SDKs optimize apps for OpenCL

Advanced Micro Devices Inc. has introduced a set of software development tools and solutions to enable developers to optimize their applications for OpenCL standards. These advanced tools create a foundation for software companies to realize the full potential of the AMD Fusion Family of Accelerated Processing Units (APUs), harnessing the combined compute power of AMD's high-performance CPUs and GPUs across a wide array of heterogeneous computing platforms. As a result, developers can create innovative experiences such as HD video, 3D gaming, video conferencing and intuitive user interfaces, to truly differentiate their applications in the market.

Among the new offerings is the gDEBugger product, which was created by experts from AMD's Israeli research center, based on AMD's acquisition of startup company Graphic Remedy in October 2010. gDEBugger is an advanced OpenCL and OpenGL debugger, profiler and memory analyzer. The AMD gDEBugger provides developers with the ability to debug OpenCL kernels running on AMD GPUs, and step through their source code while examining kernel variables and data. This product, which is a plug-in designed to work with Microsoft Visual Studio, includes all of gDEBugger's previous features and capabilities.

Additional developer solutions include a Parallel Path Analyzer (PPA), Global Memory for Accelerators (GMAC) and Task Manager tools, which are being developed by Multicoreware in collaboration with AMD. These new tools and solutions, expected to be available in Beta during Q3 of this year, are designed to make OpenCL GPU development easier and more efficient.

Parallel Path Analyzer (PPA) is a profiling tool for developing applications that optimize both GPU and CPU load. The PPA visualises data transfers and kernel execution, identifies system-wide critical paths and locates data dependencies.

The Global Memory for Accelerators (GMAC) API provides a framework in which a developer can create applications leveraging the immense compute capabilities of OpenCL, but without the overhead of having to explicitly manage multiple data buffers across the separate address spaces of GPU and CPU.

The Task Manager API provides a framework for managing compute tasks in a heterogeneous multicore environment. OpenCL kernels can be automatically scheduled to execute on an available and task-appropriate device, providing dynamic load balancing, optimizing use of available compute resources and removing the burden of explicit schedule handling.

The new tools expand AMD's line of developer solutions that are available on the AMD Developer Central website, including software development kits, libraries, compilers, webinars and educational support.