On Mon, Feb 26, 2018 at 04:47:31PM +0000, Tomasz Kłoczko wrote: > On 26 February 2018 at 07:11, Zbigniew Jędrzejewski-Szmek <zbyszek@xxxxxxxxx > > wrote: > > > On Mon, Feb 26, 2018 at 05:24:28AM +0100, Kevin Kofler wrote: > > > Zbigniew Jędrzejewski-Szmek wrote: > > > > But size is easy: > > > > $ du -s /var/tmp/inst-O* > > > > 74428 /var/tmp/inst-O3 > > > > 56136 /var/tmp/inst-O3-lto > > > > 61992 /var/tmp/inst-Os > > > > 43572 /var/tmp/inst-Os-lto > > > > > > The current default is -O2, not -O3, though. > > > > Heh, I'm a bit surprised: > > > > 74428 /var/tmp/inst-O3 > > 56136 /var/tmp/inst-O3-lto > > 61992 /var/tmp/inst-Os > > 43572 /var/tmp/inst-Os-lto > > 71268 /var/tmp/inst-O2 > > 50276 /var/tmp/inst-O2-lto > > 47928 /var/tmp/inst-O0 > > 32688 /var/tmp/inst-O0-lto > > > > So... I'd say that lto *does* give a measure benefit. > > (Note, it's all unstripped.) > > Looking only on total file size is not proper method checking code size. > Especially such method may be totally misleading if binaris are not > stripped. > Can you repeat above and show output of the command "size /var/tmp/inst-*"? After stripping: 24808 /var/tmp/inst-O2 21060 /var/tmp/inst-O2-lto 22240 /var/tmp/inst-O3 21812 /var/tmp/inst-O3-lto 23256 /var/tmp/inst-Os 19980 /var/tmp/inst-Os-lto (24808-21060)/24808 ≈ 15%. Zbyszek _______________________________________________ devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx
