This is the mail archive of the
mailing list for the GCC project.
Re: Patch to fix gcc.c-torture/compile/20010102-1.c on IA64 HP-UX
On Thu, Oct 30, 2008 at 12:38:09PM -0600, Jeff Law wrote:
> Peter Bergner wrote:
>> On Thu, 2008-10-30 at 10:52 -0700, Steve Ellcey wrote:
>> --- config/ia64/predicates.md (revision 140417)
>> +++ config/ia64/predicates.md (working copy)
>> @@ -585,6 +585,6 @@ (define_predicate "ar_pfs_reg_operand"
>> (define_predicate "basereg_operand"
>> (match_operand 0 "register_operand")
>> - return REG_P (op) && REG_POINTER (op);
>> + return REG_P (op) && REG_POINTER (regno_reg_rtx[ORIGINAL_REGNO (op)]);
>> If you recall, that did solve the ICE you were hitting. The rerename.c
>> change was just handling a case that looked like should be handled, but
>> wasn't necessary for fixing your IC
> We shouldn't be hacking up backends to deal with this problem, it's
> entirely the wrong approach.
Why? REG_POINTER can't (reliably) work on hard registers, because through
the life of one function one hard register might contain in some places
a pointer and in others non-pointer. While pseudos should be always either
pointer or non-pointer. IMHO (almost) all uses of REG_POINTER should
be changed to work on ORIGINAL_REGNO.