An Early Look At LLVM Clang 13 Performance On AMD Zen 3

Overall, Clang 13 on this AMD EPYC 7003 Milan server was showing some speed improvements out of this forthcoming compiler.

We weren't expecting any big breakthroughs as no "killer features" come to mind for LLVM 13 or the Clang 13 front-end, but seems to be yet another nice evolutionary step forward for this open-source compiler stack that is already competing well with GCC on x86_64 and AArch64.

For those wondering about the build time performance of Clang 13 compared to Clang 12, some benchmarks were run there as well in seeing how the build times compare. Overall the performance was close to the same but compiling FFmpeg and MPlayer both took measurably longer than when built under Clang 12. This though isn't necessarily bad since the addition of new optimization passes and other features in turn can drive up the compile times for yielding faster compiled binaries.

More LLVM Clang 13 benchmarks will come next month around the time of the stable release on more hardware.