[llvm-dev] LLVM Weekly - #306, November 11th 2019

Alex Bradbury via llvm-dev llvm-dev at lists.llvm.org
Mon Nov 11 14:07:33 PST 2019

LLVM Weekly - #306, November 11th 2019

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

Welcome to the three hundred and sixth 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.

As many of you know, I co-founded a not-for-profit collaborative engineering
company, lowRISC. We've had a [huge
last week, publicly announcing the world's first open source silicon root of
trust - OpenTitan.

## News and articles from around the web

The call for papers for the 4th LLVM performance workshop at CGO [has been
The deadline for submissions in teh 5th Jan, and the workshop will take place
on Feb 23rd in San Diego.

Herb Sutter has posted a [trip report from the Autumn ISO C++ standard

## On the mailing lists

* Mehdi Amini started a thread on [allowing LLVM contributions through GitHub
This generated a huge amount of discussion, pro and con.

* Levi Aul is seeking feedback on supporting an [unlimited width integer
David Chisnall [provided some

* Ehud Katz posted an RFC on [exposing the waymarking
from the 'Use' class as an independent utility.

* Bardia Mahjour sent out the [loop optimisation working group meeting
for last week.

## LLVM commits

* The 'freeze' instruction was added to LLM IR, which can be used to stop
propagation of undef and poison values.

* GlobalISel saw a number of documentation improvements including docs on
KnownBits analysis, debugging with the block extractor, and GMIR and
GenericOpCode references.

* SystemZ gained support for the GHC calling convention.

* DFAPacketizer was updated to allow up to 64 functional units.

* DWARFDebugLocLists started to moved to an incremental parsing model.

* InstCombine will now canonicalise shift+logic+shift to reduce the dependency
chain. [d9ccb63](https://reviews.llvm.org/rGd9ccb6367a1).

* The `-consider-local-interval-cost` options for the greedy register
allocator is now enabled by default for AArch64. This has very minor impact
which the commit message documents.

* Development of the data dependency graph continues with the addition of Pi
Blocks. [f0af11d8](https://reviews.llvm.org/rGf0af11d86f8).

## Clang commits

* Syntax nodes were added for most comment statements.

* Clang's sysroot computation will now do a better job for the RISC-V target
in the case that no GCC install is detected.

## Other project commits

* The old Python-based build system was dropped from libclc.

* std::midpoint was optimised for integers.

More information about the llvm-dev mailing list