GCC 4.6/4.7 vs. LLVM-Clang 3.0/3.1 Compilers

With LLVM/Clang 3.1 due out next week, here's a look at the compiler performance of the GCC 4.6 and 4.7 compilers compared to LLVM-Clang 3.0 and a recent LLVM-Clang 3.1 SVN snapshot.

This testing was done from an Intel Sandy Bridge platform in late April while new Intel Ivy Bridge compiler benchmarks and other results are due out soon. There's also some AMD FX-8150 Bulldozer compiler benchmarks and some other earlier Clang 3.1 results.

Aside from GCC 4.6.3, GCC 4.7.0, LLVM-Clang 3.0, and LLVM-Clang 3.1 SVN, AMD's Open64 5.0 x86_64 compiler was tossed into the mix. However, for the Intel Sandy Bridge platform and the set of open-source test profiles used, AMD's Open64 flavor only ended up working for a single test profile.
As far as the features of LLVM/Clang 3.1, many of the key items are mentioned in this article. The key highlights of GCC 4.7.0 are listed here. The results begin in this article and are then continued on OpenBenchmarking.org.
See many more results on OpenBenchmarking.org. You can also run phoronix-test-suite benchmark 1204215-SU-LLVMCLANG23 to facilitate an automated side-by-side compiler comparison on your system's software/hardware via the Phoronix Test Suite.