Am 03.07.2012 um 16:10 schrieb Darell Tan:
> Hi Hannes,
>
> I took a quick look at the avrftdi code and realised it re-implemented
> a lot of the pgm functions. In theory, it should have been enough to
> implement just the cmd() function to "buffer" the op itself, to tell
> the serial engine to generate the CLK and MOSI signals. And of course
> the setpin() function to turn off/on LEDs or the RESET line.
Yes. In fact, that is what my first prototype did. However, without paged
read/write it would is rather slow. Especially with the old USB1.1-devices.
I tie the /RESET line to TMS, so you could even leave that out - it is changed
automagically. And who needs LEDs? :)
>
> If I had the time, I would implement TPI over avrftdi too, but it
> seems like I need to copy/paste (or re-implement) the TPI logic into
> avrftdi, which is not an elegant solution. If the code in avr.c could
> be re-used, then it would have been really simple to get TPI working
> for avrftdi.
I've had only looked at the TPI protocol definition, not what is already
available. But I agree that implementing TPI with avrdude would be rather
hack-ish.
My current approach would be to transmit 2 bytes with the lower 12 bits as the
TPI-frame and the higher 4 bits as '1' ("idle"). For reading, I would read in 2
bytes, with the TPI-frame in the upper 12 bits.
Of course, it would be possible to transmit one byte and 4 bits (since MPSSE
can clock single bits). But I think this could incur more overhead (shifting
bits around, more control information than data).
Luckily, PDI is very similar (without the /RESET line), so we could possibly
get away with implementing the "physical" layer once.
Best regards,
Hannes
_______________________________________________
avrdude-dev mailing list
avrdude-dev@nongnu.org
https://lists.nongnu.org/mailman/listinfo/avrdude-dev
