| List: | Commits | « Previous MessageNext Message » | |
| From: | Tor Didriksen | Date: | September 17 2010 8:51am |
| Subject: | Re: bzr commit into mysql-next-mr-opt-team branch (tor.didriksen:3215) WL#1393 | ||
| View as plain text | |||
On 2010-09-16 15:00, Konstantin Osipov wrote: > * Tor Didriksen<tor.didriksen@stripped> [10/09/16 16:44]: > >> optimized, from next-mr-bugfixing, configured with cmake (no arguments) >> >> $ll sql/mysqld >> -rwxr-xr-x 1 clustra 44304518 2010-09-16 14:15 sql/mysqld* >> >> >> same branch, but removed the explicit template instantiation >> >> $ll sql/mysqld >> -rwxr-xr-x 1 clustra 44048388 2010-09-16 14:15 sql/mysqld* >> >> so executable size shrinks when you let gcc decide what to do with >> the templates. >> > Nice. I guess we need to get these numbers for a couple more > major platforms and remove the switch. > > we use this for g++ only, and we link with gcc. do we link with the OS native linker on any platform? > On the other hand it's not a complete evaluation: without explicit > template instantiation there is a tendency for uncontrolled growth > of the use of templates, so it would be nice to read up on the > modern compiler/linker's ability to eliminate duplicate code from the > resulting executable. > > well, we cannot use stl anyways, since we link with gcc rather than g++ so any 'uncotrolled growth' must come from our own home-grown templates. -- didrik
