Alex Bradbury via llvm-dev
2019-Mar-25 19:22 UTC
[llvm-dev] LLVM Weekly - #273, March 25th 2019
LLVM Weekly - #273, March 25th 2019 ================================== If you prefer, you can read a HTML version of this email at <http://llvmweekly.org/issue/273>. Welcome to the two hundred and seventy-third 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](https://www.linkedin.com/in/alex-bradbury/). Subscribe to future issues at <http://llvmweekly.org> and pass it on to anyone else you think may be interested. Please send any tips or feedback to <asb at asbradbury.org>, or @llvmweekly or @asbradbury on Twitter. ## News and articles from around the web LLVM 8.0.0 [has been released](http://lists.llvm.org/pipermail/llvm-dev/2019-March/131157.html). Amongst a whole host of other changes and improvements, the WebAssembly backend is no longer experimental. ## On the mailing lists * Jeroen Dobbelaere posted an RFC on [full restrict support for LLVM](http://lists.llvm.org/pipermail/llvm-dev/2019-March/131127.html), i.e. support for restrict other than just on function arguments. * Louis Dionne is [seeking feedback](http://lists.llvm.org/pipermail/libcxx-dev/2019-March/000282.html) on declaring a libc++ ABI v2 as stable. Concerns were raised about supporting multiple stable ABIs in parallel, though these weren't shared universally. * Chandler Carruth [responded](http://lists.llvm.org/pipermail/llvm-dev/2019-March/131137.html) to the thread on scalable vector types IR with concerns on whether the benefit outweighs the costs, arguing even if the ISA supports variable widths the common case will be to tune for specific widths anyway. Bruce Holt [responded to this](http://lists.llvm.org/pipermail/llvm-dev/2019-March/131188.html). * Steven Wu [shares an RFC](http://lists.llvm.org/pipermail/llvm-dev/2019-March/131244.html) on upstreaming an Objective-C metadata reader library to LLVM. ## LLVM commits * A new ARCOptAddrMode pass was added to generate postincrement loads/stores. [r356606](https://reviews.llvm.org/rL356606). * The RISC-V backend learned to optimise sequences of multiple SELECTs, which is useful as there is no conditional move instruction. [r356741](https://reviews.llvm.org/rL356741). * Target definitions and MC layer support were added for RISC-V RV32E. [r356744](https://reviews.llvm.org/rL356744). * MVTs were defined for v3i32, v3f32, v5i32, and v5f32. These will be used by AMDGPU. [r356531](https://reviews.llvm.org/rL356531). * AArch64 floating point materialisation was improved. [r356390](https://reviews.llvm.org/rL356390). * A new Remark/RemarkParser abstraction was added. [r356491](https://reviews.llvm.org/rL356491). * AliasAnalysis learned to cache BasicAA/anyAA state. [r356783](https://reviews.llvm.org/rL356783). * Numeric block IDs are now supported in LLVM IR. [r356789](https://reviews.llvm.org/rL356789). ## Clang commits * The creduce-clang-crash.py helper was improved, adding support for preprocessing the file before reducing and for trying to remove some command line arguments. [r356636](https://reviews.llvm.org/rL356636). * A new OpenMP clang-tidy module was added, along with some initial checks. [r356800](https://reviews.llvm.org/rL356800), [r356801](https://reviews.llvm.org/rL356801), [r356802](https://reviews.llvm.org/rL356802). ## Other project commits * `<filesystem>` is now built as part of the libcxx dynamic library. [r356518](https://reviews.llvm.org/rL356518). * Polly gained an "LLVM" OpenMP backend (as opposed to just supporting GNU OpenMP). [r356434](https://reviews.llvm.org/rL356434). * The utils/compare.py script in the LLVM test-suite repo gained a geomean summary row in its output. [r356545](https://reviews.llvm.org/rL356545).