Intel has released the Intel SDK for OpenCL Applications 2012 to support 3rd generation Intel Core processors. Supported seamlessly by the Intel HD Graphics Driver, OpenCL developers can now use compute resources across both the Intel CPU and Intel HD Graphics 4000. The SDK also supports interoperability with graphics and media APIs including Microsoft DirectX, OpenGL, and the Intel Media SDK.
AMD today announced a collaboration with Adobe Systems Incorporated to optimize a new set of GPU-accelerated features for Adobe products including the newly announced Adobe Photoshop CS6. Implementing GPU acceleration and incorporating OpenCL optimization improves the end-user experience by dramatically speeding up critical imaging features and generating real-time results when editing with key tools in Adobe Creative Suite 6.
Adobe announced that they will be using their newly enhanced Adobe Mercury Playback Engine that incorporates OpenCL heterogeneous compute for the very first time on a number of Apple MacBook Pro laptops with AMD Radeon graphics.
The OpenCL hardware acceleration features in the Generic Graphics Library (GEGL) were originally developed as part of Google's Summer of Code in 2009 and 2011 and finished over the winter with sponsorship from AMD and help by MulticoreWare. On supported systems, this feature should greatly speed up processor intensive operations. Users have to start the application with the command line parameter GEGL_USE_OPENCL=yes to use OpenCL.
Altera Corporation announced that goHDR achieved a significant reduction in development time and a dramatic increase in performance leveraging Altera's OpenCL for FPGAs. Working closely with Altera, goHDR ported its proprietary C-code to the OpenCL standard and implemented the code in an FPGA in less than a week - a process that typically requires 3-6 months using a traditional HDL flow.
MulticoreWare Inc. has announced the Multicore Cross-platform Architecture (MXPA), an OpenCL runtime and compiler, enabling cross-architecture performance from a single, unified codebase. MXPA is designed to preserve the expression of parallelism in OpenCL, independent of vendor hardware and software stack implementation. MXPA currently supports x86 architectures and is being extended to a wide range of platforms from CPUs to DSPs with minimal vendor support.
Last December AMD released preview drivers that contained much of new the functionality defined in OpenCL 1.2. The new APP SDK includes beta drivers that contain a complete, beta level implementation of the OpenCL 1.2 specification for 32 and 64-bit Windows and Linux.