On Jan 10, 12:58 am, Robert Bradshaw <rober...@math.washington.edu>
wrote:
> On Jan 9, 2009, at 7:46 AM, Ronan Paixão wrote:
>
>
>
>
>
> > Em Qui, 2009-01-08 às 14:41 -0800, mabshoff escreveu:
> > <SNIP>
>
> >> The gcc memory requirements are not consistent, i.e. gcc 3.4.x does
> >> compile the LinBox extension using about 350 MB peak or so. For gcc
> >> 4.x it seems that maybe 750 MB is peak consumption. So I am not sure
> >> what to do, i.e. if we make this dependent on the gcc version used we
> >> should maybe print a big, fat warning, but go on to build.
>
> >> Michael
>
> > WTF?!? 750 MB to compile? I'd consider that ridiculous, but that's  
> > just
> > me (and all the people out there who doesn't have that amount of free
> > RAM around).

g++, especially 4.0.x and later are absolute pigs when compiling
templated C++ code. And it doesn't have to be heavily templated, i.e.
I remember somebody adding a *single* template to CoCoALib's server
and all the sudden it required 400MB to compile with gcc 3.4.x.

Also: When comparing compile speed for g++ 4.x vs. MSVC 2005 or alter
MSVC absolutely trashes g++ in compile speed over a wide range of test
cases while producing faster code on average. The gcc folks have to do
a lot to catch up to that. Something like the heavily templated LinBox
illustrates why g++ *really* needs to be fixed.

> Me too, but I don't know whether to blame that on gcc or the (highly  
> templated) library, or if anything can be done about it. Though sage  
> is entirely useable on machines with much less RAM than that, it's  
> towards the low end of people trying to compile from source.

Well, given that you are compiling about 5 million lines of code
having faster hardware is a big plus. And RAM is dirt cheap theses
days, but that doesn't help you when your box is already maxed out.

> - Robert

Cheers,

Michael
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to sage-devel@googlegroups.com
To unsubscribe from this group, send email to 
sage-devel-unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URLs: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---

Reply via email to