LLVM Weekly - #641, April 13th 2026
Welcome to the six hundred and forty-first issue of LLVM Weekly, a weekly newsletter (published every Monday) covering developments in LLVM, Clang, and related projects. LLVM Weekly is brought to you by Alex Bradbury. Subscribe to future issues at https://llvmweekly.org and pass it on to anyone else you think may be interested. Please send any tips or feedback via email: asb@asbradbury.org, or Mastodon: @llvmweekly@fosstodon.org / @asb@fosstodon.org, or Bluesky: @llvmweekly.org / @asbradbury.org.
I’m in Dublin for EuroLLVM this week - hopefully I’ll see a bunch of you here!
News and articles from around the web and events
Kavin Gnanapandithan blogged about tracking down and fixing a performance regression in LLVM RISC-V code generation. It walks through the whole process of spotting a regression on Igalia’s LNT instance, analysing the problem, and finding and patching the issue.
Fangrui Song wrote about recent performance improvements to the LLD ELF
linker.
Comparing current main against LLVM 22.1 “Release+Asserts clang
–gc-sections link is 1.37x as fast as lld 22.1; Chromium debug with
–gdb-index is 1.07x as fast.” Read the post for more details, as well as why
mold and wild are faster still.
According to the LLVM Calendar in the coming week there will be the following (note, it’s possible some may be impacted by EuroLLVM):
- Office hours with the following hosts: Aaron Ballman, Alexey Bader, Alina Sbirlea, Phoebe Wang, Johannes Doerfert.
- Online sync-ups on the following topics: Flang, MLIR C/C++ frontend, LLVM qualification, modules, libc++, lifetime safety, LLVM/Offload, Clang C and C++ language working group, OpenMP for flang, HLSL, reflection.
- For more details see the LLVM calendar, getting involved documentation on online sync ups and office hours.
On the forums
“Zorojuro” wondered about having an auto-formatter for CMake. Respondents are interested in principle, but would like to see a specific tool proposed and an evaluation of of its output.
I don’t generally repost job postings, but I’ll make an exception for the LLVM Foundation who are hiring a DevOps Infra Engineer. For other job opstings, be sure to check out the job postings category on Discourse.
Ferdinand Lemaire provided an update on the MLIR dialect for WebAssembly, pointing to a PR that’s looking for more reviewers and noting that it’s still being pushed forward, but with lower priority due to changes in plans.
LLVM 22.1.3 was released.
Fabian Ritter started an RFC discussion on clarifying LLVM IRs concurrent memory model, proposing several changes to the LangRef.
LLVM commits
inline_historymetadata was introduced, which can be attached to a call instruction and will indicate if the call instruction itself has been inlined from the referenced functions. be62f27.An
llvm-testing-toolsPython package was added to enable packagingsplit-fileandFileCheckfor distribution on PyPI. a974f0a.A Hexagon ELF backend was added to JITLink. 0d64040.
The work to remove trivial VP intrinsics from the RISC-V backend continues. 7c1805d, f1aa984.
llvm.masked.{udiv,sdiv,urem,srem}intrinsics were added. 9ba7745.Common library functions like malloc, calloc, printf were implemented in the UB-aware IR interpreter. f149ab6.
LLVM’s CMake build system now supports statically linking libxml2. 11a8026.
Clang commits
-Wnon-portable-include-path-separatorwas added to warn on#includepath with backslashes. 9a1860c.Inline HLSL intrinsics were moved to TableGen. 88af280.
A patch landed that undoes some of the performance regression for concepts between clang-21 and clang-22. c73872a.