LLVM Weekly - #240, August 6th 2018

Welcome to the two hundred and fortieth 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

LLVM 7.0.0 rc1 has been tagged. Release blockers are tracked here.

Rust 1.28 was released.

On the mailing lists

• Amara Emerson has shared an update on GlobalISel design and goals. Compile time performance remains withing 5% of FastISel.

• Greg Bedwell has shared materials from EuroLLVM describing the DExTer tool for tracking debuginfo quality and updated on its status.

• Raphael Isemann shares that nightly LLDB benchmarks and flame graphs are now available.

• Sergey Dmitriev posted an RFC to allow dropping OpenMP offload targets at link time.

• Rodrigo Caetano Rocha is seeking feedback on a new proposed function merging optimisation.

• Paul Robinson announces that preliminary DWARFv5 support is in Clang/LLVM 7.0.

• There has been a large amount of new discussion regarding support for variable-length vectors. Bruce Hoult shares his expectation of how RISC-V vector support will end up while Robin Kruppe clarifies that there are really two notions of vector length throughout these discussions.

LLVM commits

• Profile loading functionality for XRay was implemented. r338825.

• Pseudo-iterator adaptors implementing 'early increment' style loops were added to LLVM's ADT library. r338955.

• The MachineOutliner for AArch64 learned to save LR to a register rather than the stack. r338278.

• llvm-mca documentation was extended to describe the instruction flow. r338319.

• Rudimentary C++11 support was added to the MS demangler. r338324.

• A very subtle miscompile due to a bug in EFLAGS copy lowering for X86 was fixed. Chandler suggests maintainers of out-of-tree branches using the X86 backend may want to cherry-pick this fix. r338481.

• The Hexagon backend will now simplify the control-flow graph after atomic expansion. r338774.

• Assembler/disassembler support for Arm SVE is now complete.
  r338261.

Clang commits

• Support for a new implicit conversion sanitizer was added. r338288.

• Clang learned to support the [[clang::lifetimebound]] attribute. This indicates the lifetime of a function result is related to one of the function arguments. r338464.

• Clang's ABI lowering for composite types on AArch64 and ARM was adjusted to match GCC and the AAPCS. r338279.

Other project commits

• LLDB now supports syntax highlighting, meaning ANSI color escape sequences are emitted with printed source code. r338662.

• Support for weighted mutation leveraging was aded to libFuzzer, which gives mutations weight based on how much new coverage they provide. r338776

• String conversions for integral types, <charconv>, was added to libcxx. r338479.