Benchmarks Of LLVM Clang 6.0 Through Clang 10.0 Compilers
At the end of 2019 I ran some GCC 5 through GCC 10 compiler benchmarks while here are the similar tests conducted on the LLVM side for seeing how the Clang C/C++ compiler performance has evolved over the past few years.
This round of testing was done on LLVM Clang 6.0 through LLVM Clang 9.0 while building in the release modes and also testing with LLVM Clang 10.0 as the development code as of December. All the CFLAGS/CXXFLAGS remained the same throughout testing.
Testing was done on the same Intel Core i7 5960X Haswell-E system that was used in the GCC testing, given the mature Haswell state in these compilers going back to the oldest releases tested.
The performance of Clang 6 through Clang 10 of the generated binaries was a lot less eventful on this Core i7 system than during the GCC testing. Compiler times are noticeably higher as expected due to the additional compiler passes with newer compilers.
Of more than 50 tests, Clang 9 was the fastest followed by Clang 10.
The numbers aren't too exciting overall, but there are a few standouts. You can see all those numbers via this OpenBenchmarking.org result file.