LLVM Weekly - #91, Sep 28th 2015
Welcome to the ninety-first 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. 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@asbradbury.org, or @llvmweekly or @asbradbury on Twitter.
The canonical home for this issue can be found here at llvmweekly.org.
News and articles from around the web
Some slides and videos from cppcon have started to appear. See the slides from Bjarne Stroustrup's keynote (which focus on the Core Guidelines project), as well as Herb Sutter's slides on supporting ownership annotations in C++ (this will be familiar to anyone who has used Rust). Videos are starting to appear on the CppCon Youtube channel.
A Microsoft blog says something about their plans to rejuvenate MSVC. "We will continue to invest in improving our compiler with a goal towards making it fully standards compliant".
On the mailing lists
Jason Kim has posted a summary of a recent in-person meeting about ThreadSanitizer, Android, and AARch64.
Chris Matthews is planning to work on improving performance change tracking in LNT, and has posted an RFC on his plans.
The discussion/debate in the 'trouble with triples' thread has rumbled on. There's been a lot of in-depth discussion, which I'm afraid I don't have the time to study and try to summarise fairly. Do dive in if this topic is important to you. Otherwise, I hope we'll see a summary/RFC once a way forward is found.
Jia-Ju Bai is looking for a way to extract basic loop information from LLVM IR. There's a good range of suggested starting points: LoopInfoWrapperPass, ScalarEvolution, InductiveRangeCheckElimination, and Polly.
Wolfgang Pieb is interested in extending the liveness of the 'this' pointer for a better debugging experience. His question is whether there's a better way of doing this than creating a new 'fake use' intrinsic. Kevin Smith suggests always storing the 'this' pointer in memory, but Wolfgang is concerned about the overhead of this approach.
LLVM commits
The AArch64 machine reassociation code has been refactored to be target-independent. r248164.
LLVM's SafeStack now supports Android. r248405.
A new target hook has been added for optimizing register copies. r248478.
Operand bundles are now supported for CallInst and InvokeInst. Initial support was also added to LLVM bitcode. r248527, r248551.
Clang commits
The iOS/OSX localizability checks have been improved. r248350.
Some more PS4 toolchain code landed. r248546.