[llvm-dev] LLVM Weekly - #258, December 10th 2018

Alex Bradbury via llvm-dev llvm-dev at lists.llvm.org
Mon Dec 10 12:47:51 PST 2018

LLVM Weekly - #258, December 10th 2018

If you prefer, you can read a HTML version of this email at

Welcome to the two hundred and fifty-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

LLVM 7.0.1-rc3 [has been
The intent is that this is the last release candidate.

Stephen Kelly has blogged on [using build tools with

## On the mailing lists

* John Brawn
about a fascinating regalloc issue. A regression in bzip2 is triggered when
there is one more available register, resulting in worse allocation behaviour.
Quentin Colombet [provides some

* Adam Nemet [kicks
off](http://lists.llvm.org/pipermail/llvm-dev/2018-December/128322.html) a
second round of discussions on matrix support in LLVM. This new proposal takes
a more scaled back approach.

* Anastasia Stulova has started an RFC thread on [preserving the address space
of 'this' in C++

* Lingda Li is seeking feedback for an RFC on [the implementation of the
OpenMP 5.0 declare mapper

* Hans Wennborg has proposed the [8.0 release
which would see the branch taking place on 16th January and final tag on 27th

* James Y Knight is keen on [moving forwards with the plan for the git
After another week with [no extra
it looks like there may be consensus.

* Tom Stellard
that a bug fix for 7.0.0 will necessarily lead to an ABI change for a
clang-build libLLVM-7.so. He proposes leaving the fix from 7.0.1 and shipping
a special 7.0.2/7.1.0 release so users have the option of accepting this ABI
change or not.

* Michael Kruse
to no longer use LoopIDs as identifiers.

## LLVM commits

* A new llvm.experimental.widenable.condition intrinsic was added. This allows
the choice between two alternative correct solutions for the same problem or
to express guards. [r348593](https://reviews.llvm.org/rL348593).

* It is now possible to mark variadic register operands as defs.

* Initial support for the 'TAPI' TextAPI proposal landed. This allows
conversion between binary shared object stubs and textual counterparts.

* SimpleLoopUnswitch learned to preserve MemorySSA.

* A new interface was introduced to observe changes in GISel passes.

* The IR constant folder no longer assumes all functions are 4-byte aligned.

## Clang commits

* Support for pretokenized headers was removed. This was an experimental
feature that was ultimately considered failed.

* Portability of the Clang python utility scripts to Python 3 was improved.
[r348129](https://reviews.llvm.org/rL348129), and more.

* The iterator checker learned to warn on decrements past the begin() and
end() of containers. [r348425](https://reviews.llvm.org/rL348425).

## Other project commits

* DragonFlyBSD and OpenBSD support was added to the OpenMP runtime.

* The WebAssembly lld linker gained support for the event section specified in
the exception handling proposal. [r348703](https://reviews.llvm.org/rL348703).

More information about the llvm-dev mailing list