Support RAJA and Scientific Applications on RVV Architectures

Project Snapshot

In this work, we aim to make RVV more accessible to scientific applications by integrating it into the RAJA performance-portability framework. RAJA is a C++ library primarily developed at Lawrence Livermore National Laboratory that offers loop-based abstractions and multiple execution backends to deliver portable performance across heterogeneous systems. We contribute a new RVV backend to RAJA’s vectorization API, enabling RVV-aware optimizations within RAJA-based applications.