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: A blast from the past


  >     Thu Apr  4 07:34:17 1991  Richard Kenner  (kenner at 
vlsi1.ultra.nyu.edu)
  > 
  > Wow!  That's over a decade ago.
Yup.  Long long ago.

  > Obviously I can't remember for sure, but my *guess* is that I felt
  > canonicalization would move the SUBREG operand to OP0.  Unless that's 
wrong,
  > I think the only way OP1 could be a SUBREG is if both it and OP0 were.
  > I don't know why this code didn't try to handle that case.
  > 
  > Sorry, it's just too long ago to know more.
OK.  So it seems to me that with the way swap_commutative_operands_p works
these days that we prefer to put the subreg into OP1, in fact the only time
I can see that it would be in OP0 would be if they are both SUBREGs.

With that in mind, I'd like to put in a set of parallel simplifications
which operand on OP1.  

This will expose a latent bug in the combiner, but I want it exposed so that
we'll fix it :-)

jeff


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