The Khronos Group has ratified OpenCL 1.1, a programming standard for parallel execution of tasks across multicore processors, the standards-setting organisation said. The OpenCL standard, which includes a C-like programming language with APIs (application programming interfaces), enables parallel task execution across hardware including CPUs and graphics processing units (GPUs). Apple, IBM, Intel, Nvidia and Advanced Micro Devices are among the companies promoting OpenCL.
OpenCL is gaining increased adoption as systems combine CPUs and GPUs to execute certain tasks faster. IBM and Dell recently said they would offer GPUs in servers as a way to boost the performance of specific scientific and commercial applications.
The new standard is an update from the OpenCL 1.0 standard, which was ratified in December 2008. The new standard adds additional functionality to provide better performance and programming flexibility, the Khronos Group said in a statement.
OpenCL competes with DirectCompute, an API from Microsoft that harnesses the parallel processing power of multicore CPUs and graphics chips. The GPU is known to be better at processing graphics-intensive applications as opposed to standard applications.
However, AMD officials have said regular applications like antivirus potentially could harness the parallel processing capabilities of GPUs. AMD plans to release laptop chips codenamed Fusion that integrate full graphics processors and CPUs inside a single chip. The chips are designed to support OpenCL on the GPU and CPU to provide improved system performance.
Graphics chip company Nvidia said it has already released an OpenCL 1.1 driver, available on its website to registered users. The download link was not immediately provided by the company.