On Mon, Oct 23, 2000 at 08:40:34PM -0400, Michael 'Maki' Kato wrote:
> I always thought that the 4000PC came chronologically before the 4000SC, is 
> this not the case?
> 

I'm guessing that maybe some Magnum 4000PC's were manufactured AFTER the 4000SC
came out.  I don't know.
...
Yesterday I contacted LSI Logic in an attempt to get the tech specs for the
"L1A7385" I/O ASIC.  The woman I spoke with told me that they don't catalog
their parts using the L1A numbers.  I explained to her that the chip was in 
the same family as the R4030 and R4230.  She called me back and said she 
couldn't find the specs for ANY of these chips, since they were too "old".
So I hit a dead end.  Next I set out feverishly disassembling the ARC firmware,
which I'm still doing as we speak.  The only place I found it checking the
ASIC_REVISION (register 8), was at offset 0xd94.  If it's less than 3 (which
mine is), it skips over checking for L2 cache, which didn't make much sense.
Also, the only place in ARC that touches the EXT_IMASK register is at offset
0xdc0, where it enables all interrupts.  I also found out that the I/O ASIC
has its own I/O cache, which is being completely ignored by NetBSD.  I wonder
if this is causing a problem.  Don't know.  I wouldn't mind having some more
eyes helping me go through disassembling ARC.  Without access to tech specs,
this seems the only avenue for information at this point.  I would recommend
using linux's include/asm-mips/jazz.h and jazzdma.h for reference to the
I/O ASIC's registers.  It proved very helpful.  NetBSD's pica.h is woefully
incomplete!  In any event, I see no evidence in ARC as of yet that would
indicate that my I/O ASIC were any different from a 4030/4230.  I'm not sure
what to try next...

-Mark