 |
|
|
|
|
Company Success Stories - Numerical Algorithms Group (NAG)
Numerical Algorithms Group Teams up with AMD to Provide Mathematical Software Components for AMD Opteron™ Processor
Profile
Numerical Algorithms Group (NAG)
http://www.nag.co.uk/
October 1, 2003
"With the work we’ve done with AMD, we’ve demonstrated that accuracy and high performance are not mutually exclusive. - Brian Ford, CEO of NAG"
The Numerical Algorithms Group (NAG) was founded by university researchers in the United Kingdom 30 years ago and since then has focused on building mathematical and statistical components, often called math libraries, used by software developers.
“Researchers embed math libraries in their applications to solve complex problems,” said Brian Ford, CEO of NAG. “It’s critical that a core math library be both extremely efficient and accurate. We have found that the best way to accomplish this is to work closely with microprocessor vendors.”
According to Ford, NAG was intrigued by AMD’s approach to 64-bit computing, called AMD64, and was anxious to work with AMD on building the math libraries that would support development of software applications for the first AMD64 product, the AMD Opteron™ processor.
Although 64-bit systems have been around for a number of years, Ford notes that some users, specifically in high-performance cluster computing, have been left behind due to incompatibility and poor price performance.
“It’s a laudable goal to make it easy for the user to migrate from 32-bit computing to 64-bit computing,” he said. “I think it’s an excellent approach.”
Memory Addressability Driving 64-bit Clusters
High-performance computing clusters today are largely built on 32-bit, x86 processors. These systems execute complex calculations at a cost that is significantly less than 64-bit supercomputers. They have been widely adopted in university and government research settings, but are also prevalent in commercial businesses such as the finance industry.
“In finance, for example, you want to optimize the return for clients based on their risk tolerance so you calculate what’s called the ‘efficient frontier’,” Ford said. “Ideally, within your portfolio, you’ll be right on that line as it represents the best balance between risk and return.”
In mathematical terms, this is an optimization problem, a complex calculation with many variables that change as stock prices change. According to Ford, NAG provides a number of optimization routines that can perform these calculations and counts a number of large financial institutions among users of its math libraries.
“Users of these large clusters have massive amounts of data they want to address,” Ford said. “Unfortunately, 32-bit processors and 32-bit applications are sometimes too limited in memory addressability. I’m convinced that these users are going to need to move to 64-bit computing to address their increasingly sophisticated problems.”
AMD and NAG Building 64-bit Foundation
Ford recognizes that the simultaneous 32-bit and 64-bit computing capabilities offered by the AMD Opteron™ processor will provide high performance computing users with an easier migration path.
“Because it offers the flexibility for customers to migrate to 64-bits when they’re ready, we think the AMD Opteron processor is an extremely interesting technology that is going to achieve strong acceptance in the market,” he said. “Obviously we want to be well positioned and ready for that shift.”
With that in mind, NAG has worked closely with AMD to develop the AMD Core Math Library (ACML), which is now available free on the AMD Web site at http://www.developwithamd.com/acml.
To provide software application developers with higher-level mathematical functions, NAG plans to make available later this year a set of NAG Libraries optimized for the AMD Opteron processor.
“A unique aspect of our relationship with AMD is that we worked very closely with AMD to develop ACML,” he said. “With the work we’ve done with AMD, on both ACML and libm, we’ve demonstrated that accuracy and high performance are not mutually exclusive. Our technical goal has been successfully achieved.”
About AMD
AMD is a global supplier of integrated circuits for the personal and networked computer and communications markets with manufacturing facilities in the United States, Europe, Japan, and Asia. AMD produces microprocessors, Flash memory devices, and support circuitry for communications and networking applications. The company was founded in 1969 and is based in Sunnyvale, California (NYSE: AMD).
For more information, contact Phil Hughes, AMD public relations at 512-602-4797.
© 2003 Advanced Micro Devices, Inc. All rights reserved.
AMD, the AMD Arrow logo, AMD Opteron, and combinations thereof, are trademarks of Advanced Micro Devices, Inc. Other company and product names used are for identification purposes only and may be trademarks of their respective companies. |
|
|
|
|
