The problem is nothing to do with anyones network it's the armx6
I've just spent a fortnight rewriting the EtherTH driver - I find it
easier to experiment in C rather than assembler - but in the end it
suffers from the same problems as the original EtherTH ie slow writes of
large files.
This isn't program specific, LanManFS has the same problems as LanMan98.
The slow writes are erratic. I test by transferring a 10mb file and
reading it back.
When connected to my laptop by wire I always read at about about 5.5MB/s
but writing can make it vary between 1 and 5.5MB/s. Just when you think
you have a break through and get 5 5.5MB/s transfers in a row it will
start being eratic again. If I use wireless on my laptop writes are
still eratic at around 0.5MB/s to 0.9 MB/s and reading is always around
0.9MB/s so it's not the sending speed required that is the problem.
You will notice that when writing is slow the hourglass appears in
LanMan98, at this point LanMan98 is waiting for a reply to a packet that
it has sent. However the laptop hasn't received the packet and when
sending resumes the laptop receives not only the original packet but
also a retried packet which the armx6 sockets send as part of the
protocol because it thinks the first one is lost.
Tweaking threshold values make little or no difference. It's nothing to
do with flow control - not networking flow control anyway - as no pause
packets are sent or received while while transferring data.
I had thought maybe the reads are overwhelming the computer - they
happen under interrupts and you have to remember that even when sending
you are receiving acknowledgements of packets all the time - but slowing
them down makes matters worse.
Anyway I'm still playing with it.
--
Colin
---
To alter your preferences or leave the group,
visit //www.freelists.org/list/armini-support
List-related queries to info@xxxxxxxxxxxx
