[llvm-dev] LLVM Weekly - #338, June 22nd 2020

Mon Jun 22 10:17:33 PDT 2020

LLVM Weekly - #338, June 22nd 2020
==================================

If you prefer, you can read a HTML version of this email at
<http://llvmweekly.org/issue/338>.

Welcome to the three hundred and thirty-eighth 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

Applications are
[open](http://lists.llvm.org/pipermail/llvm-dev/2020-June/142322.html) for the
LLVM Foundation Board of Directors. They are due by 30th July and the new
board will begin their term on 12th August. Additionally, the BoD will be
having their [first open board
meeting](http://lists.llvm.org/pipermail/llvm-foundation/2020-June/000207.html)
on July 10th.

A virtual Austin area LLVM meetup is [being planned for Jun
24th](http://lists.llvm.org/pipermail/llvm-dev/2020-June/142577.html).

The first Workshop on LLVM Compilers and Tools for HPC [is taking place on
June 25th and
26th](http://lists.llvm.org/pipermail/llvm-dev/2020-June/142581.html) as a
free virtual event.

SCRT have written up a blog post on [using libTooling to transform source code
to evade anti-virus
detection](https://blog.scrt.ch/2020/06/19/engineering-antivirus-evasion/).

The Fedora community are discussing a [proposed policy
change](https://fedoraproject.org/wiki/Changes/CompilerPolicy) which would
allow some packages to built using Clang/LLVM.

## On the mailing lists

* Chris Lattner posted an RFC on [introducing an LLVM "incubator"
process](http://lists.llvm.org/pipermail/llvm-dev/2020-June/142548.html),
allowing a straight-forward path for projects to be started within LLVM
without being added to the monorepo.

* Mehdi Amini started a discussion on [renaming the `master`
branch](http://lists.llvm.org/pipermail/llvm-dev/2020-June/142445.html). This
generated significant (but not universal) positive response, with much
discussion centering around the timeline for any move and considerations such
as matching GitHub's new default branch name.

* Manuel Klimek no longer has time to maintain the LLVM project's Phabricator
instance and [proposes two options going
forwards](http://lists.llvm.org/pipermail/llvm-dev/2020-June/142458.html) -
either another volunteer steps up or LLVM switches to GitHub PRs. Many
responses to the thread discussed the pros/cons of code review on GitHub.

* Y Song started a thread on [options for disabling certain LLVM optimisations
in the Clang
frontend](http://lists.llvm.org/pipermail/cfe-dev/2020-June/065894.html). Eli
Friedman explained his concern that [adding special-purpose attributes doesn't
solve the underlying
problem](http://lists.llvm.org/pipermail/cfe-dev/2020-June/065910.html). Y
Song followed up with [two possible paths
forwards](http://lists.llvm.org/pipermail/cfe-dev/2020-June/065935.html).

* Baptiste Saleil posted an RFC on [supporting the new PowerPC MMA
instructions in
Clang/LLVM](http://lists.llvm.org/pipermail/cfe-dev/2020-June/065918.html).

* Louis Dionne kicked off an RFC thread on [adding a top-level monorepo CMake
file](http://lists.llvm.org/pipermail/llvm-dev/2020-June/142384.html).

* Djordje shared a proposal to [introduce an LLVM debug info checker
utility](http://lists.llvm.org/pipermail/llvm-dev/2020-June/142331.html) to
check the preservation of debug info through optimisations. Responses
discussed the pros/cons of this approach vs debugify.

* John Regher [proposed setting up
alive.llvm.org](http://lists.llvm.org/pipermail/llvm-dev/2020-June/142328.html).
It is [now live](https://alive2.llvm.org/).

* Roman Lebedev proposed [adding integer intrinsics for integer abs, min,
max](http://lists.llvm.org/pipermail/llvm-dev/2020-June/142257.html) providing
a long list of bugs and problems that can be attributed to now having these
intrinsics.

* All pre-merge checks [are now running on
BuildKite](http://lists.llvm.org/pipermail/llvm-dev/2020-June/142301.html).

* Renato Golin started a discussion on [dealing with spam in the LLVM IRC
channel](http://lists.llvm.org/pipermail/llvm-dev/2020-June/142291.html).

* Kristof Beyls [shared the mitigations that have been implemented for
straight-line speculation vulnerability
CVE-2020-13844](http://lists.llvm.org/pipermail/llvm-dev/2020-June/142109.html)
(the thread is a couple of weeks old, apologies I missed it previously).

* Jeremy Morse shared a lengthy proposal on [using a new LiveDebugValues
implementation](http://lists.llvm.org/pipermail/llvm-dev/2020-June/142368.html)
that aims to produce richer information and be faster.

* Erich Keane started a thread to discuss [codifying LLVM's preferred style
for braces in single line
blocks](http://lists.llvm.org/pipermail/llvm-dev/2020-June/142271.html).

## LLVM commits

* The AArch64 backend gained intrinsics and codegen for BFloat load/store and
matrix multiplication.[508a476](https://reviews.llvm.org/rG508a4764c0e),
[10b6567](https://reviews.llvm.org/rG10b6567f497).

* The WebAssembly backend gained 64-bit versions of all load and store
operations. [d9e0bbd](https://reviews.llvm.org/rGd9e0bbd17b0).

* LLVM documentation was updated to specify rules for updating debug
locations. [b4459b5](https://reviews.llvm.org/rGb4459b597a6).

* Support was added for efficient lookups of AssertingVH/PoisoningVH.
[5f565c0](https://reviews.llvm.org/rG5f565c04194).

* A `--dwarf64` option was added to llvm-mc to generated DWARF64 debug info.
[a8ec9de](https://reviews.llvm.org/rGa8ec9de406e).

* CallPrinter gained heat colouring.
[369d00d](https://reviews.llvm.org/rG369d00df60d).

* The MVE tail-predication pass was implemented using
@llvm.get.active.lane.mask. [d152251](https://reviews.llvm.org/rGd1522513d4c).

* The Arm backend was updated to support lowering of half-precision FP
arguments and returns. [a255931](https://reviews.llvm.org/rGa255931c405).

## Clang commits

* A webkit.UncountedCallArgsChecker was added to the static analyzer.
[a7eb369](https://reviews.llvm.org/rGa7eb3692e76).

* libTooling gained a parser for a string representation of `RangeSelector`.
[9ca50e8](https://reviews.llvm.org/rG9ca50e887db).

## Other project commits

* The OMPT-Multiplexing header file was added to overcome the restriction of
loading a single tool at once.
[e9b8ed1](https://reviews.llvm.org/rGe9b8ed1fd7c).

* Flang gained infrastructure for unit tests using Google Test.
[93f602b](https://reviews.llvm.org/rG93f602b339f).

* MLIR patterns can now match any root kind.
[80d7ac3](https://reviews.llvm.org/rG80d7ac3bc7c).

* LLVM's libc gained strcmp and long double fabsl and truncl.
[05e1612](https://reviews.llvm.org/rG05e1612c434),
[1390182](https://reviews.llvm.org/rG139018265bf).

* An implementation of `<numbers>` was added to libc++.
[4f6c4b4](https://reviews.llvm.org/rG4f6c4b473c4).