.E. processor.
Editor's note: The performance results in this series were obtained using Versions 1 and 2.1 of the Cell Broadband Engine Software Developer Kit (SDK). The current version of the SDK, the IBM Software Development Kit for Multicore Acceleration, Version 3.0, has recently become available and offers many enhancements in functionality, ease of use, and performance over the earlier versions. While the results documented in this article are correct for the earlier versions of the SDK, different results will be obtained with SDK 3.0. Watch for updates to the articles in this series that will describe the latest performance improvements obtained using SDK 3.0.
Introducing the
application
The example application modified in this article is a piece of
code used to price a European Option to highlight the benefits of the
Cell/B.E. blade. A European Option is a simple financial contract with strict
terms and properties that gives the buyer the right to trade a given asset at a
specific price on a specific date. It is generally an option that can be
exercised only at the end of its life. By contrast, an American Option can be traded
at any time between its purchase date and the date at which the contract expires.
Because a European Option is traded on a fixed
date, it is a simpler calculation to perform because the time variability of the
American Option is removed.
You can use several different models price a European Option, depending on
the type of asset that underlies it. For example, an option based on currency is
calculated using a slightly different model than an option based on
futures. In the example described in this series, the calculation is based on a
simple Monte Carlo simulation technique. You will generate 200,000,000 uniform, pseudo-random numbers. These numbers are
transformed to a log-normal distribution using a Box-Müller transform.
Using the random numbers generated, you will execute the financial model repeatedly to
simulate a random walk. The final stage of the analysis will be the calculation of
the relevant statistics, such as the minimum, maximum, and average and the 95
percent quartile for losses.
Getting the most
performance out of Cell/B.E. technology
This work has led to a few observations to share with those
readers who would like to perform a similar porting exercise.
Offloading for performance enhancement
The performance data shows that a single SPU is capable of delivering almost
half the performance of an Intel® general purpose processing core. As such, the
first recommendation has to be to use the performance of the SPUs by
offloading as much of the computation onto the SPUs as possible.
Exploiting the SIMD
To really make the SPU perform well, exploiting the SIMD nature of the
SPU is vital. Although the compiler technology, especially that provided by XLC,
provides some capability to auto-SIMDize
the input program code, the very
flexibility of a language such as C means that this technology does not always
make a particularly good approximation of the best SIMD code.
As such, to take advantage of the SMID nature fully, it might be more
appropriate to actually write the SIMD code yourself rather than relying on the
compiler to do it. The key to doing this is to make sure that the way you have
chosen to rewrite the algorithm to exploit SIMD is optimal for the given
application and the input data. In certain situations, such as the case where the
source code is heavily templated C++, or where the code makes significant use of
object-oriented constructs, you might find that starting from scratch is a much
quicker way to implement application code.
Drawing some conclusions
Many organizations are investigating the use of acceleration technologies to
improve the performance of their algorithms, because their abilities to exploit
ever-increasing numbers of general-purpose processors must be reconciled against
the challenges that they face with space, power, and cooling issues. So, how is
the Cell/B.E. system positioned against these other accelerator technologies?
At one end of the spectrum are the general-purpose processors from Intel and AMD
that make up the majority of the computational infrastructures used by these
organizations. The huge numbers of systems based on these processors give them
several advantages over other technologies. There is a large supply of
professionals skilled in working with them, which leads to lower skills costs, a
lot of application development tooling, and therefore a large base of ISVs
supporting these platforms. Couple these facts with the relatively easy code
porting to these platforms, and it has meant that the majority of application codes
run by the business run on systems based on these processors.
These general-purpose processors are very much regarded as jack-of-all-trades
devices. They do suffer, though, in that they deliver relatively low performance
for the area of the chip, and they do this with relatively high power consumption per
computation.
At the other end of the spectrum is a number of what can only be described as
esoteric technologies. These include Field Programmable Gate Arrays (FPGAs),
Complex Programmable Logic Devices (CPLDs), and Graphics Processing Units (GPUs).
Though there is a wide range of different devices here, these all typically offer
high performance for their chip area and consume much less power per computation
than the general purpose processors. Given this, these devices are highly
efficient at addressing a very narrow task scope.
You might ask why their adoption has not been wider. This is basically because
they are very much the opposite of the general-purpose devices: the skills to
program them are rare and relatively expensive. If you join this with the lack of
application development tooling, it leads to quite hard code creation and porting
for these platforms.
The performance capabilities of these platforms are well understood, but
realizing this performance is proving to be a challenge for many organizations.
Assuming that you can find the skills to do this, the porting process is generally
both slow and costly. This in turn leads to a miniscule ISV base for these
platforms today.
So how does the Cell Broadband Engine technology compare? You might expect that
the nature of the Cell Broadband Engine environment means that it would be closer
to the FPGAs and GPUs than the general-purpose processors. This is not the case.
The facts that the Cell/B.E. environment runs a standard Linux® operating system
and has a growing range of development tools, ISV applications, and standard libraries
means that it is actually much closer to the general-purpose processors than might
be expected. Working with other customers in this space has simply confirmed this.
They all regard the Cell/B.E. platform as significantly easier to port code to
than FPGAs and GPUs.
In an industry where time-to-market is a key concern, this places the Cell/B.E.
platform in a unique position: for the workloads that can exploit its
capabilities, the Cell/B.E. environment offers excellent performance advantages at
only a slightly longer time-to-market than the general purpose devices.
Regardless of the technology used, this series introduced that code
optimization is critical to get the best out of any chosen platform. The days of
simply recompiling application code to fully exploit a new processor are over.
Simple recompile will forever be relegated to a second-class effort position when
it comes to effective porting. In the future, programmers will have to use
increasingly advanced programming techniques that take into account the management
of both code and data parallelism to get the best out of any architecture. The
abilities of programmers to adapt to these new technologies will be critical to
the success of organizations that are finding that they can no longer achieve
commercial advantage simply by throwing more power at the problem.
Constraints over power, space, and cooling issues and an increasing pressure to
innovate are causing organizations to reconsider strategies when it comes to
designing the compute infrastructures of the future. A system based on a processor
as capable as the Cell Broadband Engine processor has to be a strong contender, not only in terms of its
computational power, but also its data movement and manipulation abilities.
Couple this with a number of strong customer proof
points, a growing ecosystem of tools, support from key Independent Software
Vendors, and the results of experiments such as this one to see that the Cell/B.E.
platform shows that it is extremely credible for taking on the technology
challenges of these organizations.
Acknowledgments
Many other individuals contributed (both knowingly and unknowingly) to this
piece of work. The authors wish to acknowledge their kind contributions. Without
this assistance, this paper would never have been written.
Resources Learn
- Use an RSS
feed to request notification for the upcoming articles in this series. (Find out more about RSS feeds of developerWorks content.)
- Check out the original whitepaper,
"Porting Financial Markets Applications to the Cell Broadband Engine Architecture"
(alphaWorks, June 2007). The original whitepaper combines
the contents of this entire series. It also
provides a tidy introduction to the Cell/B.E. architecture, and it explains why the
processor is important, especially for compute-intensive financial market
applications.
- See
"Introduction to the Cell Multiprocessor"
(IBM Journal of Research and Development, 2005) for an introductory
overview of the Cell/B.E. multiprocessor's history, the program objectives and
challenges, the design concept, the architecture and programming models, and the
implementation.
- Go to "Porting practices: Compute-intensive applications"
(developerWorks, June 2007) for help bringing a compute-intensive
application to the Cell/B.E. architecture.
- Read "Tech tips: SPU vector intrinsics at your fingertips"
(developerWorks, May 2007) for a handy list to keep you on the right side of common
Cell/B.E. SPU vector intrinsics. This article was extracted from a longer article,
"Programming high-performance applications on the Cell BE processor, Part 5").
- Review "Cell Broadband Engine Architecture and its first implementation"
(developerWorks, November 2005) for an up-close look at the performance
figures and characteristics of the first implementation.
- Explore the
QuantLib project, which is a free/open-source library
written in C++ with a clean object model for modeling, trading, and risk
management in real-life. It is exported to different languages such as C#,
Objective Caml, Java™, Perl, Python, GNU R, Ruby, and Scheme.
- Learn more about the
Mersenne-Twister.
It is a very fast, pseudo-random number-generating algorithm that uses memory very
efficiently. The algorithm has a far longer period and far higher order of
equidistribution than any other implemented generator.
- Refer to
"Implementation of a Mixed-Precision in Solving Systems of Linear Equations on the CELL Processor"
for details on the implementation of code to solve a linear system of equations
using Gaussian elimination in single precision with iterative refinement of the
solution to the full double-precision accuracy.
- Bring up the
Software Development Kit 2.1 Installation Guide Version 2.1
(PDF) to walk through installation, configuration, and many of the basics that
you need to know to get started with development. Two companion pieces,
"Cell/B.E. SDK 2.1: Setting up Fedora Core 6"
and
"Cell/B.E. SDK 2.1: Understanding the terminology"
(developerWorks, April 2007), can help you get the requisite FC6 up and running,
and they provide a quick reference to Cell/B.E. terminology.
- To learn more on Cell/B.E. programming, try the
developerWorks series:
- Refer to the Cell
Broadband Engine documentation section of the IBM Semiconductor Solutions Technical Library for a wealth of downloadable manuals,
specifications, and more.
- Sign up for the developerWorks newsletter
and get the latest developer news and Cell/B.E. happenings delivered to your inbox each week.
Check Power Architecture when you sign up to receive Cell/B.E. news in your newsletter.
Get products and technologies
Discuss
About the authors | | | John is currently leading a worldwide emerging technologies team within IBM Systems and Technology Group. He has several roles competing for his time, all of which revolve around advising organizations on how best to exploit new technologies. John has been working for IBM for over 20 years in a variety of technical roles. He worked in Distributed Systems Development in Austin before the launch of the RS/6000, and he holds several patents in the areas of security and systems software. Before taking his current role, he was the European technical leader for grid computing. |
| | | Ingo Meents joined IBM nine years ago and works currently as an IT Architect in IBM Global Engineering Solutions (GES). His current focus is to provide IBM customers with knowledge of the latest Cell/B.E. software technology by consulting, educating, briefing, and creating solutions for this platform. Before his work on the Cell/B.E. platform, he was lead architect for a modeling, simulation, and production planning solution used by the IBM 300mm semiconductor line in Fishkill. Starting as a research student at IBM, Ingo Meents received his doctor's degree from the University of Clausthal in 2001. |
| | | Olaf Stephan joined IBM in 1998 and currently works as an IT Specialist in IBM Global Engineering Solutions (GES). His focus is to provide IBM customers with knowledge of the latest Cell/B.E. software technology by consulting, educating, briefing, and development for this platform. Before his work on the Cell/B.E. platform, he worked in the areas of data management, data warehousing, business intelligence, and data integration. Olaf holds a Masters degree in Electrical Engineering, specializing in Communications Technology, from the University of Applied Sciences, Koblenz, Germany. |
| | | Horst has over 10 years of experience in the application of simulation methods and the development of mathematical models in different areas. He is currently leading a development team in IBM Global Engineering Solutions (GES) that is working on a simulation and planning solution used by IBM 300mm manufacturing in Fishkill and by external customers as well. Horst is also the European subject matter expert for the GES supply chain offerings. In addition, Horst regularly gives lectures at universities about simulation and mathematical modeling. Horst is a member of a standardization group for simulation and optimization. |
| | | Sei Kato is a staff member in IBM Research, Tokyo Research Laboratory. He joined IBM in 2002 after receiving his PhD in Mathematical Science from the University of Tokyo. After joining IBM, Sei has worked on modeling and simulating the performance of Web systems. His is currently working on the acceleration of financial calculations and on large-scale traffic simulations. |
Rate this page
| |
