LLVM Weekly - #354, October 12th 2020
Welcome to the three hundred and fifty-fourth 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 to asb@asbradbury.org, or @llvmweekly or @asbradbury on Twitter.
I try to keep LLVM Weekly focused and to the point, but I do have to share that my incredibly talented wife Carrie Anne has been awarded an MBE for services to education. I couldn't be more proud of everything she's done for inclusive computing education.
News and articles from around the web
LLVM 11.0.0 has now been released. Thank you to everyone involved in preparing this release.
Luís Marques has written a blog post about using differential testing to rapidly find and fix missed optimisation opportunities in the RISC-V LLVM backend.
On the mailing lists
-
Paul C. Anagnostopoulos posted some proposals to enhance TableGen's ability to analyze and manipulate DAGs.
-
Serge Guelton has produced a prototype replacing llvm-build with pure CMake and is seeking feedback for this "llvmbuildectomy".
-
Sean M Bartell shared minutes from the code size birds of a feather session at the LLVM Dev Meeting.
-
Reid Kleckner posted notes from a chat about dbg.value at the LLVM Dev Meeting.
-
Juneyoung Lee distributed a follow-up and plan to a round-table discussion on undef and poison at the LLVM Dev Meeting.
-
Kristóf Umann shared a survey of dataflow analyses in Clang.
-
Andrzej Warzynski proposed an RFC on improving clang -help.
-
The GlobalISel roundtable at the LLVM Dev Meeting generated follow-up posts from Dominik Montada on register bank select and multi-stage legalisation.
-
Siva Chandra distributed a summary of discussions at an llvm-libc round table, including a Q&A with questions such as "What is the timeline you believe until at least one operating system is fully supported?" (answer: ~2 years until Clang can build fully against LLVM libc).
-
Adrien Guinet proposed an
apple_abifunction attribute to compiler functions for the Apple ARM64 ABI when targeting other ARM64 OSes. -
Jeroen Dobbelaere shared minutes from an Alias Analysis round table at the LLVM dev meeting.
LLVM commits
-
The vector reduction intrinsics are no longer "experimental". 322d0af.
-
Loop terminology in LLVM documentation has been revised with the addition of new diagrams. c3f12dd.
-
The AMDGPU MC layer now has improved detection of and error messages for unsupported instructions. e2452f5.
-
The ImplicitNullChecks pass was updated to support complex addressing mode. 35cb45c.
-
The LoopInterchange and FixIrreducible passes have been ported to the new pass manager. 9c21c6c, 0689dab.
-
Initial infrastructure was committed for the C-SKY backend. e1c38dd.
Clang commits
-
Documentation was added on APINotes. 85d5064.
-
A DirectoryWatcher implementation was added for Windows. 5d74c43.