LLVM Weekly - #176, May 15th 2017

Welcome to the one hundred and seventy-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. 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.

News and articles from around the web

Version 0.6 of the Julia programming language is nearing release (currently at rc1), and has many changes and new features.

The LLVM Foundation recently celebrated its 3rd birthday.

On the mailing lists

• It's getting close to the point where GlobalISel is enabled for -O0 on AARch64. Kristof Beyls has updated his benchmark results.

• Daniel Neilson has written a new RFC on element-atomic memory intrinsics.

• Jorge Gorbe Moya has written an RFC on adding new outputs to the clang -save-temps option, specifically generating both unoptimized and optimized textual LLVM IR.

• Sjoerd Meijer has written an RFC on supporting _Float16 in Clang for ARMv8.2-A.

• Hiroshi Inoue has written an RFC on improving scalar replacement of aggregates for method arguments.

• Andrew Kaylor is seeking advice and feedback on his efforts to support the FENV_ACCESS pragma.

LLVM commits

• The parallel algorithm support code introduced in LLD has now been moved to LLVM. r302748.

• A new LiveRangeShrink pass was added to move instructions within a basic block in a way that minimises the live range. r302938.

• A new custom event logging intrinsic was introduced for XRay. r302405.

• A new variant of computeKnownBits has been added, with an easier to use interface. r302437, r302925.

• Experimental intrinsics have been added for horizontal vector reductions. They can be used on AArch64. r302514, r302678.

• An extra operand has been added to the CALLSEQ_START SelectionDAG node to track the stack size. r302527.

• Attributes can now be set on global variables in LLVM IR. r302794.

• A profile-based cost analysis for partial inlining has been implemented. r302967.

Clang commits

• A handy new typo heuristic has been added to give better feedback if a template identifier has been mistyped. r302615.

• A new builtin, __xray_customevent(), has been added to support the XRay custom event logging intrinsic. r302492.

Other project commits

• The placement of orphan sections in LLD has been significantly optimised. r302903.

• LLDB's Error class has been renamed to Status. r302872.