The AMD Stream SDK 1.4 offers improvements in data handling, data sharing and DirectX interoperability. DirectX is the set of application programming interfaces in Windows for handling multimedia tasks.
AMD and graphics chip rival Nvidia offer tools for developers to tap the graphics processing unit not just for handling graphics, but for computational computing typically handled by the general processor, or CPU. Developers of applications related to scientific research, oil and gas exploration and financial modeling and risk assessment like to take advantage of the GPU's more advanced parallel processing capabilities.
The enhancements in the latest AMD SDK include support for multiple GPUs in a single program, 8-bit and 16-bit integer types, memory pinning to optimize data transfers and asynchronous stream write calls. In addition, AMD has provided access to thread-level data sharing and additional hardware features exposed in the Compute Abstraction Layer.
AMD has also incorporated in the SDK technology to take advantage of enhancements in Brook+, the programming language for using graphics hardware for general-purpose computations. AMD offers its FireStream processor line for computational computing.
Nvidia competes with AMD with its Tesla line of general purpose GPUs and its CUDA development tools that enable programmers to use the C programming language to code algorithms for execution on the GPU. Nvidia claims CUDA isn't difficult for developers because it uses C, a language they're already familiar with.
The ATI Stream SDK v1.4 can be downloaded at no charge from AMD's Web site.