[llvm-dev] LLVM Weekly - #239, July 30th 2018

[Alex Bradbury via llvm-dev]

LLVM Weekly - #239, July 30th 2018
==================================

If you prefer, you can read a HTML version of this email at
<http://llvmweekly.org/issue/239>.

Welcome to the two hundred and thirty-ninth 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

GCC 8.2 has been
[released](https://gcc.gnu.org/ml/gcc-announce/2018/msg00002.html).


## On the mailing lists

* Kirill Bobyrev
[proposes](http://lists.llvm.org/pipermail/llvm-dev/2018-July/124888.html)
moving the google-benchmark library into the main LLVM repository. It's
currently included in libcxx and the test-suite, and clangd would now like to
use it. Responses are positive so far.

* Craig Topper starts a discussion on [improving the performance of dot
product reduction
loops](http://lists.llvm.org/pipermail/llvm-dev/2018-July/124786.html).

* Philip Reames
[shares](http://lists.llvm.org/pipermail/llvm-dev/2018-July/124829.html) a
couple of ideas for extensions of the loop unswitch cost model.

* Alexandros Lamprineas is [seeking
feedback](http://lists.llvm.org/pipermail/llvm-dev/2018-July/124871.html) on
re-enabling GVNHoist by default.

* Matthew Arsenault starts a discussion about [the behaviour of the
llvm.minnum/llvm.maxnum
intrinsics](http://lists.llvm.org/pipermail/llvm-dev/2018-July/124770.html).
He points out that the behaviour doesn't match IEEE-754 as the LangRef
currently suggests. Stephen Canon [helpfully
clarifies](http://lists.llvm.org/pipermail/llvm-dev/2018-July/124870.html)
recent changes to the IEEE-754 standard with respect to these operations.


## LLVM commits

* The MachineOutliner is now enabled by default under -Oz for AArch64. This
offers an additional 4.5% code size improvements.
[r338133](https://reviews.llvm.org/rL338133).

* A new ArmCodeGenPrepare backend pass was added in order to perform type
promotion on icmp operands. [r337687](https://reviews.llvm.org/rL337687).

* The BPF backend gained a new `-bpf-expand-memcpy-in-order` option.
[r337977](https://reviews.llvm.org/rL337977).

* MC support for the Armv8.2-A crypto extensions was added.
[r338010](https://reviews.llvm.org/rL338010).

* The RISC-V backend gained support for the 'interrupt' function attribute.
[r338047](https://reviews.llvm.org/rL338047).


## Clang commits

* Clang learned to use non-zero memset when possible for automatic variables.
[r337887](https://reviews.llvm.org/rL337887).

* Clang's OpenMP documentation was updated to describe recent changes.
[r338049](https://reviews.llvm.org/rL338049).


## Other project commits

* The compiler-rt profiling runtime is now supported on Fuchsia.
[r337881](https://reviews.llvm.org/rL337881).

* `<span>` was implemented in libcxx.
[r337804](https://reviews.llvm.org/rL337804).

* A design document was committed to libcxx for an upcoming `file_time_type`
change. The described change, using `__int128_t` was also committed this week.
[r337880](https://reviews.llvm.org/rL337880),
[r337960](https://reviews.llvm.org/rL337960).

* `<filesystem>` was implemented in libcxx. The symbols are placed in
libc++fs.a rather than libc++.so.
[r338093](https://reviews.llvm.org/rL338093).

* The hotspot and srad kernels from the Rodinia benchmark were added to the
LLVM test suite. [r338168](https://reviews.llvm.org/rL338168),
[r338169](https://reviews.llvm.org/rL338169).