LLVM Weekly - #215, Feb 12th 2018

Welcome to the two hundred and fifteenth 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 https://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.

I became a father last week, and Archie has been keeping my wife and I busy ever since. More so than usual, please do drop me an email if there's something you think should definitely be featured in LLVM Weekly.

News and articles from around the web

LLVM 6.0.0-rc2 has been tagged.

Slides and recordings from the LLVM Devroom at FOSDEM 2018 are now available.

On the mailing lists

• Tanya Lattner is looking for project proposals for Outreachy internships. "LLVM is participating in the Outreachy May 2018 to August 2018 internships. We are seeking mentors to propose projects that Outreachy interns can work on during their internship. For our first time with this program, the LLVM Foundation will be sponsoring 1 intern. Ideally we would like a few project proposals for applicants to pick from. These can range from entry level documentation projects to more technical."

• Davide Italiano started an RFC discussion on improvements to the LLDB test suite and future directions. He proposes making aspects of testing more like LLVM, sticking to LLVM policies regarding reverting commits that break the buildbots, and removing support for unmaintained languages (Java/Go/Swift).

• Quentin Colombet started an RFC thread on bumping the bitcode version in LLVM 6.0, due to the changes in fast math flags in bitcode.

• Florian Hahn proposes making LoopVectorize aware of superword-level parallelism (SLP) operations.

• David Li has written a short summary of developments in profile-guided optimisation with LLVM over the past 18 months.

• Michael Haidl started a discussion on upstreaming PACXX. "he main goal for my research was to develop a single-source programming model equal to CUDA or SYCL for accelerators supported by LLVM (e.g., Nvidia GPUs). PACXX uses Clang as front-end for code generation and comes with a runtime library (PACXX-RT) to execute kernels on the available hardware."

LLVM commits

• LLVM now supports representing discriminated unions in DWARF, as required by Rust. r324426.

• Code generation for Armv8.2-A FP16 has been added. r324321.

• The MergeICmps pass is now enabled by default. r324465.

• utils/update_llc_test_checks.py and utils/update_test_checks.py have been updated to share more code. r324803.

Clang commits

• The trivial_abi attribute has been added. This indicates the annotated C++ class/struct/union to be passed or returned using the C ABI for the underlying type. r324269.

• The flags -fdiscard-value-names and -fno-discard-value-names were added to allow user control over whether names should be discarded in IR output. r324498.

• A TemplateInstantiationObserver class was added. This is used in order to implement the TempLight C++ template debugging tool. r324808.

Other project commits

• LLD now has a man page. r324512.

• The sanitizer allocator has seen a variety of local cache improvements. r324257.