My home server was also built with 'native' and it's unixbench score was: 231.3 (24.6 min)
# /bin/sh -c "gcc -v -x c -E -mtune=native /dev/null -o /dev/null 2>&1 | grep mtune | sed -e 's/.*mtune=//'" athlon-4 Here it didn't return 'generic' Just out of curiosity, I've removed 'native' and left only 'CPUTYPE?=athlon-xp' After rebuild of OS, it's unixbench score is now: 319.8 (22.4 min) That is a 38.26%, global speed gain! Now I'm 100% sure, 'native' should always be avoided! > It is acceptable to set CPUTYPE and to append certain flags to CFLAGS > in make.conf, and CPUTYPE is used in many ports for > architecture-dependent build options. Default CFLAGS are are '-O2 -pipe' Which are recommended and proven to be safe? > The use of CPUTYPE usually has a relatively modest affect on general > benchmarks in the base system, because a lot of code there is > intentionally generic, and is compiled with an older toolchain using > conservative flags, so aggressive optimizations that can take fuller > advantage of CPU extensions are not used. I believe because OS stability, is a top priority. > However, CPUTYPE can have a dramatic affect on certain applications > when used with appropriate compiler flags (-O3, etc.), particularly > those in some ports. So, CPUTYPE is mainly for boosting performance of ported apps. Regarding compiler flags (-O3, etc.) I use them, only if port offers them. > > b. > Domagoj Smolčić _______________________________________________ freebsd-hackers@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to "freebsd-hackers-unsubscr...@freebsd.org"
