On Mon, 13 May 1996 18:07:44 -0700 
 Andrew Carey <careya@peak.org> wrote:

 > I'd recommend using the mux (dcm) if you have a hp300 lower than a 380 (and
 > even then..).  Using a plug-in dca board (98644A, same as the internal
 > serial port), I got numerous silo overflows and max throughput of around
 > 120 bytes/sec.

I definitely agree with Andy, here.

Actually, on a 380, 375, 345, and any 400-series, the DCA has a FIFO, so 
its not nearly as lame as it could be...however, the DCM card has much 
better hardware buffering and interrupt control...

 > Once I replaced the board with a dcm, all of my problems basically
 > disappeared.

...although, I wouldn't say that:

dcm0 port 0: uart overflow
[ . . . ]

10 of those after 2 days at 38400 ... however, that might be due to:

hpib0 at scode7 ipl 3: Internal HP-IB
dca0 at scode9 ipl 5: working fifo
dcm0 at scode13 ipl 3 flags 0xe
scsi0 at scode14 ipl 4: 32 bit dma, async, scsi id 7
le0 at scode21 ipl 5: address 08:00:09:06:48:bb

Note the interrupt levels...that's partially lameness on my part, for 
having the dcm card set wrong, however, under heavy disk and network 
loads, my dcm is going to have trouble dealing with interrupts because of:

#define splbio()        spl5()
#define splnet()        spl5()
#define spltty()        spl5()
#define splimp()        spl5()  

The reason for this flatness is because any device can be jumpered for 
any level from 3-5.  To, to maintain:

	bio <= net <= tty <= imp

...you have to make them flat.

After 1.2, I'm planning on fixing this, to calculate the required levels 
during autoconfiguration...the eventual goal is to be able to jumper the 
SCSI at ipl3, ethernet at ipl4, and both serials at ipl5.  The code would 
magically set:

	splbio == spl3
	splnet == spl4
	spltty == spl5
	splimp == spl5

That way, while my system is busy dealing with the disk and ethernet, it 
won't be blocking serial interrupts...that's the theory, anyhow :-)

The i386 port does something like this, although the i386 uses a mask, 
rather than a level scheme.

(Ok, so that was a bit of a tangent... :-)

----save the ancient forests - http://www.bayarea.net/~thorpej/forest/----
Jason R. Thorpe                                       thorpej@nas.nasa.gov
NASA Ames Research Center                               Home: 408.866.1912
NAS: M/S 258-6                                          Work: 415.604.0935
Moffett Field, CA 94035                                Pager: 415.428.6939