Re: [avrdude-dev] ppi speed-tuning ;-)

[Jan-Hinnerk Reichert]

On Thursday 20 November 2003 04:58, Brian Dean wrote:

> With this patch also applied, that time went from 24.41 seconds
> down to 20.84 seconds, so that gained us another 15% or so.
>
> The first patch was the biggie, though.  It's not often we get to
> see a 6 times speed-up!

This patch should give about 40% on paged memory. That was the reason 
for writing it. However, the 15% on non-paged memory are a nice 
goodie.

So we should be equal to STK500 there, too.

I'm still wondering why ioctl() is so slow. Has anyone any ideas?
AFAIK, an outb() or inb() takes approx 1us, but calculation shows that 
on ioctl() that does a single port access takes almost 3us.

Has anyone got a PCI I/O-Card to check if it is faster there?

> And the best thing about it - the code still verifies correctly :-)

;-)))

> Thanks, Jan-Hinnerk and Ted for going through all these variations,
> figuring out what works and what doesn't.
>
> BTW, Jan-Hinnerk, for not even having a PPI programmer, you are
> doing pretty well :-)

I've written an AVR910-based programmer at a time where there were no 
paged memory devices and polling was commented out in Atmel's own 
AVR910-code (seems like they weren't able to get it working ;-)

So, I've spend quite some time understanding how this polling stuff 
works. I'm glad that this knowledge now gives some benefits to a 
greater audience.

/Jan-Hinnerk