This is the mail archive of the gcc-bugs@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]

Re: RFC: Jump to const_int


> On Fri, Jan 12, 2001 at 01:43:22PM -0500, John David Anglin wrote:
> > (jump_insn 8 3 9 (set (pc)
> >         (const_int -1161904448 [0xbabebec0])) -1 (nil)
> >     (nil))
> 
> Indeed this is a problem.  One solution is to modify computed_jump_p
> so that it recognizes this.  I'll probably do that.  But another is
> to change the predicate to match the constraints on indirect_jump so
> that this value is pulled into a register right from the start.
> 
> 
> r~
> 
> 
>         * config/vax/vax.md (indirect_jump): Use register_operand.

I already was doing a test bootstrap using register_operand.  I can confirm
that this fixes the testsuite failure.  However, now that you have fixed
computed_jump_p, I think that address_operand may be a slightly better
choice for indirect_jump.  

I see that you have also fixed the insns that didn't have contiguous operands
and I think the sets that produced various non lvalue warnings.  Regarding
situations with more than 255 arguments for a call, it looks like this
will now cause an abort.  Was that the intent?

Thank's for the fixes.

Dave
-- 
J. David Anglin                                  dave.anglin@nrc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)

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