This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: GCC/powerpc AltiVec expectations.


I am not sure if this is true on Mac OS X, but the Mac OS classic ABI allocates registers in numeric order if I remember correctly. Therefor its much easier to use VRSAVE to know from (0 to N) are used and (N+1 to 31) are not. In a SVR4 ABI the allocation is not as straight forwarded and thus the OS has to look at each bit position of VRSAVE to determine what to save and restore.

- kumar

On Tuesday, July 2, 2002, at 12:46 PM, Alex Rosenberg wrote:


on 7/2/02 10:32 AM, Aldy Hernandez at aldyh@redhat.com wrote:

On Tue, Jul 02, 2002 at 09:52:46AM -0700, Alex Rosenberg wrote:
on 7/2/02 7:43 AM, Aldy Hernandez at aldyh@redhat.com wrote:

vrsave generation is on by default in gcc, though since the minority
OS need it, it should be switched to off.
"Minority?"
linux doesn't need it.

i believe macos x doesn't need it.
Mac OS X requires it.


on 7/2/02 10:36 AM, Matt Thomas at matt@3am-software.com wrote:
NetBSD won't need it either.  I assume OpenBSD/FreeBSD will just reuse
the NetBSD code so they probably won't either.
I thought Matthew Green once mentioned VRSAVE support for NetBSD to me, but
I can't find a reference to that from the port-powerpc@netbsd.org list.

+------------------------------------------------------------+
| Alexander M. Rosenberg <mailto:alexr@_spies.com> |
| Nobody cares what I say. Remove the underscore to mail me. |






Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]