CUDA is a powerful parallel processing technology. High-performance NVIDIA graphics cards contain up to 480 individual processing units that run in parallel to render a video image. If you have a computational application that can be parallelized, you can utilize a graphics card or NVIDIA Tesla card as a low-cost, multi-core parallel processor. CUDA parallelized applications can run hundreds of times faster than they would on a normal x86 processor.
CUDA software libraries allow developers writing in standard programming languages such as C++ to conveniently access the graphics processing units. In a typical CUDA application, the same instructions operate on different data in parallel in each processor core. When all application data has been processed, the results can be combined into a single solution result. Examples of applications that can take advantage of CUDA include matrix arithmetic, radar data analysis, orbital calculations, medical image processing and financial market analysis.
Concurrent’s RedHawk Linux offers complete CUDA support specially optimized for real-time performance. Maximum process dispatch latencies of typical CUDA applications are significantly reduced via RedHawk optimization. RedHawk provides a standard pre-included CUDA parallel computing SDK with examples.
Fully integrated CUDA solutions are available from Concurrent on iHawk and ImaGen platforms equipped with NVIDIA graphics or Tesla cards.