LLVM Clang 6.0 vs. 5.0 Compiler Performance On Intel/AMD Linux

With the exception of the dual Xeon Gold setup, the other systems were seeing improvements with the AVX-heavy Himeno pressure solver.

But with more optimizations and features added to LLVM and Clang, the compiler time with the 6.0 release does appear measurably slower across the board than with 5.0.

The C-Ray multi-threaded ray-tracer also appears slowed down by LLVM 6.0.

In some other tests, the Core i7 8700K was back to showing some minor performance improvements with LLVM Clang 6.0.

At the end of the day though, we've only been seeing minor performance changes of LLVM Clang 5.0 vs. 6.0 on popular Intel and AMD platforms with Linux x86_64. With LLVM/Clang having matured a lot the past several years and now nearly neck-and-neck with GCC in many workloads, long ago are the days where there would be substantial performance changes between releases. It may be more interesting in the ARM space and I do have some fresh compiler benchmarks on my TODO list there as well as a fresh look at Clang 6.0 vs. GCC 8.1 when that GNU compiler release happens in the next couple of weeks.