LLVM Weekly - #437, May 16th 2022

LLVM Weekly - #437, May 16th 2022

If you prefer, you can read a HTML version of this email at http://llvmweekly.org/issue/437.

Welcome to the four hundred and thirty-seventh 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, or @llvmweekly or @asbradbury on Twitter.

Last week proved you can have too much of a good thing - apologies to those of you who received multiple copies of LLVM Weekly by email Mailgun have acknowledged a bug on their end that led to the email being sent multiple times. I’m assured the problem has now been resolved.

News and articles from around the web

Stefan Gränitz wrote up a EuroLLVM 2022 trip report.

On the forums

LLVM commits

  • Work is underway to remove TableGen’s reliance on global state. A recent patch removes the use of global Record state. 2ac3cd2.

  • A pass was added to work around erratum for the Cortex-A57 and Cortex-A72 related to fused AES instruction pairs. 3a24df9.

  • The semantics of the ADDCARRY/SUBCARRY ISDOpcodes were clarified. 3f64f03.

  • 8 and 16-bit switch conditions are now avoided where possible on X86. cd19af7.

  • The DXILOpLowering pass was added to lower LLVM intrinsics to DXIL ‘op’ functions. The DXILTranslateMetadata pass was also added. 85285be, 0c7f7f1.

  • The initial subtarget definition and early tests were added for AMDGPU gfx11. 18ed279.

  • A syntax highlighting description for TableGen in Textmate’s JSON format was committed. f0792c7.

Clang commits

  • Typo corrections are now suggested for preprocessor directives. a247ba9.

  • The clang-offload-packager tool (to bundle device files) was added. 26eb042.

  • The -ffile-reproducible flag was added. 6398f3f.

Other project commits

  • The Flang driver was moved to the MLIR coding style. 1e462fa.

  • Documentation on LLVM’s libc stdio functions was added. 12aae7d.

  • std::to_chars performance was improved for bases 2, 8, and 16. 6c11aeb.

  • An AMDGPU MLIR dialect was added. f1f05a9.

  • MLIR now supports operation agnostic pass managers. c2fb9c2.