Re: [avrdude-dev] Bad programming speed on AT90S8535 using AVRdude

[Theodore A. Roth]

On Tue, 18 Nov 2003, Theodore A. Roth wrote:

> 
> 
> On Tue, 18 Nov 2003, Brian Dean wrote:
> 
> > On Tue, Nov 18, 2003 at 06:20:00PM -0500, Brian Dean wrote:
> >
> > > I didn't try Jan-Hinnerk's patch on FreeBSD, but I tried your file
> > > with the latest cvs and with your (Ted's) patch.  The result in both
> > > cases is 13.02 seconds real time as reported by the shell.  AVRDUDE
> > > reports 12.54 seconds of programming time in both cases, i.e., no
> > > difference and seems to be on par with the STK500 and JTAGICE.  I also
> > > verified this with the STK500 on FreeBSD and got the same result
> > > (12.52 seconds as reported by AVRDUDE).  So, essentially, FreeBSD
> > > shows no difference in programming time when using the parallel port
> > > programmer vs the STK500.
> >
> > I tried this on a slower machine, a Pentium II 300 MHz.  It programmed
> > the board in 39 seconds (parallel port) and 12.53 seconds (STK500).
> > This is closer to what I remember the difference being.  Also, way
> > back when I tested before, the 300 MHz machine was my primary machine.
> > I'm now using an Athlon 2200 which gave 12.5 seconds for the parallel
> > port, STK500, and JTAGICE, so I'm thinking this is about as short as
> > the programming time is going to get, regardless of the programmer.
> >
> > However, I wouldn't expect that big of a difference in the
> > timing. between the two systems for the parallel port.  Surely this is
> > not a CPU bound process.
> >
> > Maybe differences in motherboard chipsets?
> >
> > I do note that my older system is running:
> >
> >      FreeBSD vger.bsdhome.com 4.7-STABLE FreeBSD 4.7-STABLE #0: Wed Nov  6 
> > 19:01:57 EST 2002     address@hidden:/scratch/usrobj/usr/src/sys/VGER  i386
> >
> > While my newer, faster system is running:
> >
> >      FreeBSD neutrino.bsdhome.com 4.9-STABLE FreeBSD 4.9-STABLE #0: Tue Nov 
> > 18 14:49:25 EST 2003     address@hidden:/usr/obj/usr/src/sys/NEUTRINO  i386
> >
> > These are about a year apart.  If I get a chance, I'll update my older
> > system to the latest code, just to rule out a change in the OS to
> > explain these big differences in parallel port programming time.
> 
> Interesting. I'm on a PIII-450 here at work. I'll try it on my home
> system (dual Athlon 1800) later and see if there is a difference.
> 
> I've digging on google for some reason why and found nothing. I also
> tried tweaking the ppdev stuff in avrdude and came up with nadda.

On my athlon system with current, unpatched cvs source:

  Writing | ################################################## | 100% 17.62s
  avrdude: 32052 bytes of f written
  avrdude done.  Thank you. 
  real    0m18.074s

With my patch:

  Writing | ################################################## | 100% 16.57s 
  avrdude: 32052 bytes of f written
  avrdude done.  Thank you.
  real    0m17.046s

And again for grins, uisp is a tad slower than avrdude:

  Uploading: flash
  ################################
  (total 32052 bytes transferred in 19.04 s (1684 bytes/s)
  Polling: count = 126, min/avg/max = 2.31/2.33/2.79 ms 
  real    0m19.441s

Looks like linux just isn't as efficient with it's ppdev driver as FreeBSD 
is with ppi.

Ted Roth