[llvm-dev] LLVM Weekly - #382, April 26th 2021

Mon Apr 26 11:07:04 PDT 2021

LLVM Weekly - #382, April 26th 2021
===================================

If you prefer, you can read a HTML version of this email at
<http://llvmweekly.org/issue/382>.

Welcome to the three hundred and eighty-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](https://www.linkedin.com/in/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 at asbradbury.org>, or
@llvmweekly or @asbradbury on Twitter.

## News and articles from around the web

Josh Haberman blogged about [making use of the new musttail attribute in Clang
for protobuf
parsing](https://blog.reverberate.org/2021/04/21/musttail-efficient-interpreters.html).

## On the mailing lists

* Christian Kühnel queries whether [anyone is doing code review on the mailing
  lists any
  more](https://lists.llvm.org/pipermail/llvm-dev/2021-April/150129.html),
  triggering a discussion about whether reviews by email should be phased out.
  Philip Reames [pointed
  out](https://lists.llvm.org/pipermail/llvm-dev/2021-April/150134.html) that
  the commit lists do have email responses that don't make it to Phabricator
  or post-commit email discussions.

* David Blaikie started a discussion on [refining the definition of optnone
  with interprocedural
  transformations](https://lists.llvm.org/pipermail/llvm-dev/2021-April/149960.html).
  A number of respondents were in favour of adding a 'noipa' attribute. Paul
  Robinson wrote up [a
  proposal](https://lists.llvm.org/pipermail/llvm-dev/2021-April/150107.html)
  on optnone, following extensive discussion.

* Michał Górny posted an RFC thread on [improving protocol-level compatibility
  between LLDB and
  GDB](https://lists.llvm.org/pipermail/lldb-dev/2021-April/016836.html).

* Paul C. Anagnostopoulos is [seeking
  feedback](https://lists.llvm.org/pipermail/llvm-dev/2021-April/150146.html)
  on the return value of the `!find` TableGen operator when there is no match.

* Tobias Hieta [sought feedback on proposals to better automate LLVM
  releases](https://lists.llvm.org/pipermail/llvm-dev/2021-April/150064.html).

* The discussion about emitting unaligned vector moves
  continued with Paul Robinson [pointing
  out](https://lists.llvm.org/pipermail/llvm-dev/2021-April/149986.html) that
  the previous behaviour is needed in some cases for backwards compatibility.

* Arthur Eubanks split off a thread to discuss [deprecation of the old pass
  manager](https://lists.llvm.org/pipermail/llvm-dev/2021-April/150100.html).

## LLVM commits

* The M68k disassembler was committed.
  [2221185](https://reviews.llvm.org/rG2221185776c3).

* The CSKY backend gained an assembly parser and support for basic
  instructions. [8ba622b](https://reviews.llvm.org/rG8ba622bae114),
  [4216389](https://reviews.llvm.org/rG4216389c2653),
  [4bb60c2](https://reviews.llvm.org/rG4bb60c285cb3).

* TableGen now supports the assert statement in multiclasses.
  [a5aaec8](https://reviews.llvm.org/rGa5aaec8f4e89).

* A new peephole optimisation was added to the AArch64 backend to remove
  redundant cmp after cset.
  [2ec1610](https://reviews.llvm.org/rG2ec16103c685).

* An implementation of MacOS's 'otool' was added.
  [1a3f886](https://reviews.llvm.org/rG1a3f88658a02).

* Documentation was added on LLVM's CI system.
  [7f9717b](https://reviews.llvm.org/rG7f9717b922d4).

* The memcpy optimiser was updated to work for variable sizes, as long as the
  sizes in the memcpy-memcpy or memcpy-memset pairs are equal.
  [f5446b7](https://reviews.llvm.org/rGf5446b769a79).

* IR intrinsics were added for the RISC-V Zbp extension.
  [b49337b](https://reviews.llvm.org/rGb49337bbb9de).

* The X86 backend gained support for AMX fast register allocation.
  [3b8ec86](https://reviews.llvm.org/rG3b8ec86fd576).

## Clang commits

* The new amdgpu-arch tool can list all AMD GPUs installed.
  [15be0c4](https://reviews.llvm.org/rG15be0c41d2e5).

* clang-scan-deps gained support for clang-cl.
  [8c16c8b](https://reviews.llvm.org/rG8c16c8b7ef7b).

* Collecting a profile without a filesystem is now documented.
  [d4ee603](https://reviews.llvm.org/rGd4ee603c8f21).

* The `clang_builtin_alias` attribute was introduced, allowing an alias to be
  introduced for a Clang builtin.
  [1088643](https://reviews.llvm.org/rG108864397d26).

## Other project commits

* A Language Server Protocol implementation and vscode language extension was
  added for MLIR. [751c14f](https://reviews.llvm.org/rG751c14fc4218),
  [b810e3a](https://reviews.llvm.org/rGb810e3a552e2),
  [52fad38](https://reviews.llvm.org/rG52fad38d28c1).

* A status page was added for libcxx's Format library.
  [ca177dc](https://reviews.llvm.org/rGca177dcda301).

* Initial ORC runtime directories and build system files were added to
  compiler-rt. [5e537ea](https://reviews.llvm.org/rG5e537ea1d7a9).

* Platform-specific implementations of fma and fmaf were added to LLVM's libc.
  [95934c3](https://reviews.llvm.org/rG95934c3a37cb).