LLVM Weekly - #177, May 22nd 2017
Welcome to the one hundred and seventy-seventh 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 [email protected], or @llvmweekly or @asbradbury on Twitter.
News and articles from around the web
The repository for flang, the
Fortran frontend for LLVM is now available. Presumably there will be more
information and announcements shortly. The -Mextend compiler options,
allowing source lines up to 132 characters in length is cute.
A new Eclipse blog post discusses how clangd may be used from Eclipse.
On the mailing lists
-
Keno Fischer kicked off an interesting thread on propagating memory copies.
-
Gábor Horváth has written an RFC on adding API notes to to Clang, reviving a previous discussion.
-
Marc-André Laperle is seeking feedback on adding indexing support to clangd.
-
Kristof Beyls reports that GlobalISel numbers now look better vs the current O0.
-
Evgenii Stepanov shared an RFC on supporting CFI-icall with ThinLTO.
LLVM commits
-
The last of the major pieces of NewGVN has landed. This handles the equivalence between a phi of ops and an op of phis. r303444.
-
64-bit wide vectorisation has been enabled on AArch64. r303116.
-
A new optimisation pass to replace some forms of LEA on x86 has been added. r303183.
-
The PPC backend learned to eliminate more compare instructions. r303500.
Clang commits
-
Clang will now cast the result of alloca to refer to the appropriate address space. r303370.
-
clangd has been refactored into smaller components. r303067.