11
15
2026

Announcing OpenACC Technical Report 26-1

We’re pleased to announce the publication of the OpenACC Technical Report 26-1 at the SC26 conference. This work culminates 18 months of effort following the release of OpenACC 3.4 in June 2025. This technical report is a snapshot of the progress the OpenACC Technical Committee has made on the specification, in hopes of gathering community feedback before we release our next version, targeted for 2027. While the text in this document is far from final and will most likely change, including potentially adding or subtracting features, the committee wanted to share the direction that OpenACC is heading.

Below you’ll find highlights of the major new changes in the specification.

     * Executable Graphs - A means to portably capture and execute reusable graphs that include data and compute operations.

     * User Annotations - The profiling interface has been expanded to include user-provided annotations, which allow OpenACC regions to be labeled for use by profiling tools.

     * Prefetching - A means of providing hints to the implementation when data should be fetched into hardware caches, rather than requiring the allocation of a special scratchpad, as supported by the cache directive.

     * Static & Unroll - New modifications were added to vector loops to allow unrolling before (static) or after (unroll) the partitioning of loop iterations to parallel execution units.

     * Additional clarifications to the existing specification.

We hope you—our community members and specification users—review and comment on the new features. Many of these features are being prototyped by our implementers in hopes of gathering both implementer and user feedback before the final approval of the next specification. Please send your feedback to feedback@openacc.org.

If you’re attending SC26 in Chicago, we invite you to discuss TR 26-1 and your experiences using OpenACC at the OpenACC Birds of a Feather (BOF) on Tuesday at 5:15 PM Central Time.

Author

Jeff Larkin
Jeff Larkin
Jeff Larkin is a Senior Developer Technologies Software Engineer at NVIDIA Corporation, where he focuses primarily on porting and optimizing HPC applications. Jeff is an active contributor to the OpenACC and OpenMP standards bodies and is the Chair of the OpenACC Technical Committee. Prior to joining NVIDIA, Jeff worked in the Cray Supercomputing Center of Excellence. He holds a M.S. in Computer Science from the University of Tennessee and a B.S. in Computer Science from Furman University.