With LLVM/Clang 3.2 being released next month and the code branching occurring this month, here's some new benchmarks from the latest SVN development snapshot as of this weekend. LLVM/Clang 3.2 SVN benchmarks were compared to the earlier LLVM 3.1 and 3.0 releases for reference.
Among the features for LLVM/Clang 3.2 is an automatic loop vectorizer, LLVM's Polly has seen improvements, the AMD GPU back-end, and many other compiler infrastructure changes that will be discussed in future Phoronix articles as the official 3.2 release is on final approach.
There's already been a few LLVM/Clang 3.2 benchmarks of the development state from last month on a different Linux x86 system while the testing today is a 3.0/3.1/3.2 comparison from an Intel Core i7 720QM "Clarksfield" processor running from a Lenovo ThinkPad W510 notebook. On the operating system side was Ubuntu 12.10 x86_64 with the Linux 3.5 code. LLVM 3.0/3.1/3.2 were all built from source in a release/optimized mode (using the "--disable-assertions --enable-optimized" configure options).