Subject: Re: Heavy /dev/lpt use burdens system.
To: None <current-users@netbsd.org>
From: Thor Lancelot Simon <tls@rek.tjls.com>
List: current-users
Date: 09/10/1999 16:57:46
On Fri, Sep 10, 1999 at 02:13:10PM -0600, Brook Milligan wrote:
>    It seems that the printer driver causes an incredible number of interrupts
>    per seconds, thousands of them.  Probably as many as your machine can
>    handle.  You should try the interruptless driver, but what is really needed
>    is for someone to fix the ltp driver, perhaps to use DMA.
> 
> I have been trying the interruptless driver, but it goes incredibly
> slowly.  Like a line a minute or something.  I was thinking of
> switching to the interrupt driver, but now I'm not so sure that will
> help.

I guess I don't understand what people seem to think is wrong with the
interruptful version of the driver.

Where I used to work, the main office printer was a Lexmark postscript
printer with 32MB of RAM connected by a very long parallel cable to our
main fileserver, router, and development machine -- for a long time, the
only "server" in our office.  This printer was used to print primarily
simple text and occasionally small graphics and most jobs fit quite nicely
in its 32MB of memory such that the engine speed wasn't an issue; it'd
slurp down all the PostScript as fast as the interface could send it.

When the machine was a 200MHz Pentium Pro, I saw the printer, when in use,
generate 3000-5000 interrupts per second fairly regularly.  The lpt
interrupt service routine's pretty lightweight and this did *not* seem
to impact performance in any way -- note that lots of other people were
using this machine for all kinds of stuff and weren't shy about complaining.

The machine was for some months, a little later, a 433MHz AlphaPC164.  I saw
*this* machine peak at over 12,000 'lpt' interrupts per second.  Again, this
did not appear to impact overall performance in any way.