As Uwe Bonnes wrote: > Running my test (as discussed on libftdi) from yesterday with current SVN 1007 > slow dowm enourmous:
Could you compare the trace data (from -vvvv) between both versions? The major difference between previous versions and r1007 is that the programmer's paged_load and paged_write methods are now called once per page, rather than once for an entire memory region. Actually, paged_load is still called once for the entire region, when a full region of the entire memory has been requested. I think about changing that though, so we've got a uniform interface in the lower layers, and can remove all loops and boundary checks there, relying on the caller calling the method for exactly one page. But, you could use that for a comparison: run the command with -U flash:r:foo.hex:i, this will cause the entire flash memory to be read. Is this also still faster then the verification read in r1007? p.s.: Thanks for the report! It's for a reason that I did that change right after a release, so we've got enough time to shake out all bugs before the next one. -- cheers, J"org .-.-. --... ...-- -.. . DL8DTL http://www.sax.de/~joerg/ NIC: JW11-RIPE Never trust an operating system you don't have sources for. ;-) _______________________________________________ avrdude-dev mailing list [email protected] https://lists.nongnu.org/mailman/listinfo/avrdude-dev
