LLVM Weekly - #522, January 1st 2024
Welcome to the five hundred and twenty-second 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 http://llvmweekly.org and pass it on to anyone else you think may be interested. Please send any tips or feedback to asb@asbradbury.org, @llvmweekly or @asbradbury on Twitter, or @llvmweekly@fosstodon.org or @asb@fosstodon.org.
Ten years of LLVM Weekly!
Today marks a massive milestone, LLVM Weekly has now been published every single week (without failure) for the past ten years. Thank you for reading! I’ve written up some thoughts and reflections on the newsletter, and also tried to answer some common questions about how it’s put together.
It has of course been a quiet week in terms of commits and development discussions due to many contributors being on holiday, so the issue is a bit more brief than usual.
News and articles from around the web and events
Yet another batch of videos from the 2023 LLVM Developers' Meeting were posted on YouTube.
Serge Guelton wrote up some in-depth experimental with
-ftrivial-auto-var-init.
If you’ve ever wanted the ability to convert simple functions defined in LLVM IR to an Excel spreadsheet performing the same computation, then Jordan Rose has you covered with CellLVM.
At the time of writing, reviews.llvm.org is down. However, Fangrui Song has blogged about work to move it to a static archive.
According to the LLVM calendar in the coming week there will be:
- Office hours with the following hosts: Anastasia Stulova, Quentin Colombet, Johannes Doerfert.
- Online sync-ups on the following topics: OpenMP, Flang, RISC-V, SPIR-V.
- For more details see the LLVM calendar, getting involved documentation on online sync ups and office hours.
On the forums
William Moses announced that EnzymeCon will take place March 25-27th at MIT and is accepting talk submissions until 20th Jan.
Nikolai Kholiavin pointed out that some uses of the IntrinsicInst class are potentially undefined behaviour. Nicolai Hähnle agreed with this assessment and suggested a potential solution involving migrating towards “Ref” classes.
Returning to a previous discussion on type annotations for libclang’s Python binding, “linux4life798” pointed to an issue they opened on moving this forward (which has also generated some discussion).
LLVM commits
Codegen of overflow arithmetic intrinsics for M68k was improved. b80e1ac, 2476e2a.
In the X86 backend, MC layer support for APX (providing additional GPRs) variants of arithmetic instructions were added. d79ccee.
MC layer support was added for the experimental RISC-V Zimop (‘may-be-operations’) and Zcmop extensions. 1d76692, 13cdee9.
MC layer support was introduced for the RISC-V Zicfiss (shadow stack) extension. 3dc0638.
The
noundefattribute can now be deduced for return values. 1228bec.
Clang commits
The Clang static analyzer release notes were prepared for the LLVM 18 release. d488adb.
The StdVariantChecker was documented. 1d2fab7.