| 1 |
El mié, 04-02-2009 a las 14:31 +0100, Momesso Andrea escribió: |
| 2 |
> On Wed, Feb 04, 2009 at 08:45:50AM -0430, Sebastián Magrí wrote: |
| 3 |
> [snip] |
| 4 |
> > > |
| 5 |
> > > Often on gentoo related IRC chanels comes someone who asks why his |
| 6 |
> > > firefox-bin (or openoffice-bin or *-bin) runs faster than his |
| 7 |
> > > built-from-source firefox. |
| 8 |
> > > |
| 9 |
> > > Usually chan's gurus answer that upstream packagers use all the possible |
| 10 |
> > > compiler optimizations (CFLAGS LDFLAGS etc.) for the given package, |
| 11 |
> > > while the average gentoo users keeps a set of "system wide very safe |
| 12 |
> > > optimizations" that are good for most packages, but not the best for |
| 13 |
> > > every particolar package. |
| 14 |
> > > |
| 15 |
> > > Is that statement correct? |
| 16 |
> > > |
| 17 |
> > > ======= |
| 18 |
> > > TopperH |
| 19 |
> > > ======= |
| 20 |
> > |
| 21 |
> > I've always felt the compiled openoffice faster than the binary one, but |
| 22 |
> > if it is not the case portage also gives you the chance of establishing |
| 23 |
> > per-package optimisations on '/etc/portage/env/' or in the paludis |
| 24 |
> > bashrc, so if one user wants an particular app to go faster, he can |
| 25 |
> > research about the best way to build this one. This way, the user can |
| 26 |
> > keep the very safe optimisations for the rest of the system and some |
| 27 |
> > -unsafe optimisations- for the packages he want. |
| 28 |
> > |
| 29 |
> > It is more about choices... |
| 30 |
> |
| 31 |
> Sure, I've used per-package optimizations myself in some particular |
| 32 |
> cases, but that's not the point. |
| 33 |
> |
| 34 |
> A package manteiner *should* know better than an average user which |
| 35 |
> optimizations will tune better their own package. |
| 36 |
> |
| 37 |
> My question can be put like this: Do binary distro's per package |
| 38 |
> optimiziations override the benefit of having arch specific |
| 39 |
> optimiziations that gentoo allows? |
| 40 |
> |
| 41 |
> |
| 42 |
> ======= |
| 43 |
> TopperH |
| 44 |
> ======= |
| 45 |
|
| 46 |
It does, but I am almost sure that most of the binary distro's package |
| 47 |
maintainers can't ship a package with hard optimisations because it will |
| 48 |
possibly work fine on his box but not in the user's box. There is where |
| 49 |
we heard histories about binary distros users compiling their apps to |
| 50 |
improve it's performance, possibly breaking their system at the same |
| 51 |
time. |
| 52 |
|
| 53 |
Gentoo maintainers *should* also know better than the users which |
| 54 |
optimisations can be given to the user for a package to build and work |
| 55 |
fine... Other case is when it represents a risk of having unstable apps, |
| 56 |
in that case dropping optimisations is necessary in order to have more |
| 57 |
stable apps. |
Archives