On Feb 24, 6:53 am, water <waterloo2...@gmail.com> wrote:
Hi,
> I change to :
> CFLAGS="-O0 -march=pentium-m -pipe"
> CXXFLAGS="-O0 -march=pentium-m -pipe"
No, that is not what I wrote. *Unset* them completely since clisp is
miscompiled by gcc 4.x if you look at it from the wrong angle. I am
surprised you even get the CFLAGS in since we overwrite them
explicitly, but its build system might pick up the CXXFLAGS - I have
no idea why that would be.
In general: I seriously doubt you will see any statistical relevant
improvements from "-march=pentium-m" for example since the truly speed
critical components in Sage have assembly kernels or set CFLAGS to
*proven* settings that work. That "let's compile things with maximum
optimization" crowd that represents some of the Gentoo community (I
used to run a couple servers on Gentoo and know the system well) does
often not understand that code compiled with more optimizations will
also miscompile more often that not. This is not based on conjecture
but by various other cases in this group where people did the same
thing and ended up with broken doctests in Sage due to miscompilations
if they even got Sage to compile at all. Currently Sage given you card
blanche to do what ever you want in regard to CFLAGS & friends, but
once I get to revamp the build system this March (see ticket
#2999-3001) I will make it so that you cannot overwrite the CFLAGS &
friends for Sage unless you have carefully read the documentation and
*when* it fails it will print a message that clearly states that you
played with CFLAGS & friends. And in that case my canonical answer
will be "You picked expert mode and you broken it, either do not do
that if you don't know what you are doing or fix it yourself". No hard
feelings, but after the 5th report of someone shooting themselves in
the foot I just have better things to do :)
> fail to compile clisp 2.46 too.
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
-~----------~----~----~----~------~----~------~--~---
