egcs-1.1 status

[Michael Hayes]

Jeffrey A Law writes:

 > Fixing regparm to work means killing SMALL_REGISTER_CLASSES.  Which
 > would be a good thing, but is a huge amount of work.

Kill, kill, kill!  What approach is required to remove this horrid
hack?  A rewrite of reload?  I have to define this for the sake of one
rare insn that requires a restrictive register class that clashes with
incoming function argument registers.