This is the mail archive of the gcc@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]

Re: question regarding regrename and failure of 950704-1.c on main


> It's got to be merely exposing a latent problem.  Dig deeper.  Again,
> why do we have reg:di 22 and reg:di 23.  That's simply wrong.

As noted, DI 23 is one of the original arguments.  The current definition
of HARD_REGNO_MODE_OK forces DI mode to even registers.  Thus, the
register choice in this case is not consistent with our definition
of HARD_REGNO_MODE_OK.

DI 22 should not have been selected because it overlaps DI 23
(HARD_REGNO_NREGS wasn't checked?).  I will try to find this bug.

I think we should revamp HARD_REGNO_MODE_OK to use odd registers
for REGNO < 27, plus registers 28 and 30.  This will allow direct
use of the incoming argument values, and the outging return value.
What do you think?

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]