LLVM Clang 12 Performance Benchmarks On AMD Ryzen 9 5950X

Compile times are comparable with Clang 12, even with new optimization passes and other features continuing to be added.

With the AOBench renderer there were some small improvements for the runs involving the -march=native targeting.

But with the vast majority of the tests, there wasn't much change to find out of LLVM Clang 12 with the C/C++ benchmarks run on the Ryzen 9 5950X with its initial Zen 3 support.

Of the benchmarks carried out, dav1d was the main winner of Clang 12 when being built with -march=native settings. At least there haven't been any major regressions found in my Clang 12 benchmarking thus far. It's also not too surprising the lack of major optimizations considering how mature LLVM/Clang 12 is at this point on the x86_64 front and often competing neck-and-neck with GCC -- stay tuned for a very interesting GCC 11 vs. Clang 12 showdown this spring. With the new Znver3 support in LLVM 12 not including any optimized scheduler model or other optimizations beyond the basic enablement work that landed prior to the feature freeze, the lack of uplift isn't that surprising.