RapidMind is has announced software intended to help developers leverage the power of newfangled multi-core processors based on popular x86 chips built by Intel and AMD.
Shipping in December, RapidMind Multi-core Development Platform v3.0 allows developers to get the most out of processors like quad-core AMD Opteron and Intel Xeon processors. Previously, RapidMind has supported processors like the nVidia GPU, ATI Radeon, and IBM Cell.
While chips are getting more processor cores and more power, few applications have realized potential performance gains because of complexity in parallelizing across multiple cores, RapidMind said. Developers in most cases still limit themselves to single-core processing capabilities. RapidMind enables applications to utilize multi-core processing power without the need for complex, multi-threaded programming.
"If your application hasn't been written in a way to take advantage of multiple cores, it won't," said RapidMind President/CEO Ray DePaul.
With the use of multiple cores resulting in slower clock speeds, developers need to account for multiple cores, DePaul said. "The onus is really on the software developers to figure out how to use multiple cores or their software literally will start to slow down," he said.
Critical to RapidMind is a runtime platform with simple interfaces for developers to pass work to the RapidMind product, DePaul said.
"We have a platform, so when software is developed on top of RapidMind, we take care of distributing across multiple cores," he said.
Scaling to an unlimited number of CPU cores, the platform has been used in development of applications for database transactions, three-dimensional visualization, financial analysis, seismic analysis, and other applications.
Developers can use their regular development environment, such as Visual Studio or Eclipse. New applications can be built or existing applications can be retrofitted to take advantage of multi-core processors.
The RapidMind platform supports Windows and Linux. Future processor architectures that may be supported include systems like SPARC and IBM Power.