LLVM/Clang 3.3 Branched With Many New Features
With plans to release in June as a major update to LLVM and dependent projects, all relevant code-bases were branched from trunk this afternoon for LLVM 3.3. In particular, the release is planned for 5 June.
Among the exciting features to be found with LLVM 3.3 include:
- An SLP vectorizer for potential performance improvements of straight-line code.
- Loop vectorizer improvements and it's now enabled by default for the -O3 optimization level.
- The Clang compiler is now C++11 feature-complete and there's early support for C++14 / C++1y.
- Better Intel AVX2 support.
- The AMD R600 Radeon GPU LLVM back-end, with various features and GLSL 1.30 support. The AMD GPU LLVM back-end is needed for RadeonSI (HD 7000 series) support and OpenCL/GPGPU support in general.
- 64-bit ARM / AArch64 compiler support.
- IBM SystemZ support.
- Compiler performance improvements.
And there's still a whole lot more, which will be covered in various articles and benchmark results to be shared in the coming weeks on Phoronix.