Here's my take on this situation, it feels very wrong.
I'm 99% sure there are individual GCC flags which can achieve the same performance targets without resorting to using -O3, for instance -funroll-all-loops.
Using such flags would be safer than enabling -O3 wantonly and we won't end up with bloated binaries.
I'm 99% sure there are individual GCC flags which can achieve the same performance targets without resorting to using -O3, for instance -funroll-all-loops.
Using such flags would be safer than enabling -O3 wantonly and we won't end up with bloated binaries.
Comment