GCC 8/9 Land Fix For "-march=native" Tuning On Modern Intel CPUs

[phoronix]

Phoronix: GCC 8/9 Land Fix For "-march=native" Tuning On Modern Intel CPUs

The other day we reported on a GCC 8 regression where Skylake and newer CPUs with "-march=native" haven't been performance as optimally as they should be. Fortunately, that patch was quickly landed into the GCC SVN/Git code for GCC 9 as well as back-ported to GCC 8...

http://www.phoronix.com/scan.php?page=news_item&px=GCC-Lands-March-Native-SKL-Fix

[cl333r]

Is LLVM also having this issue?

[busukxuan]

Wonder if AMD's Ryzen IPC claims (8% slower than Kaby Lake was it?) was based on GCC. But IIRC they said it was about the same as Haswell/Broadwell, rather than an even bigger IPC difference, so I guess their tests didn't have this problem.

[FireBurn]

Is LLVM also having this issue?

No it won't