[llvm-dev] LLVM Weekly - #141, Sep 12th 2016

[Alex Bradbury via llvm-dev]

LLVM Weekly - #141, Sep 12th 2016
=================================

If you prefer, you can read a HTML version of this email at
<http://llvmweekly.org/issue/141>.

Welcome to the one hundred and forty-first 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.

As well as having the LLVM Cauldron last week, it was also an [eventful
weekend](https://twitter.com/MissPhilbin/status/774985421914796033) so please
accept my apologies if I've missed anything important in this issue!


## News and articles from around the web

Last Thursday we successfully held the inaugural [LLVM
Cauldron](http://llvm.org/devmtg/2016-09/) in Hebden Bridge. I'd like to thank
all the speakers, attendees, sponsors, and helpers who made the event such a
success. Simon Cook from Embecosm very kindly recorded all the talks and has
managed a crazy-fast turnaround time to get them all edited and uploaded ready
for you to view. See [the video playlist
here](https://www.youtube.com/playlist?list=PLOGTP9W1DX5X9dNzT02d-CP-OWOR4kKpq).
Simon's been so fast I haven't even collected PDFs of the slides from all the
speakers yet, but you can expect to find them linked to on the [LLVM Cauldron
page](http://llvm.org/devmtg/2016-09/#schedule) in the next day or two.

Neil Henning has written up [part two of his series on compiling a toy
language using
LLVM](http://www.duskborn.com/cleaning-up-the-parser-mpc-llvm-for-the-neil-language-2/).
This post explains key concepts such as LLVM Values and Types, symbol tables,
and type tables.

[NLVM](https://github.com/arnetheduck/nlvm), a new LLVM-based compiler for the
[Nim programming language](http://nim-lang.org/) has been
[announced](http://lists.llvm.org/pipermail/llvm-dev/2016-September/104527.html).
Currently it passes 90% of the upstream Nim test cases.


## On the mailing lists

* The events of September 4th 2016 will be passed down through the generations
of LLVM developers as the day the bots got lazy (or at least, the [day we
noticed](http://lists.llvm.org/pipermail/llvm-dev/2016-September/104543.html)).
A configuration change to Zorg introduced a bug which made all builds rebuild
from the same revision, without including the latest changes. This did cause
[some
hassle](http://lists.llvm.org/pipermail/llvm-dev/2016-September/104564.html)
when the configuration was fixed and new failures surfaced. The good news is
the CIncident is now over and life can resume as normal.

* MSVC2013 has had a stay of execution, the plan is now to [retire support for
it until 15th
October](http://lists.llvm.org/pipermail/llvm-dev/2016-September/104683.html).

* Tom Stellard is managing the upcoming 3.9.1 stable release and has requested
that [requests for merges now go through bugzilla rather than the mailing
list](http://lists.llvm.org/pipermail/llvm-dev/2016-September/104587.html). As
he explains, this workflow makes things much easier for him to keep track of.

* Ever wondered what the difference is between LLVM's libunwind and 'nongnu'
libunwind? David Chisnall [has the
answer](http://lists.llvm.org/pipermail/cfe-dev/2016-September/050649.html).

* Daniel Cederman is a software engineer at Cobham Gaisler working on
toolchain support for LEON3/LEON4 (which implement SPARC). He has shared an
[RFC on adding support for REX, their new 16-bit dense instruction
encoding](http://lists.llvm.org/pipermail/llvm-dev/2016-September/104579.html).
These shorter instructions overlap with the 32-bit encoding space, so require
instructions to enter and exit REX mode.


## LLVM commits

* In GlobalISel, type information is now stored in MachineRegisterInfo rather
than in every MachineInstr. [r281035](http://reviews.llvm.org/rL281035).

* LLVM's OCaml bindings can now be built out-of-tree.
[r280642](http://reviews.llvm.org/rL280642).

* The C++ Itanium demangler from libcxxabi has been imported in to LLVM.
[r280732](http://reviews.llvm.org/rL280732).

* A new field has been added to the Instruction TableGen class. AsmVariantName
is used to indicate instructions that are only available in specific assembler
variants. [r280952](http://reviews.llvm.org/rL280952).


## Clang commits

* Work has started to support export declarations as specified in the C++
Modules TS. [r280999](http://reviews.llvm.org/rL280999).

* A Visual Studio extension for clang-tidy has been added to the
clang-tools-extra repository. For now it only supports graphically configuring
clang-tidy settings. [r280840](http://reviews.llvm.org/rL280840).

* The Microsoft pragma intrinsic extension (used to tell the compiler to use a
builtin version of the function) is now supported, though is somewhat
redundant in Clang as it will already use a builtin version of a function when
possible by default. [r280825](http://reviews.llvm.org/rL280825).


## Other project commits

* A new mechanism to improve crash info has landed in LLDB. Disassembly will
be parsed in order to try to reconstruct the sequence of dereferences and
offsets that were applied to a known variable to obtain an invalid pointer. It
is exposed through a 'frame diagnose' command.
[r280692](http://reviews.llvm.org/rL280692).

* In a monster commit, LLDB has been completely reformatted to conform to
clang-format's LLVM style. Python files have also been reformatted using
autopep8. [r280751](http://reviews.llvm.org/rL280751).

* Support for EfficiencySanitizer on MIPS64 has been added (with a few tests
still be fixed). [r280795](http://reviews.llvm.org/rL280795).

* LLDB has gained support for Intel's Memory Protection eXtensions (MPX).
Specifically, it can now access the MPX registers.
[r280668](http://reviews.llvm.org/rL280668).

* libc++ can now be built against an external threading library other than
pthreads. [r281179](http://reviews.llvm.org/rL281179).