[ELF] Speed up ICF by hashing relocations and propagating over an edge list (#233697) ICF's initial hash covers section contents and, since https://reviews.llvm.org/D56986, two rounds of relocation target hashes, but not what constantEq compares: offsets, types, addends, merge-section offsets, and names of undefined or preemptible targets. Sections that differ only in relocations, e.g. the string or external function they reference, share a hash at every depth, which is why D56986 found no gain from more rounds. segregate() separates them with a quadratic number of equalsConstant calls, which dominates a Chromium --icf=all link. Hash what constantEq compares, and propagate hashes over relocations to other candidates for 7 rounds. Each round saves up to one segregate() pass. Also parallelize the symbol fold loop. --threads=8, --icf=all, output byte-identical at --threads=1 and 8: ``` link ICF segregate passes Chromium (CREL) 3.46 -> 2.08 s 1614 -> 233 ms 9 -> 2 TensorFlow (RELA) 2.07 -> 1.89 s 290 -> 120 ms 4 -> 2 Firefox (RELA) 560 -> 490 ms 117 -> 47 ms 6 -> 2 clang (RELA) 240 -> 230 ms 37 -> 22 ms 5 -> 2 ``` LLM-aided
Welcome to the LLVM project!
This repository contains the source code for LLVM, a toolkit for the construction of highly optimized compilers, optimizers, and run-time environments.
The LLVM project has multiple components. The core of the project is itself called “LLVM”. This contains all of the tools, libraries, and header files needed to process intermediate representations and convert them into object files. Tools include an assembler, disassembler, bitcode analyzer, and bitcode optimizer.
C-like languages use the Clang frontend. This component compiles C, C++, Objective-C, and Objective-C++ code into LLVM bitcode -- and from there into object files, using LLVM.
Other components include: the libc++ C++ standard library, the LLD linker, and more.
Consult the Getting Started with LLVM page for information on building and running LLVM.
For information on how to contribute to the LLVM project, please take a look at the Contributing to LLVM guide.
Join the LLVM Discourse forums, Discord chat, LLVM Office Hours or Regular sync-ups.
The LLVM project has adopted a code of conduct for participants to all modes of communication within the project.