Subject: Re: AltiVec and *jmp.
To: Matt Thomas <>
From: Jason R Thorpe <>
List: tech-userlevel
Date: 07/02/2002 11:36:20
On Tue, Jul 02, 2002 at 02:31:17PM -0400, Allen Briggs wrote:

 > My concern here was whether or not it's likely to ever be a binary
 > compatibility snag.  After actually waking up today, I guess it
 > wouldn't be moreso than, say, compiling with -m7400 anyway.

-m7400 doesn't change the size of data structures, does it?

 > It seems reasonable to me to do the #define for setjmp() and to
 > not worry about restoring for signals.  Pthreads, however, should
 > probably be saving/restoring.

The #define won't work -- __setjmp14 isn't the name that programs
use.  You'd need to use __RENAME().

For pthreads, etc. you need to make sure that the ucontext_t stuff on
the nathanw_sa branch has support for AltiVec.  Best to do it now before
the stuff is brought down from the branch.

        -- Jason R. Thorpe <>