The Programming Models group at BSC has published a new release of the OmpSs-2 programming model (version 2022.11).
In this release, we have defined a generic interface between the LLVM compiler and Nanos6 runtime to support kernels targeting hardware accelerators. Currently, only CUDA kernels are supported, but other programming models for accelerators will be supported in the future with the same generic interface. This release also contains several bug fixes and performance improvements that improve lock contention and reduce task creation overheads. Finally, we have added support for Ovni, a new instrumentation module that will become the default in the next release. You can check all changes here.