LLVM Weekly - #161, Jan 30th 2017
Welcome to the one hundred and sixty-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 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
Microsoft have announced a new open source DirectX shader compiler based on Clang and LLVM.
LLVM 4.0.0-rc1 source and binaries are now available. Time to get testing!
On the mailing lists
-
Johannes Doerfert has posted an RFC for adding parallel extensions to LLVM IR.
-
Chas Boyd has written to the mailing list to say a little more about Microsoft's use of LLVM and Clang in the new DirectX Shader Compiler.
-
Manuel Klimek reports that work is starting on ClangD, a daemon to provide access IDEs etc with access to Clang services.
-
Philip Herron is curious about selection of negative addition rather than subtraction. Matt Arsenault explains how this behaviour can be overridden.
-
Guy Blank has posted an RFC on making i1 illegal in X86 codegen.
LLVM commits
-
NewGVN gained support for dead and redundant store elimination. r293258.
-
A new MemorySSA updater has been introduced. r293356.
-
A new guard-based loop predication optimisation has been added. This will enable more guards to be hoisted out of loops. r293064.
-
lit has been expanded to allow boolean expressions in REQUIRES, XFAIL, and UNSUPPORTED lines. r292896.
-
Early if-conversion has been implemented for AMDGPU, but is disabled by default for now. r293016.
-
The llvm-xray graph subcommand now colors edges and vertices based on statistics. r293031.
-
A new OptimizationRemarkEmitter pass was added, allowing MIR passes to emit optimisation remarks just like IR passes can. r293110.
-
The MIPS backend gained support for the static relocation model with the N64 ABI. r293279.
-
A number of new intrinsics for constrained floating point operations have been added. r293226.
Clang commits
-
A new TableGen-based generator for command line argument documentation was added. r292968.
-
Comment reflowing has been implemented for clang-format. r293055.