Re: glibc-2.1 upgrade headaches. Any ideas??

CaT (cat@zip.com.au)
Sun, 7 Mar 1999 14:35:48 +1100 (EST)


Arvind Sankar wrote the following:
>
> On Sun, Mar 07, 1999 at 02:09:17PM +1100, CaT wrote:
> > I've been happily doing it for a while with -O2 optimisation and
> > -march=pentiumpro and haven't had a single kernel crash.
> >
> > I'd use -O6 but I can't find any info on what the hell this does
> > as far as optimisation goes.
>
> The biggest you can go to currently is -O3, I believe. Anything beyond that
> turns on the same flags.

Right. That's what I figured. But a lot of people seem to go on about how -O6
should be used for best performace... I've found that somewhat confusing...

> > of -O2 (I compiled X without it and it's working fine. Hasn't crashed) or
> > is it something the kernel needs? And, what exactly does this do? I've read
> > the manpage but what that has doesn't mean much to me. I wish there was
> > info on not just what an optimisation flag does but also how it might effect
> > performance and what sideeffects it might have. (this is probably real hard
> > to do but it would be way useful. I don't mind long compile times as long
> > as runtime performance is imporved)
> >From the gcc manual:
>
> `-fomit-frame-pointer'
> Don't keep the frame pointer in a register for functions that
> don't need one. This avoids the instructions to save, set up and
> restore frame pointers; it also makes an extra register available
> in many functions. *It also makes debugging impossible on some
> machines.*
>
> This option is never turned on by default (on i386), so the kernel Makefile
> turns it on.

Uhuh. Does this prevent the use of gdb on the executable and can it be
used safely on libraries?

> Obviously, it should improve run-time performance, unless the gcc code
> optimizer is so buggy that it generates faster code with less registers
> available ;)

Hehe :)

-- 
CaT (cat@zip.net.au)                       URL: http://www.zip.com.au/dev/null

An electricity provider of New Hampshire, US has advised it's customers that in the event of a power failure they can log on to its website for more information... - Paraphrased from the New Scientist, Feb 6, 1999

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/