[llvm-dev] LLVM Weekly - #146, Oct 17th 2016

[Alex Bradbury via llvm-dev]

LLVM Weekly - #146, Oct 17th 2016
=================================

If you prefer, you can read a HTML version of this email at
<http://llvmweekly.org/issue/146>.

Welcome to the one hundred and forty-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 schedule for the 2016 LLVM Developers' Meeting [is now
available](https://llvmdevelopersmeetingbay2016.sched.org/).

The LLVM Devroom at FOSDEM 2016 [has been
announced](http://lists.llvm.org/pipermail/llvm-dev/2016-October/105815.html).
The call for papers is now open and runs to December 1st.

Stefan Gränitz reports that the first LLVM hackday in Berlin will [take place
on October
22nd](http://lists.llvm.org/pipermail/llvm-dev/2016-October/106049.html) at
2pm.


## On the mailing lists

* Renato Golin reports that the survey about the proposed GitHub move is
[ready to
go](http://lists.llvm.org/pipermail/llvm-foundation/2016-October/000131.html).
It will hopefully soon be posted. In the time being, you can read the proposal
covering the move [here](http://llvm.org/docs/Proposals/GitHubMove.html).

* Justin Lebar has posted an RFC on [reducing the number of set classes in
LLVM's ADT
library](http://lists.llvm.org/pipermail/llvm-dev/2016-October/106079.html).
He proposes that we move to four set classes, down from the current ten.

* Johannes Doerfert has kicked off discussion in [preparation for the upcoming
BoF on parallel
IR](http://lists.llvm.org/pipermail/llvm-dev/2016-October/106051.html).

* Will Dietz has sharted discussion in [preparation for the upcoming BoF on
shipping software as LLVM
IR](http://lists.llvm.org/pipermail/llvm-dev/2016-October/106077.html).

* Paul Robinson
[reports](http://lists.llvm.org/pipermail/llvm-dev/2016-October/106063.html)
that the DWARF committee are taking comments on the draft of DWARF v5 until
November 30th.

* Zachary Turner has proposed added a new [type-safe formatting library in
LLVM's
codebase](http://lists.llvm.org/pipermail/llvm-dev/2016-October/105836.html).
Chandler Carrutth shared some [musings on why LLVM might want to build its
own](http://lists.llvm.org/pipermail/llvm-dev/2016-October/105856.html) rather
than rely on other popular C++ libraries.

* Teresa Johnson shared an [RFC on raising the minimum supported GCC version
to 4.8](http://lists.llvm.org/pipermail/llvm-dev/2016-October/105955.html).
There seems to be a lot of support for this.

* Peter Collingbourne posted an RFC on [improving support for references to
absolute
symbols](http://lists.llvm.org/pipermail/llvm-dev/2016-October/105800.html).


## LLVM commits

* Initial support for the 'RegCall' calling convention on X86 has landed. This
convention puts as many values as possible in registers.
[r284108](http://reviews.llvm.org/rL284108).

* The BitCodeFormat documentation has been updated.
[r284246](http://reviews.llvm.org/rL284246).

* Global variables representing each Target are now declared behind accessor
functions. Out of tree backends may want to replicate this change.
[r283702](http://reviews.llvm.org/rL283702).

* Support for tail duplication during placement was merged.
[r283934](http://reviews.llvm.org/rL283934).

* Loops with break statements can now be fully unrolled.
[r284053](http://reviews.llvm.org/rL284053).

* A new `llvm.addressofreturnaddress` intrinsic has been added. This is used
to implement the Microsoft `_AddressofReturnAddress` builtin.
[r284061](http://reviews.llvm.org/rL284061).

* A lengthy proposal on moving to GitHub has been added.
[r284077](http://reviews.llvm.org/rL284077).


## Clang commits

* Clang gained a new option, `-fsave-optimization-record`,to save the YAML
optimisation record to a file. [r283834](http://reviews.llvm.org/rL283834).

* Semantic analysis and code generation for C++17 'overaligned allocation' was
added. [r283789](http://reviews.llvm.org/rL283789).

* CUDA deferred diagnostics are now emitted during semantic analysis rather
than codegen. [r284158](http://reviews.llvm.org/rL284158).


## Other project commits

* libcxx gained `std::optional`. [r283980](http://reviews.llvm.org/rL283980).

* libcxx now maintains an ABI changelog.
[r284185](http://reviews.llvm.org/rL284185).

* C++17 aligned allocation functions have been added to libcxx.
[r284206](http://reviews.llvm.org/rL284206).