Re: pegasosII crash at boot with -current

Matt Thomas <[email protected]> Wed, 1 Dec 2010 13:17:17 -0800
Newsgroups gmane.os.netbsd.ports.powerpc,gmane.os.netbsd.ports.ofppc
Message-ID <[email protected]>
On Dec 1, 2010, at 12:45 PM, Michael wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Hello,
> 
> On Dec 1, 2010, at 3:35 PM, Allen Briggs wrote:
> 
>> On Wed, Dec 01, 2010 at 09:29:03PM +0100, Frank Wille wrote:
>>> But our G4s should be able to execute AltiVec code (I have a 7447A) - maybe
>>> something is not correctly set up at this early point in main()?
>> 
>> Should the kernel be able to execute AltiVec code?  Or should it be
>> treated like the FPU?
> 
> I dimly remember running into the same problem years ago when gcc3 ( or was it 4 ? ) started to emit altivec instructions if the CPU flags allowed it to. The solution was to build all kernels with -mno-altivec when using -mcpu=74xx.
> IIRC the consensus back then was to treat altivec like the FPU and don't use it in the kernel.

You can use it, but only if treated carefully (save/restore explicitly, etc.)
Some of the pmap routines use altivec for copies.