Idiom-driven innermost loop vectorization in the presence of cross-iteration data dependencies in the HotSpot C2 compiler

• Main Author: Sjöblom, William
• Format: Others
• Language: English
• Published: Linköpings universitet, Programvara och system 2020
• Subjects: compiler; vectorization; SIMD; Java; HotSpot; code optimization; reductions; prefix sums; parallel programming; data-level parallelism; Computer Sciences; Datavetenskap (datalogi)
• Online Access: http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-172789

This thesis presents a technique for automatic vectorization of innermost single statement loops with a cross-iteration data dependence by analyzing data-flow to recognize frequently recurring program idioms. Recognition is carried out by matching the circular SSA data-flow found around the loop body’s φ-function against several primitive patterns, forming a tree representation of the relevant data-flow that is then pruned down to a single parameterized node, providing a high-level specification of the data-flow idiom at hand used to guide algorithmic replacement applied to the intermediate representation. The versatility of the technique is shown by presenting an implementation supporting vectorization of both a limited class of linear recurrences as well as prefix sums, where the latter shows how the technique generalizes to intermediate representations with memory state in SSA-form. Finally, a thorough performance evaluation is presented, showing the effectiveness of the vectorization technique.