[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: floor() broken on EV4?

On Tue, Feb 20, 2007 at 12:31:48AM -0500, prescott@phys.ufl.edu wrote:
> Or pilot error :-(  I finally got a chance to look more into this, 
> and I see that I have gaffed.  I did not have the math-emu module 
> loaded.  Once loaded, floor() (and other stuff) works fine.

> Maybe everyone except me knew to have math-emu loaded?  I was 
> kind of surprised it was not built-in.  Is this intended?  

Hmm, I would guess it was modularized with the idea that it was unnecessary
bloat in the kernel for newer CPUs that don't need it?

I tend to disagree with this, because the cost on systems that don't need it
is very small, whereas the RAM penalty of modularization on systems that
/do/ need math-emu is comparatively high.  Unfortunately, this config
setting has been unchanged since at least July 2005 when the Debian kernel
configs were first centralized, and changing the setting will change the
kernel ABI, so this realization probably comes too late to fix for etch r0.

But, I've confirmed that modprobing math-emu fixes all the other known IEEE
problems on ev56 as well, as described in
<http://lists.debian.org/debian-alpha/2006/06/msg00073.html>.  Thanks for
figuring this out for us :)

Steve Langasek                   Give me a lever long enough and a Free OS
Debian Developer                   to set it on, and I can move the world.
vorlon@debian.org                                   http://www.debian.org/

Reply to: