On Jan 23,  1:00pm, Adam Glass wrote:
> > > the sun3 is a 68020 + a proprietary mmu.  i didn't do/copy the
> > > cachectl stuff because i wasn't sure it would be necessary.  I can
> > > make it a temporary no-op though.
> > 
> > Oh, good. One of my 3/60's has a 68030 in it. Though this is not
> > normal, it is still good to leave that code in.
> 
> i can drop the code back in.  the basic problem is that i don't
> understand it.  In particular, which part of it is real, which isn't.
> which doesn't matter on a 68020, which matters on a 68030.  The
> original code has ifdefs by hp model number which don't help me any.

  The only ifdefs I saw in the Amiga file was for the HP370 and dealt
with the external cache.  The cachectl routine only called two routines
to deal with the cache:  DCIU() and ICIA().  I think these routines are
"Data Cache Invalid User" and "Instruction Cache Invalidate All".  The
Amiga routines didn't do anything in DCIU(), but I did a cpusha dc for
the 68040 in all the DCI?() routines.

  The change I made to ld.so to handle the 68040 caches was to call
cachectl with CC_EXTPURGE|CC_IPURGE, since that argument would cause the
data cache to be pushed and the instruction cache invalidated.

Michael

-- 
Michael L. Hitch			INTERNET:  osymh@montana.edu
Computer Consultant			BITNET:  OSYMH@MTSUNIX1.BITNET
Office of Systems and Computing Services
Montana State University	Bozeman, MT	USA

------------------------------------------------------------------------------