[llvm-dev] LLVM Weekly - #186, Jul 24th 2017

[Alex Bradbury via llvm-dev]

LLVM Weekly - #186, Jul 24th 2017
=================================

If you prefer, you can read a HTML version of this email at
<http://llvmweekly.org/issue/186>.

Welcome to the one hundred and eighty-sixth 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](http://asbradbury.org). 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

The LLVM 5.0.0 release branch [has been
created](http://lists.llvm.org/pipermail/llvm-dev/2017-July/115594.html), and
LLVM trunk is now 6.0.0. The bug tracking release blockers is
[here](https://bugs.llvm.org/show_bug.cgi?id=33849). The second release
candidate is scheduled for August 9th, and the release for August 23rd.

The 9th LLVM Social in Berlin [is
scheduled](http://lists.llvm.org/pipermail/llvm-dev/2017-July/115708.html) for
this Thursday at 7pm. Yegor Derevenets will be giving a talk about
decompilation. Be sure to
[register](https://www.meetup.com/LLVM-Social-Berlin/events/241197713) if you
would like to attend.


## On the mailing lists

* River Riddle has an RFC on [adding a new interprocedural
outliner](http://lists.llvm.org/pipermail/llvm-dev/2017-July/115666.html).
Outlining finds common code sequences and extracts them to separate functions,
in order to reduce code size. River has shared the current implementation and
a range of code size benchmark results, as well as a [useful summary of
differences vs the MachineOutliner
pass](http://lists.llvm.org/pipermail/llvm-dev/2017-July/115697.html).

* Jakub Kuderski has shared an [update on DominatorTree changes and support
for incremental
updates](http://lists.llvm.org/pipermail/llvm-dev/2017-July/115530.html).

* Ayman Musa is [seeking input on recent efforts to generate X86 memory
folding tables from
TableGen](http://lists.llvm.org/pipermail/llvm-dev/2017-July/115734.html).

* Tim Northover kicked off a discussion on [GlobalISel legalization
guarantees](http://lists.llvm.org/pipermail/llvm-dev/2017-July/115614.html).


## LLVM commits

* An initial scheduling model for AMD's Zen microarchitecture has been added.
[r308411](http://reviews.llvm.org/rL308411).

* The AArch64RedundantCopyElimination pass learned to removed more zero
copies. [r308849](http://reviews.llvm.org/rL308849).

* Support has been added for the IBM z14 processor.
[r308194](http://reviews.llvm.org/rL308194),
[r308195](http://reviews.llvm.org/rL308195),
[r308196](http://reviews.llvm.org/rL308196).

* Documentation was added on debugging instruction scheduling models.
[r308476](http://reviews.llvm.org/rL308476).

* The size of a ThinLTO minimized bitcode file has been further reduced by
stripping unneeded information. [r308760](http://reviews.llvm.org/rL308760).


## Clang commits

* An initial version of clang-diff has been committed. This is part of a GSoC
project, allowing a structural AST diff.
[r308731](http://reviews.llvm.org/rL308731).

* Clang's CMake scripts have been improved so the version information can be
extracted when building as a git submodule.
[r308205](http://reviews.llvm.org/rL308205).

* The huge lib/Basic/Targets.cpp file has been split up into per-target files.
[r308791](http://reviews.llvm.org/rL308791).

## Other project commits

* GDB index creation in LLD Has been sped up. Thanks to this, linking clang
with gdb index now takes 20.8 rather than 34.1 seconds.
[r308544](http://reviews.llvm.org/rL308544).

* A recursive taskloop scheduling strategy has been added for OpenMP
taskloops. [r308338](http://reviews.llvm.org/rL308338).

* Polly gained a new ForwardOptree pass, which can reduce scalar dependences.
[r308825](http://reviews.llvm.org/rL308825).