>>>>> "Jason" == Jason Thorpe <thorpej@SJ.Xenotropic.COM> writes:

[on the ZRLB change]

	As a followup, I got pleanty of fifo overflows last night, but
the other end didn't quit from the number of errors.  The observed
throughput is about 1-1.2K receiving... but that could be affected by
any number of problems... (plus I was using telnet, too).

	If I read the code correctly, the ring will hold upto 64 bytes
of in or output as it stands (256).  This is increased to 1024 with
the change to 4096.  I think that the router was quitting because it
was having trouble even sending a control structure for PPP (which
could easily be more than 64 bytes) and eventually quitting.

	This change seems to prevent that.  I don't know if I'd want
to be using a 19.2K plain serial connection with this machine for
much.  However, with PPP it is very usable.

>> The next thing that I'd like to know about this mess is whether or
>> not the priority (or somesuch) can be bumped on this thing.  My
>> 68030 based Amiga manages to run 19200 w/o any loss without even
>> the help of a fifo on the serial port.  Why is the Sun4 (about 2
>> times faster) having problems?

Jason> Actually, I've always been under the impression that Amiga's
Jason> had Really Buff serial ports in general...Also might be a
Jason> side-effect of the way interrupts are handled.

	Not really.  Single byte fifo serviced by an interrupt.  The
only thing special about them is a clock that allows them to run at
the 'special' MIDI speed.  On the stock systems, even with a 68030,
many serial applications have trouble at 38.4 or 56... but some work.

Jason> And, really, a 4/260 is twice as fast as an Amiga?!  :-)

	Well... If a 68030 put's all it's MIPS in one basket (mine
rarely does since it's a news/mail hub), then it's somewhere between 3
and 5.  Something like a Sun3/260.  The Sun4/260, as I understand, is
about 10 mips or so.

	That said, with this graphics card and disk system, the Sun4
doesn't seem at all faster than the Amiga.  I especially hate how the
machine seems to lock up under disk accesses.