This is the mail archive of the
libc-help@sourceware.org
mailing list for the glibc project.
Re: Pow very slow on x86_64
- From: philippe marguinaud <pmarguinaud at yahoo dot com>
- To: Carlos O'Donell <carlos at systemhalted dot org>
- Cc: libc-help at sourceware dot org
- Date: Thu, 29 Jan 2009 02:07:27 -0800 (PST)
- Subject: Re: Pow very slow on x86_64
- Reply-to: pmarguinaud at yahoo dot com
Yes, I am sure I am using pow@libm. I have recompiled the libm to have a look at what is going on. For some reason, when the exponent is exactly 0.25 ( 0.24999 or 0.250001 would not make it ), I end up in a function called "slowpow" whose name tells you what it does.
This is pretty annoying, since I have very large arrays I need to **0.25. For the time being, I use sqrt(sqrt()) which as fast as expected.
Philippe
--- On Wed, 1/28/09, Carlos O'Donell <carlos@systemhalted.org> wrote:
> From: Carlos O'Donell <carlos@systemhalted.org>
> Subject: Re: Pow very slow on x86_64
> To: "philippe marguinaud" <pmarguinaud@yahoo.com>
> Cc: libc-help@sourceware.org
> Date: Wednesday, January 28, 2009, 10:59 AM
> On Tue, Jan 27, 2009 at 10:33 AM, philippe marguinaud
> <pmarguinaud@yahoo.com> wrote:
> > What people think ?
>
> Are you sure you are calling glibc's pow function?
> e.g. Does readelf show that the binary needs pow() from
> libm?
>
> Contact your distribution vendor to see if they have any
> comments?
>
> Cheers,
> Carlos.