This is the mail archive of the gcc-patches@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: [patch] fix bitfield constraints for m68k


On Wed, Mar 10, 2004 at 05:26:09PM -0500, Kazu Hirata wrote:
> Hi Richard,
> 
> + (match_operand:SI 2 "general_operand" "dn")
> 
> In this kind of case, should we let reload handle a non-compile-time
> constant?  Or should we tighten the predicate to not accept a
> non-compile-time constant?  Or is it up to one's taste?

depends on what alternative instruction patterns are available.
In this case it would be some shift and mask which would require
the same kind of reload as the bit field insn plus more instructions,
clearly a bad tradeoff.

In other cases things may be different. It might even be possible
to improve binutils and dynamic loader to handle this special case.

Richard


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