[llvm-dev] LLVM Weekly - #390, June 21st 2021

[Alex Bradbury via llvm-dev]

LLVM Weekly - #390, June 21st 2021
==================================

If you prefer, you can read a HTML version of this email at
<http://llvmweekly.org/issue/390>.

Welcome to the three hundred and ninetieth 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

Uma Srinivasan will be presenting "From Packing Decimals With Cobol to
Optimizing Tweets With Scala: A Journey Through Space, Time, & Culture with
Compilers" at the [Women in Compilers and Tools June
Meetup](https://www.meetup.com/meetup-group-ifwtlvwd/events/278899544/). June
24th, 6pm PDT.

A new book on LLVM has been published - [Learn LLVM 12 by Kai
Nacke](https://www.packtpub.com/product/learn-llvm-12/9781839213502).


## On the mailing lists

* The [thread suggesting a migration to
  Discourse](https://lists.llvm.org/pipermail/llvm-dev/2021-June/150823.html)
  has seen continued discussions. To pick a few messages:
  * Christian Kühnel
    [shared](https://lists.llvm.org/pipermail/llvm-dev/2021-June/151172.html)
    a [draft migration
    user-guide](https://github.com/llvm/llvm-iwg/blob/main/discourse_migration/userguide.md).
  * James Y Knight
    [reported](https://lists.llvm.org/pipermail/llvm-dev/2021-June/151150.html)
    that the [Open Source Community
    Infrastructure](https://www.osci.io/tenants/) would be willing to provide
    Mailman3 hosting.
  * John McCall provided a [summary of issues with mailing
    lists](https://lists.llvm.org/pipermail/llvm-dev/2021-June/151215.html).
    "Now, forums have their own usability annoyances, without question. [...]
    My point is that those problems should not be treated as blockers when we
    have equal or worse problems with mailing lists that we've just come to
    accept."

* Reid Kleckner [summarised responses to the RFC on
  LLD-as-a-library](https://lists.llvm.org/pipermail/llvm-dev/2021-June/151184.html).

* Tom Stellard announced that [12.0.1-rc2 has been
  tagged](https://lists.llvm.org/pipermail/llvm-dev/2021-June/151223.html) and
  an extended deadline for requesting backports.

* Matt Morehouse responded to an old thread on first-class metadata with a
  [list of potential use
  cases](https://lists.llvm.org/pipermail/llvm-dev/2021-June/151198.html).

* Gulfem Savrun Yeniceri
  [proposes](https://lists.llvm.org/pipermail/llvm-dev/2021-June/151154.html)
  adding a binary id to profiles, in order to aid matching executables to
  profiles.

* John McCall responded to the RFC thread on byte types in LLVM, [summarising
  how this may relate to work to add a provenance model to
  C](https://lists.llvm.org/pipermail/llvm-dev/2021-June/151199.html), and
  arguing that it would be better to start by considering inttoptr and
  ptrtoint casts.

* Luke Drummond started a conversation about [when it is appropriate to use the
  [NFC] tag](https://lists.llvm.org/pipermail/llvm-dev/2021-June/151234.html).

* LLVM GPU News Issue #14 [is
  out](https://lists.llvm.org/pipermail/llvm-dev/2021-June/151294.html).

* Kito Cheng posted an [RFC on providing a #pragma to implement the
  riscv_vector.h
  contents](https://lists.llvm.org/pipermail/cfe-dev/2021-June/068340.html),
  avoiding the parsing of ~60k lines. David Rector [pointed
  to](https://lists.llvm.org/pipermail/cfe-dev/2021-June/068346.html) recent
  work to solve a similar problem in OpenCL.


## LLVM commits

* A CustomBehaviour class was added llvm-mca, intended to allow developers to
  implement modelling for instructions that require special handling.
  [f7a23ec](https://reviews.llvm.org/rGf7a23ecece52).

* The SpeculateAroundPHIs pass was removed, with the commit message noting
  some of the regressions observed with the NewPM transition may have been
  caused by it. [e523645](https://reviews.llvm.org/rGe52364532afb).

* `DW_CFA_LLVM_` was added, as part of the work to add extensions for
  heterogeneous debugging on AMDGPU.
  [167e7af](https://reviews.llvm.org/rG167e7afcd52b).

* DebugInfo handling was added for IR outlined functions.
  [2c21278](https://reviews.llvm.org/rG2c21278e7471).

* Initial GlobalISel infrastructure was added for M68k.
  [2193347](https://reviews.llvm.org/rG2193347e72fa).

* Relocation emission for RISC-V was refactored and improved, bringing it more
  closely in line with GNU as.
  [bbea642](https://reviews.llvm.org/rGbbea64250f65).

* The MIRPrinter gained machine metadata support.
  [b9c05af](https://reviews.llvm.org/rGb9c05aff205b).


## Clang commits

* Improvements were made for named return value optimisation.
  [12c90e2](https://reviews.llvm.org/rG12c90e2e25df).

* The performance-unnecessary-copy-initialization checker was updated to fix
  some false positive cases.
  [bdd5da9](https://reviews.llvm.org/rGbdd5da9dec61).


## Other project commits

* EmitC, an MLIR dialect for emitting C++, was upstreamed to MLIR.
  [876de06](https://reviews.llvm.org/rG876de062f946).

* Identical code folding was implemented in lld-macho.
  [f27e454](https://reviews.llvm.org/rGf27e4548fc42).

* An implementation of a two-level distributed barrier was added to LLVM's
  OpenMP library. [25073a4](https://reviews.llvm.org/rG25073a4ecfc9).

* LLVM's libc gained implementation of a set of abstract elementary operations
  acting on fixed chunks of memory, intended to be used as building blocks for
  higher level abstractions.
  [454d92a](https://reviews.llvm.org/rG454d92ac3b3b).