Vespa Newsletter, December 2021

The Vespa Newsletter from December 2021 highlights significant updates to the Vespa.ai platform, focusing on enhanced tensor performance and match features. The Euclidian distance calculations using int8 have become 250% faster due to hardware-accelerated instructions, optimizing performance for applications with large datasets by using less memory. Vespa 7.514.11 introduces optimized tensor field memory alignment, potentially improving query latency by reducing memory bandwidth usage. The introduction of match-features allows for the configuration of features returned from content nodes, avoiding unnecessary recalculations and enabling their use in further re-ranking processes. Additionally, a new IntelliJ plugin for editing schema files was contributed by Shahar Ariel, enriching the development experience.