[llvm-dev] LLVM Weekly - #302, October 14th 2019

Alex Bradbury via llvm-dev llvm-dev at lists.llvm.org
Mon Oct 14 11:50:39 PDT 2019


LLVM Weekly - #302, October 14th 2019
=====================================

If you prefer, you can read a HTML version of this email at
<http://llvmweekly.org/issue/302>.

Welcome to the three hundred and second 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

The Call for Papers for for the LLVM Dev Room at FOSDEM 2020 [is now
open](http://lists.llvm.org/pipermail/llvm-dev/2019-October/135735.html). The
deadline is the 24th November, and the devroom will take place on Saturday 1st
February in Brussels.

The next set of proposed papers for the C++ standard [have now been
posted](http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2019/#mailing2019-10).


## On the mailing lists

* Tom Stellard shares an [update on the GitHub migration schedule and
plans](http://lists.llvm.org/pipermail/llvm-dev/2019-October/135803.html). The
plan is to make SVN read-only on the evening of October 21st.

* Jeremy Morse [proposes that LLVM adopts Sony's Dexter tool and uses it to
run
debuginfo-tests](http://lists.llvm.org/pipermail/llvm-dev/2019-October/135773.html).

* Tanya Lattner
[confirms](http://lists.llvm.org/pipermail/llvm-dev/2019-October/135687.html)
that the LLVM Foundation Board of Directors has voted to accept MLIR as a
project into LLVM.

* Sam Elliott
[reports](http://lists.llvm.org/pipermail/llvm-dev/2019-October/135768.html)
that the GCC Torture Suite has now been imported into the LLVM test suite.

* Josef Eisl has [written a patch](https://reviews.llvm.org/D68213) to add
support for embedding bitcode in a binary during LTO. This follows [this
earlier
RFC](http://lists.llvm.org/pipermail/llvm-dev/2019-January/129657.html). He's
actively seeking reviewers.

* David Greene kicked off an RFC thread on [end-to-end
testing](http://lists.llvm.org/pipermail/llvm-dev/2019-October/135739.html).
This generated a lot of discussion that sadly I can't fully summarise.

* Galina Kistanova has sent out a [note to buildbot
owners](http://lists.llvm.org/pipermail/llvm-dev/2019-October/135870.html) to
prepare their bot for the git switchover.

* David Tellenbach [submitted a
patch](http://lists.llvm.org/pipermail/llvm-dev/2019-October/135731.html) to
enable scheduling of CFI instructions to try to address the issue where
generated assembly differs depending on whether debug info is generated, but
[notes](http://lists.llvm.org/pipermail/llvm-dev/2019-October/135813.html)
this approach seems to be a dead-end. He will next try to post-pone insertion
of CFI instructions to after machine scheduling.

* Serge Guelton is [seeking input on stack clash protection for X86 in
LLVM](http://lists.llvm.org/pipermail/llvm-dev/2019-October/135786.html).

* The LLVM Dev Meeting is taking place next week and [Tanya Lattner is seeking
volunteer
moderators](http://lists.llvm.org/pipermail/llvm-dev/2019-October/135720.html).

* Siva Chandra has posted the first RFC on the libc-dev mailing list,
[proposing "ground truth" description files for the public
API](http://lists.llvm.org/pipermail/libc-dev/2019-October/000003.html).


## LLVM commits

* A new pass adds support for elimination of dead virtual functions during
LTO.[r374539](https://reviews.llvm.org/rL374539).

* Support for SVE in LLVM IR continues with the addition of scalable vector
size queries and IR instruction support.
[r374042](https://reviews.llvm.org/rL374042).

* The Attributor can now deduce memory behaviour of functions and arguments.
[r373965](https://reviews.llvm.org/rL373965).

* The LLVM docs have seen more re-organisation, with the "subsystem
documentation" content moving to the User Guides and Reference pages.
[r373872](https://reviews.llvm.org/rL373872).

* Constrained intrinsics were added for lrint and lround.
[r373900](https://reviews.llvm.org/rL373900).

* `update_test_checks.py` not accepts `--preserve-names` which means IR names
aren't replaced. [r373912](https://reviews.llvm.org/rL373912).

* X86 gained a new "tailcc" calling convention that guarantees tail call
optimisation. [r373976](https://reviews.llvm.org/rL373976).

* `update_cc_test_checks.py` can now handle a `clang | opt | FileCheck`
pipeline. [r374287](https://reviews.llvm.org/rL374287).

* FileCheck now supports `--ignore-case`.
[r374538](https://reviews.llvm.org/rL374538).


## Clang commits

* The new `-fgnuc-version=` option allows control of `__GNUC__` and other GCC
macros. [r374449](https://reviews.llvm.org/rL374449).

* clang-format gained initial support for producing warnings for lines that
need reformatting. [r374663](https://reviews.llvm.org/rL374663).

* The Clang interface stubs tooling now has end-to-end support for generating
ELF interface stubs. [r374061](https://reviews.llvm.org/rL374061).

* Clang's OpenMP 5.0 implementation was updated with support for the "master
taskloop" directive. [r374437](https://reviews.llvm.org/rL374437).


## Other project commits

* The GCC C torture suite was imported into the LLVM test-suite. It can be
built using CMake and run using Lit.
[r374155](https://reviews.llvm.org/rL374155) and
[r374156](https://reviews.llvm.org/rL374156).

* LLDB learned to read symbols from `.gnu_debugdata` if present.
[r373891](https://reviews.llvm.org/rL373891).

* LLDB accelerator table lookups were sped up (accelerated!).
[r374401](https://reviews.llvm.org/rL374401).


More information about the llvm-dev mailing list