altivec fix for starting_frame_offset

[David Edelsohn]

>>>>> Richard Henderson writes:

Richard> On Fri, Mar 01, 2002 at 02:28:23PM -0500, David Edelsohn wrote:
>> It is constant on Darwin, so why can't GCC rely on it in this case
>> instead of going through this mess to lie to GCC about the real positions
>> of the stack and arguments?

Richard> We aren't lying.  We're accurately describing them in way that
Richard> reflects the fact that the first argument address is 8 mod 16.

	The patch defines FIRST_PARM_OFFSET as 0.  I am a little worried
about having GCC's concept of the argument pointer not point at the first
word in the argument area.

David