Re: m68k cgen file

[Ben Elliston <bje at redhat dot com>]

wcohen wrote:

   I have been looking into developing a m68k simulator that will work
   with gcc, so one doesn't need m68k hardware to test out code. I

Join the club. :-)

   noticed that there is a start for the m68k description in m68k.cpu.
   How difficult would it be to complete this to provide a m68k
   simulator? Are there any known limitations in the current cgen that
   would make this difficult?

As far as I recall, there were two troublemakers:

	* variable length instructions.  At the time I started m68k.cpu,
	  I had little insight into cgen internals.  I don't know how
	  good the support is now--anyone else care to comment?

	* derived operands.  As you probably know, the m68k has a
	  relatively large number of addressing modes.  Ideally, a class
	  of effective address (`ea') operands could/should be used
	  to make this easier--addressing modes are almost orthogonal to
	  the instruction set in the m68k.  More than one of these is
	  required, however, as some instructions don't permit all available
	  addressing modes.

Feel free to continue with what I had, though!

Ben