Re: ECP port driver

[Robert Creager <robert@carolinecreager.com>]

Shawn,

Well, that's good enough for me.  I will shoot for
supporting 2.2 kernel (at least 2.2.14).  Shawn, would
you be willing to run a test or two when the time
comes?  I'll try not to give you anything with causes
a kernel panic :-)  I'm downloading that kernel now
(for the next few hours), so I can at least compile
against it.

Thanks,
Rob

---- "RollingHillsObs (Shawn Dvorak)"
<rollinghillsobs@cfl.rr.com> wrote:
> My 486 Linux box was retired some months ago, but I
gave this a try on a
> Pentium 166 running Redhat 6.2 (= kernel 2.2.14). 
After just a bit of
> twiddling to get the "short" module to compile, I
got the following results
> from three runs:
> 
> [root@homebase /dev]# time dd if=/dev/short0s bs=16
count=1000000 >/dev/null
> 1000000+0 records in
> 1000000+0 records out
> 3.12user 26.53system 0:32.38elapsed 91%CPU
(0avgtext+0avgdata 0maxresident)k
> 0inputs+0outputs (100major+14minor)pagefaults 0swaps
> 
> [root@homebase /dev]# time dd if=/dev/short0s bs=16
count=1000000 >/dev/null
> 1000000+0 records in
> 1000000+0 records out
> 2.89user 27.38system 0:33.10elapsed 91%CPU
(0avgtext+0avgdata 0maxresident)k
> 0inputs+0outputs (100major+14minor)pagefaults 0swaps
> 
> [root@homebase /dev]# time dd if=/dev/short0s bs=16
count=1000000 >/dev/null
> 1000000+0 records in
> 1000000+0 records out
> 3.37user 26.28system 0:32.42elapsed 91%CPU
(0avgtext+0avgdata 0maxresident)k
> 0inputs+0outputs (100major+14minor)pagefaults 0swaps
> 
> Average clock time was 32.6 seconds, and CPU
utilization was solid at 91%.
> Although a 486 w/ Linux has generally been the
agreed-upon platform for
> running the cameras, I don't think that an old
Pentium like one I used would
> be much of a stretch for any budget today.  There
isn't much price
> difference on EBay or the like for a 486 vs. an
ancient Pentium.  And,
> Pentium motherboards generally accept DIMMs as
opposed to the much-pricier
> SIMMs required by most (all?) 486s.
> 
> Shawn
> 
> 
> ----- Original Message -----
> From: "Robert Creager" <robert@carolinecreager.com>
> To: "Fraser Farrell" <fraser@trilobytes.com.au>;
"Robert Creager"
> <robert@carolinecreager.com>
> Cc: <tass@listserv.wwa.com>
> Sent: Thursday, September 06, 2001 4:56 PM
> Subject: Re: ECP port driver
> 
> 
> >
> > Hey,
> >
> > I'm hoping some fine person, with a slow box
> > (486/Pentium) and a 2.2 kernel can run a simple
test.
> > On the link below is a tarball of examples from
the
> > Linux Device Drivers 2nd edition book.  Within
said
> > tarball is a driver called short.  While it
doesn't
> > run the parallel port in ECP mode, it should give
a
> > rough idea of if the computer is capable of
> > downloading the data in a timely manner.  After
> > building and loading the module (by using make and
> > then short_load), execute the following statement:
> > time dd if=/dev/short0s bs=16 count=1000000 >
> > /dev/null
> >
> > What this will do is execute 1 million 16 byte
> > transfers (using insb), trashing the results.  The
> > /dev/short0s is the device which uses the insb
macro,
> > vs inb.  If the time of this command is less than
40
> > seconds, the computer might (should?) be able to
keep
> > up.
> >
> > For comparison, my AMD K6-2 500Mhz (roughly a
Pentium
> > II), will execute the above in about 31 seconds,
using
> > 85% CPU (it's almost CPU bound, but not quite).
> >
> > http://examples.oreilly.com/linuxdrive2/
> >
> > ---- Fraser Farrell <fraser@trilobytes.com.au>
wrote:
> > > Robert (and list),
> > >
> > > >Hmmm... 2.2 you say.  I was shooting for 2.4 -
it
> > > >makes supporting devfs easier to do - no need
for
> > > >IFDEFS.  Any particular reason you indicate
2.2?
> > >
> > > The ECP port was introduced around the era of
the
> > 486; so I was
> > > thinking of a scenario where a 486/Pentium-class
> > computer sits outside
> > > in the cold & damp to run the Mark IV. Most of
the
> > prebuilt distros
> > > with kernel 2.4 are a bit overwhelming for older
> > hardware. Or won't
> > > run at all.
> > >
> > > OTOH you may need a powerful box with kernel 2.4
to
> > run the image
> > > analysis software.
> > >
> > >
> > > >If there are serious requests for 2.2, is
anyone
> > > >willing to test some code on their platform?
> > Although
> > > >I can grab another version of the kernel, my
server
> > is
> > > >my playground, and I dislike rebooting often
:-)
> > >
> > > My server had its second reboot for this year
two
> > days ago. A power
> > > cut which outlasted the UPS. I used this
opportunity
> > to replace a
> > > noisy fan :)
> > >
> > >
> > > >Speaking of distribution specific, what
> > architectures
> > > >might someone use this on, and would that
someone
> > be
> > > >willing to test some code?
> > >
> > > Hmmm...of the boxes with ECP ports here; I can
do
> > Redhat 6.1 (kernel
> > > 2.2.12) Mandrake 8.0 (kernel 2.4.something) or
> > Slackware 7. I've also
> > > got Redhat 7.1 but nowhere to install it yet. I
> > could also try it out
> > > with a couple of the mini-distros such as Dragon
or
> > Phat.
> > >
> > > What I don't have is a Mark IV (or something
that
> > can simulate a Mark
> > > IV).
> > >
> > > It may be worth developing a utility that can
run
> > under DOS (or Linux)
> > > to simulate a Mark IV on any computer with an
ECP
> > port. Including the
> > > output of known "test pattern" images for
> > verification of your setup
> > > and data analysis.
> > >
> > > The ability to swap in a known-to-be-working
> > replacement is very
> > > helpful for troubleshooting. But having two Mark
> > IV's at every
> > > location probably isn't Tom's idea of effective
> > surveying! A simulator
> > > would also be useful for anyone without the
camera
> > who wants to write
> > > control software for it.
> > >
> > > Apologies if all this has been discussed before.
> > >
> > >
> > >
> > > cheers,
> > > Fraser Farrell
> > >
> > >
> > >
> > >
> >
> 
> 
> 
>