> On Mon, 24 Apr 2006 08:02:47 -0700, Chops wrote:
>> Ten of 115200bps is a pretty non-trivial task.  By my calculations,
>> it gives you about 4 microseconds per character to do... everything.
>
> Very true -- this issue comes up on the Rabbit Semiconductor users
> group frequently... The R3000 CPU has six UART channels and the Rabbit
> C compilers all provide interrupt driven serial library routines.
> Invariably, some newbie asks "why am I missing characters when I try to
> run 4 (or 6) serial ports at 115Kbps?".
>
> In fact, even one channel of data at 115Kbps is quite a load for most
> lowly 8 bit controllers running at a moderate clock rate.
>
> Matt Pobursky
> Maximum Performance Systems

Absolutely.

Some years ago I developed a system based on an x86 platform that used many 
high speed UARTs. The UARTs were not the standard ones used by PCs, but 
rather Zilog SCC chips. As originally designed, the system ran under 
Borland's 32-bit DOS extender. Unfortunately, it turned out to be completely 
unworkable, even with a 200MHz+ 32-bit CPU, because of the ridiculously long 
pathlengh in handling interrupts in the DOS extender. I eventually had to 
develop my own Win32 subset equivalent kernel that processed the character 
interrupts with almost no overhead. It works really well now!

A similar example is when one of my customers insisted that his PDP-73 could 
easily handle multiple 9600 bps serial data streams. Again, this turned out 
to be a disaster and in this case we had to crank the data rate down to 2400 
bps to get the application to work at all. :-(

Bob Ammerman
RAm Systems

-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist