[llvm-dev] LLVM Weekly - #142, Sep 19th 2016

Alex Bradbury via llvm-dev llvm-dev at lists.llvm.org
Mon Sep 19 05:17:47 PDT 2016

LLVM Weekly - #142, Sep 19th 2016

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

Welcome to the one hundred and forty-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](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 new board of directors for the LLVM Foundation has [been
It consists of 8 members (5 continuing and 3 new members). Congratulations to
all those elected.

Dean Michael Berris is [organising another LLVM developer meet-up in
This will take place on October 11th.

Neil Henning has shared the [next
of his blog series on compiling the toy 'Neil' language using LLVM.

## On the mailing lists

* Chris Lattner has
the discussion on moving LLVM projects to the Apache license. The current
proposal involves an exception to allow use of LLVM code with GPLv2 software.

* Hemant Kulkarni has posted an RFC on [disassembling mixed ARM and thumb in a
ELF](http://lists.llvm.org/pipermail/llvm-dev/2016-September/104942.html) by
making use of information in the `.ARM.attributes` section.

* Can def-use chains be used for static data dependency analysis? As John
yes as long as everything is in virtual registers. Otherwise you'll want to
use a iterative data-flow analysis and alias analysis.

* Chris Lattner is [not
on the proposal to add support for specifying named regex patterns in
FileCheck. He proposes instead just having several baked-in 'modes', that
expose some predefined patterns.

* Many people are not aware of this limitation so I thought it was worth
signal-boosting this post. [Comments on patch sent by email won't always be
picked up by
Specifically, Phabricator will drop email responses with an inline response.
Top-posting seems to mostly work.

* Vivek Pandya has posted an RFC on [an extension to existing register

## LLVM commits

* MCInstrDesc gained a new field, `isAdd`. For Hexagon, this is used to
identify loop induction statements.

* llvm-objdump gained the `--start-address` and `--stop-address` options.

* There is now a new metadata representation for global variables.

* All functions related to branch analysis have been renamed to conform to the
LLVM coding style. [r281535](http://reviews.llvm.org/rL281535).

* The documentation on compiling CUDA with Clang has been updated to discuss
STL support and the differences between nvcc and clang.

## Clang commits

* A new command-line option, `-fdiagnostics-show-hotness` was added. This
means any optimisation remarks will also include hotness information if
available from PGO. [r281276](http://reviews.llvm.org/rL281276).

* clang-tblgen gained a model to generate reference documentation for warning
and remark flags. [r281192](http://reviews.llvm.org/rL281192).

* A new clang-tidy check, 'misc-use-after-move' has been introduced. This
warns if an object is used after it has been moved.

## Other project commits

* Work in the direction of supporting breakpoint save and restore in LLDB has
been committed. [r281273](http://reviews.llvm.org/rL281273).

* Work has started to use lit for the LLDB test suite.

More information about the llvm-dev mailing list