ppc patch

Jeffrey A Law law@cygnus.com
Mon Aug 24 21:54:00 GMT 1998


  In message < 9808250410.AA39962@marc.watson.ibm.com >you write:
  > >>>>> Jeffrey A Law writes:
  > 
  > Mike> Jeff, the patch came in just before I left for the weekend, and am I 
  > just
  > Mike> looking at it now.  I believe the patch doesn't go far enough, since 
  > the
  > Mike> splitting code doesn't know about pre_dec addresses, and it leaves th
  > e
  > Mike> register pointing to the wrong word.  I'll fix it.
  > 
  > Jeff> David took care of the problems today.  I'll suck in his patches
  > Jeff> shortly.
  > 
  > 	I didn't quite follow Mike's comments about splitting and pre_dec
  > addresses with respect to the patterns that changed.  There are more than
  > enough subtle points in the soft-float support that more changes are
  > needed to make this addressing mode and constraint safe.
Looking more closely at the port, I don't see a problem with any
splits since it looks like the only time we split movdf/movdi insns
is when loading up a register with a constant which shouldn't trigger
the prolematical cases.


jef



More information about the Gcc-patches mailing list