LLVM/Clang 3.3 Delivers Speed Improvements

The C-Ray multi-threading performance was the same during this round of benchmarking from the Intel Core i7 system running Ubuntu 13.04.

LLVM/Clang 3.3 SVN was also much faster when handling the Smallpt path-tracing global illumination renderer.

Overall, the LLVM/Clang 3.3 performance is shaping up quite nicely at least from our multiple Intel Linux systems tested thus far. In a majority of the C/C++ test-cases, LLVM/Clang 3.3 is producing faster binaries than LLVM/Clang 3.2. Part of why LLVM/Clang 3.3 is faster is due to LLVM's loop vectorizer being enabled by default for -O3. There's also been various other improvements.

Benchmarks from AMD hardware as well as ARM platforms are forthcoming on Phoronix. The official LLVM/Clang 3.3 release is expected in June.