This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: continuing sparcv9 odyssey---need gnu as?
On Thu, Jan 17, 2002 at 12:17:45PM -0500, Brad Lucier wrote:
> .LL11461:
> sethi %hh(___G__23__23_raise_2d_type_2d_exception), %o0
> or %o0, %hm(___G__23__23_raise_2d_type_2d_exception), %o0
> st %o1, [%i2+152]
> sllx %o0, 32, %o0
> sethi %lm(___G__23__23_raise_2d_type_2d_exception), %o1
> add %o0, %o1, %o0
> ba,pt %xcc, .LL11137
> ldx [%o0+%lo(___G__23__23_raise_2d_type_2d_exception)], %o0
> .LL81:
> ld [%i2+152], %o0
> cmp %o0, 1
> bne,pt %icc, .LL10421
> add %i5, 3680, %o0
> .LL2736:
> and %i3, 3, %o0
> .LL11465:
> brz,a,pt %o0, .LL2468 <<<<<<<<< 38609
> mov -6, %i3
> cmp %o0, 1
> bne,pt %xcc, .LL11766
> and %i3, 3, %o0
> ldx [%i3-1], %o0
> and %o0, 248, %o0
> cmp %o0, 240
> be,pn %xcc, .LL11289
> and %i3, 3, %o0
> .LL11766:
> cmp %o0, 1
> bne,a,pt %xcc, .LL2468 <<<<<<<<<<<<<<<<
> mov -2, %i3
> ldx [%i3-1], %o0
> and %o0, 248, %i3
> srax %i3, 1, %i3
> cmp %i3, 124
> be,pn %xcc, .LL11289
> cmp %i3, 8
> be,pn %xcc, .LL11289
> cmp %i3, 12
> ba,pt %xcc, .LL11369
> nop
>
> All the instructions as complains about are either brz,a,pt or brnz,a,pt.
> I find the second line I marked curious, because it has the same target,
> yet as doesn't complain about it.
How far is .LL2468 from that location?
V9 predicted branches are limited to +- 256KB.
Is the function in question bigger?
Jakub