Draft of the Proposed Version 3.1 OpenMP Specification Available

The OpenMP ARB is pleased to announce the release of a draft of the OpenMP specification, version 3.1 for comment by the community. Please see: 

http://www.openmp.org/mp-documents/omp3.1-2011.0203a.pdf 

This draft will serve as the basis of an official 3.1 update of the specification, which is expected to be ratified in time for IWOMP 2011 in Chicago. All interested users and implementers are invited to review the comment draft and to provide feedback through the Draft 3.1 Public Comment section of the OpenMP Forum available at: 

http://openmp.org/forum/viewforum.php?f=9 

The 3.1 version is intended as a minor release that will not break existing, correct OpenMP applications. However, it does include several new features, most notably the addition of predefined min and max operators for C and C++ and extensions to the atomic construct that allow the value of the shared variable that the construct updates to be captured or to be written without being read. It also includes extensions to the OpenMP tasking model that support optimization of its use. 

"The release of this comment draft represents the culmination of significant effort on the part the OpenMP Language Committee," stated Language Committee Chair Bronis R. de Supinski. "We have added extensions that handle some of the most frequent user requests while also working to make the specification and its associated examples clearer. We anticipate making other minor changes to the draft as a result not only due to public comment but also to provide additional examples that will clarify the use of new features. We plan to make updates to the comment draft that capture these enhancements available during the comment period, which closes Monday, May 1, 2011." 

"Concurrent to our work on version 3.1, we have also been making progress on several significant enhancements to the specification that we expect to serve as the basis for version 4.0," de Supinski continued. "Topics under consideration include support for accelerators such as GPUs, major enhancements to the tasking model, mechanisms to support error handling and user defined reductions. I welcome inquiries from anyone interested in contributing to these directions."