LLVM 3.4 Branched, Christmas Compiler Present Planned

LLVM 3.4 has been branched and is now under a feature freeze. Over the next several weeks there will be extensive testing done of this major update to the open-source compiler stack while an official release is planned to happen right before Christmas.

LLVM 3.4 has many new features and will go head-to-head with GCC 4.9 that will be released a few months later. LLVM 3.4 has many improvements to the AMD GPU back-end that's now enabled by default, the loop vectorizer is now more widely used, there's much better AArch64 (64-bit ARM) support, the SLP Vectorizer is used by default, and a ton of other changes covered by numerous Phoronix articles already -- but of course more coverage will come as the official 3.4 release is on approach.

LLVM 3.4 was branched on Tuesday by Apple's Bill Wendling. After the branching he shared for phase one testing to happen for the next week followed by bug-fixing and a second phase of testing for early December. By mid-December, only fixes for major regressions will be accepted into the LLVM 3.4 branch. The estimated release date on LLVM 3.4 is to ship on Monday, 23 December. That release date has the possibility of slipping given recent LLVM releases and also the developers being challenged around the holidays.

The LLVM 3.4 release details can be found by this LLVMdev message.