A team of experts from IT4Innovations, NVIDIA, the University of Cologne, and the University of California has developed a new approach to rendering scientific visualisations that uses the ANARI API for parallel processing of large sci-vis data. This approach allows computations to be distributed across multiple devices, speeding up and simplifying the entire rendering process.

Scientific visualisation has evolved into a complex field, facing challenges such as the efficient rendering of large datasets. This task has become even more demanding with the emergence of various, often incompatible APIs. Today, scientists are not only concerned with how to visualise data but also with how to do so as quickly and accurately as possible, without the need to switch between different systems.

Milan Jaroš from IT4Innovations, along with NVIDIA experts Ingo Wald, Jefferson Amstutz, and Kevin Griffin; Stefan Zellmann and Stefan Wesner from the University of Cologne; and Qi Wu from the University of California, presented a new approach for data-parallel rendering of scientific visualisations that leverages the existing ANARI API. ANARI (A New API for Rendering Interfaces) is an interface for 3D rendering that runs on individual compute nodes.

This newly proposed concept, called data-parallel ANARI (DP-ANARI), enables compute nodes to exchange information during the rendering process itself. The result is efficient rendering of complex scenes with global effects such as shadows or ambient occlusion, realised by ray tracing.

This approach remains fully compatible with existing applications such as VTK, ParaView, and VisIT, saving developers time and costs associated with migrating to new technology. It also extends the ability to create high quality scientific visualisations without the need to implement complex new systems.

This new concept paves the way for more powerful and enhanced scientific visualisations while preserving compatibility with existing tools. Data-parallel rendering with ANARI represents a significant step forward in scientific visualisation and computer graphics.

Scientific paper
Standardized Data-Parallel Rendering Using ANARI
https://arxiv.org/pdf/2407.00179

(*The illustration image for this article is a collage of images from the scientific paper.)

Jefferson Amstutz on portable and scalable 3D rendering using ANARI: