Increasing Performance with Arrow and Gandiva
In this talk, Vivek will start with an overview of how Arrow represents columnar data; and how it is more efficient on modern processors. Then he will introduce Gandiva and explain: 1) How it uses LLVM and generates optimized compiled code for expressions; and 2) How it leverages SIMD instructions to gain performance. For demonstration, Vivek will use Dremioi, to show how a Data Lake engine, can use Gandiva for improved SQL query processing power. To wrap things up, Vivek will give a glimpse of on-going work in Gandiva, such as a project to improve code generation.