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]
Other format: [Raw text]

[Bug target/13186] [PPC] Internal compiler error in reload.c


------- Additional Comments From rsandifo at gcc dot gnu dot org  2003-11-25 14:03 -------
I think this is a bug in the rs6000 backend, which seems to have
a combination of:

   - movdi patterns which provide no constant->FPR alternatives.

   - movdi expanders that do nothing with constant->FPR moves,
     at least not if !TARGET_POWERPC64.

   - a definition of secondary_reload_class which says that constants
     can be moved directly into FPRs:

       /* Constants, memory, and FP registers can go into FP registers.  */
       if ((regno == -1 || FP_REGNO_P (regno))
           && (class == FLOAT_REGS || class == NON_SPECIAL_REGS))
         return NO_REGS;

I'm not sure what the intended behaviour is, but the attached patch
seems to get past the initial failure.  Unfortunately, there's not much
chance of it being right.  I guess this should really be handled in the
move expanders somehow.

Richard


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2003-11-25 14:03:34
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13186


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