JIT in ClickHouse

This blog post was originally posted on Maksim's personal blog, which we recommend for those interested in the low-level details of Performance Engineering, Query Analysis and Planning, JIT Compilation, System Programming and Distributed Systems.

In this post, I will describe what JIT compilation is, how LLVM infrastructure can be used for JIT compilation, and how JIT compilation works in ClickHouse.

Most of this post is a summary of talks that I give on C++ Russia 2021 “JIT in ClickHouse”, HighLoad 2022 “JIT compilation of queries in ClickHouse”, C++ Russia 2022 “ClickHouse performance optimization practices”, and there are also additional examples and things that I have not covered in these talks.

First, let’s start with what JIT compilation is. JIT - Just-in-time compilation. The main idea is to generate machine code and execute it in runtime. Examples of such systems are JVM Hotspot and V8. Most database systems support JIT compilation.

We have a sum function that takes two integer values, computes their sum, and returns it. In the main function, we print the result of the sum function execution with constants 1 and 2.

Leave a Comment