Thread: New programming standard for CPU / GPU systems

Together with Cray, the Portland Group (PGI) and CAPS Nvidia has formulated a latest programming , the use of heterogeneous CPU / GPU computing systems for parallel calculation is to simplify.Since it is in the involvement of companies to demonstrate a broad-based developer, is "OpenACC"an open standard.


OpenACC based in large part on the CIP and Cray-programming models and allows the developers of parallel programs, simple instructions in the form of tax statements - to pass to the compiler, in which the accelerating parts of the code to be identified - so-called directive.




The programmer must therefore make no change or adapt the underlying code. The directives allow the compiler in this way, the exact distribution of the work onto the accelerator.
Benefits through the company OpenAAC see for example in the fields of chemistry, biology, physics, data analysis and climate research.


Existing Cray compiler, PGI and CAPS to be the Nvidia CUDA support OpenACC-compatible standard in the first quarter of 2012. How developers can greatly benefit from using compiler directives-based, Nvidia tried on a website to show to report on the developer about their experiences.The far sighted advice on where to speak of enormous increases in application performance in a short time.


Developers, the directive-based programming are interested in may, for a month a free trial version of the PGI Accelerator Fortran and C compilers to use when they are at Nvidia or have registered at Cray.