AMD Donates Software Routines For Building Multithreaded Apps
AMD made 3,000 routines for multithreaded programming available as open source, hoping to get more developers building apps that make multicore chips shine.
AMD has made 3,000 routines for multithreaded programming available as open source code on Source Forge.
Many of the routines concentrate on image handling and signal processing, tasks that can be subdivided and performed in parallel on multicore chips. Dividing a task like image management into parts that are handled on separate processors yields much higher performance, but not all business logic readily yields to parallel programming approaches.
"These routines are good examples of multithreaded processing," said Margaret Lewis, AMD director of independent software vendor strategy. "If we get open source participation in their development, we'll be able to add more routines" to the library, she said.
Applications making use of multimedia presentations can benefit immediately from the library, she added.
AMD isn't known as a frequent purveyor of open source software, having its hands full designing 64-bit, multicore chips in its ongoing battle with Intel. But the AMD Performance Library, as the chipmaker calls it, is meant to encourage application builders to use software routines that make multicore chips shine, performance-wise.
The performance library has been redubbed Framewave 1.0 and made available for free download at http://framewave.sourceforge.net/.
Writing software to take advantage of parallel processing remains a new endeavor for many developers. "It works where there are activities in the application that are not highly dependent on each other and there's no conflict between them," if processed simultaneously, Lewis said.
Animation films with their predictable, frame-based processing tasks are an ideal parallel processing challenge, while business applications have no such easily divided workload, she conceded. But many Web applications have a set of tasks that can take advantage of multithreaded processing, such as building a Web page in response to a user request that requires queries to different databases and loading a combination of text and images. With Web applications, there may be performance gains still to be achieved through rethinking the nature of the multicore processing resource, she predicted.
"We will continue to work on the issue ourselves," she added, and contribute more routines to the open source community.
Building A Mobile Business MindsetAmong 688 respondents, 46% have deployed mobile apps, with an additional 24% planning to in the next year. Soon all apps will look like mobile apps – and it's past time for those with no plans to get cracking.