[gentoo-dev] Proposition for stage3.

[gentoo-dev] Proposition for stage3.

[Dan Naumov]

Hello.

I just thought I'd share an idea I had about stage3. I think it would be 
nice if it was usable for systems older then i686 and if it used -O1 or 
-O2 instead of -O3. The reasons are simple:

The majority of the people who actually want to use stage3 do so because 
of the time it takes to build the ENTIRE system from scratch. This 
mainly applies to old PC's, which are obviously older then i686. I've 
seen quite a few people on #gentoo complain that it is a major pain in 
the ass to install Gentoo on an old system.

As for -O1 / -O2 instead of the default -O3, it's just that some people 
want to be "safe". I've had all kinds of weirnesses and unexplainable 
crashes due to software compiled with -O3 and no other optimisations. 
These problems always went away after a recompile with -O2. And I am 
pretty sure I am not the only one who thinks that way, because many 
people I know got "burned" when utilising heavy opts.

Sincerely,
-- 


#-------------------------------------------------------#
|  Dan Naumov (aka Jago on IRC) || Otavantie 2, 50670,  |
|     Otava Folk Highschool     ||   Otava, Finland.    |
#-------------------------------------------------------#
|      User of Gentoo Linux : http://www.gentoo.org     |
#-------------------------------------------------------#

Re: [gentoo-dev] Proposition for stage3.

[Bart Verwilst]

On Tuesday 18 June 2002 17:52, Dan Naumov wrote:
|| Hello.

Hi!

|| I just thought I'd share an idea I had about stage3. I think it would be
|| nice if it was usable for systems older then i686 and if it used -O1 or
|| -O2 instead of -O3. The reasons are simple:
||
|| The majority of the people who actually want to use stage3 do so because
|| of the time it takes to build the ENTIRE system from scratch. This
|| mainly applies to old PC's, which are obviously older then i686. I've
|| seen quite a few people on #gentoo complain that it is a major pain in
|| the ass to install Gentoo on an old system.

Gentoo 1.3a is already available for both i586 and i686, and i think 1.3b will 
be available for even more archs.. not 100% sure which ones though.. Maybe 
i486, for the very low-end gateways out there :o) Others are k6, athlon, ...

|| As for -O1 / -O2 instead of the default -O3, it's just that some people
|| want to be "safe". I've had all kinds of weirnesses and unexplainable
|| crashes due to software compiled with -O3 and no other optimisations.
|| These problems always went away after a recompile with -O2. And I am
|| pretty sure I am not the only one who thinks that way, because many
|| people I know got "burned" when utilising heavy opts.

Well, i maintain the gentoo 1.3 series of tarballs, and i think -O3 is what 
most of the people who want to download the tarballs want. -O3 is stable for 
the base system as far as i know, and the majority wants -O3.. I think if you 
want other flags than the ones in the tarball, just use stage2, and recompile 
it yourself.. At least that's my opinion on things.. IMHO -O3 isn't all that 
crazy for general use.

-- 
Bart Verwilst
Gentoo Linux Developer, Desktop Team
Gent, Belgium

Re: [gentoo-dev] Proposition for stage3.

[Per Wigren]

-fomit-frame-pointer often gives a noticable speedup also...



Tuesday 18 June 2002 18.22 skrev Bart Verwilst:
> On Tuesday 18 June 2002 17:52, Dan Naumov wrote:
> || Hello.
>
> Hi!
>
> || I just thought I'd share an idea I had about stage3. I think it would be
> || nice if it was usable for systems older then i686 and if it used -O1 or
> || -O2 instead of -O3. The reasons are simple:
> ||
> || The majority of the people who actually want to use stage3 do so because
> || of the time it takes to build the ENTIRE system from scratch. This
> || mainly applies to old PC's, which are obviously older then i686. I've
> || seen quite a few people on #gentoo complain that it is a major pain in
> || the ass to install Gentoo on an old system.
>
> Gentoo 1.3a is already available for both i586 and i686, and i think 1.3b
> will be available for even more archs.. not 100% sure which ones though..
> Maybe i486, for the very low-end gateways out there :o) Others are k6,
> athlon, ...
>
> || As for -O1 / -O2 instead of the default -O3, it's just that some people
> || want to be "safe". I've had all kinds of weirnesses and unexplainable
> || crashes due to software compiled with -O3 and no other optimisations.
> || These problems always went away after a recompile with -O2. And I am
> || pretty sure I am not the only one who thinks that way, because many
> || people I know got "burned" when utilising heavy opts.
>
> Well, i maintain the gentoo 1.3 series of tarballs, and i think -O3 is what
> most of the people who want to download the tarballs want. -O3 is stable
> for the base system as far as i know, and the majority wants -O3.. I think
> if you want other flags than the ones in the tarball, just use stage2, and
> recompile it yourself.. At least that's my opinion on things.. IMHO -O3
> isn't all that crazy for general use.

Re: [gentoo-dev] Proposition for stage3.

[Rufiao]

I haven't build a whole system using that yet. Any packages break 
specifically because of this flag?

Per Wigren [Tue 18-Jun-2002 18:41 GMT+0200]:
> -fomit-frame-pointer often gives a noticable speedup also...

Re: [gentoo-dev] Proposition for stage3.

[Per Wigren]

Not that I have found, and my whole system is compiled with it.. (gcc 2.95.3)


Tuesday 18 June 2002 20.48 skrev Rufiao:
> I haven't build a whole system using that yet. Any packages break
> specifically because of this flag?
>
> Per Wigren [Tue 18-Jun-2002 18:41 GMT+0200]:
> > -fomit-frame-pointer often gives a noticable speedup also...
>
> _______________________________________________
> gentoo-dev mailing list
> gentoo-dev@gentoo.org
> http://lists.gentoo.org/mailman/listinfo/gentoo-dev

Re: [gentoo-dev] Proposition for stage3.

[Prashanth Aditya Susarla]

If you use gcc-2.95 then this doesn't break anything. With gcc-3.x 
however, glibc and X break with -fomit-frame-pointer. These are the two 
major packages which need to be built without this flag. About others, I 
don't know as I haven't come across any yet. In fact, I've used my regular 
flags (-march=athlon-tbird -mmnx -m3dnow -O3 -pipe -fomit-frame-pointer) 
on qt also (by editing the correspoding qmake.conf) with gcc-3.1 and there 
are no problems.

Regards,
Prashanth Aditya Susarla

On Tue, 18 Jun 2002, Rufiao wrote:

> 
> I haven't build a whole system using that yet. Any packages break 
> specifically because of this flag?
> 
> Per Wigren [Tue 18-Jun-2002 18:41 GMT+0200]:
> > -fomit-frame-pointer often gives a noticable speedup also...
> _______________________________________________
> gentoo-dev mailing list
> gentoo-dev@gentoo.org
> http://lists.gentoo.org/mailman/listinfo/gentoo-dev
> 

Re: [gentoo-dev] Proposition for stage3.

[Spider]

[-- Attachment #1: Type: text/plain, Size: 1569 bytes --]

begin  quote
On Tue, 18 Jun 2002 15:48:00 -0300
Rufiao <rufiao@gmx.net> wrote:

> 
> I haven't build a whole system using that yet. Any packages break 
> specifically because of this flag?
> 
> Per Wigren [Tue 18-Jun-2002 18:41 GMT+0200]:
> > -fomit-frame-pointer often gives a noticable speedup also...



I will personally slay the person to include this in the default system,
as -any- bug found at a higher level in the chain will be rendered
completely undebuggable  if glibc or other library that it links uses
this, The initial speedup is trivial for things such as the basesystem,
and to simply ignore all bugreports from a person who uses it is quite a
valid proposition as far as I'm concerned. 

Why the strong feelings about it? bugreports are quite vital to our QA
and future survival, and to our users if they ever want to try anything
even remotely unstable, or development (face it, quite a few do), and
the amount of trouble this flag generates further down the line is
immense.

Yes, you can leave it as a suggestion if you wish, though it should be
noted that "this will break all debugging and render all bugreports
useless, as well as it may introduce odd crashes which will not be
traceable" , But it should -not- be thought of as an default
alternative. 

Better for defaults would be -fno-exceptions  for C++ and C code, 
although this also breaks code, but more noticable in most cases.

//Spider


--
begin  .signature
This is a .signature virus! Please copy me into your .signature!
See Microsoft KB Article Q265230 for more information.
end

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]