GCC 4.8.0 vs. LLVM Clang 3.3 Compiler Performance Three-Way

The Core i7 3960X Extreme Edition results for HMMer weren't too interesting, but for the AMD FX-8350 and Core i3 3217U systems they were much faster with LLVM Clang 3.3 over the stable 3.2 release. The LLVM Clang 3.3 performance improvements were enough to make Clang faster than GCC 4.8 on the FX-8350 CPU and competitive with GCC on the Core i3 Ivy Bridge Ultrabook.

The findings with MAFFT were similar to that of the HMMer results where the AMD FX-8350 and Core i3 systems saw nice performance improvements when upgrading LLVM/Clang. Partially responsible for the performance improvements is the LLVM Loop Vectorizer being enabled by default for the -O3 optimization level. LLVM 3.3 also introduces the LLVM SLP Vectorizer, but that isn't currently enabled by default.

GCC 4.8 still has a slight edge over Clang with the BLAKE2 test for all three AMD/Intel CPUs.

GCC 4.8 has the heavy advantage for GraphicsMagick since LLVM 3.3 still lacks support for OpenMP.