LLVM Cookbook

Vectorizing IR

Vectorization is an important optimization for compilers where we can vectorize code to execute an instruction on multiple datasets in one go. If the backend architecture supports vector registers, a broad range of data can be loaded into those vector registers, and special vector instructions can be executed on the registers.

There are two types of vectorization in LLVM—Superword-Level Parallelism (SLP) and loop vectorization. Loop vectorization deals with vectorization opportunities in a loop, while SLP vectorization deals with vectorizing straight-line code in a basic block. In this recipe, we will see how straight-line code is vectorized.

Getting ready

SLP vectorization constructs a bottom-up tree of the IR expression, and broadly ...