Re: First cut of PowerPC support in NPTL

[Jakub Jelinek <jakub redhat com>]

On Tue, Mar 11, 2003 at 05:19:25PM +1100, Ian Wienand wrote:
> On Sat, Mar 08, 2003 at 04:01:04AM -0500, Jakub Jelinek wrote:
> > --- libc/nptl/allocatestack.c.jj	2003-03-05 14:43:34.000000000 -0500
> > +++ libc/nptl/allocatestack.c	2003-03-07 18:35:07.000000000 -0500
> 
> > +#ifdef NEED_SEPARATE_REGISTER_STACK
> > +	  char *guard = mem + (((size - guardsize) / 2) & ~pagesize_m1);
> > +#else
> > +	  char *guard = mem;
> > +#endif
> 
> just out of curiosity, is this split between the register stack and
> the user stack arbitrary or are there reasons why it's strictly down
> the middle?

The split is arbitrary. pthread_attr_* API doesn't have functions to
request specific sizes of IA-64 register stack and normal stack separately.

> seeing as ia64 shares the stack, should the MINIMAL_REST_STACK for
> ia64 be maybe double the size (or even bigger)?

It is bigger in more recent versions of the patch.

	Jakub