Hi Joe, We did not extensively test the patch. The two problems you found were things we did not test.
Regarding reproducing the problem of missed SFDs, you could set up a scenario similar to what I've explained in earlier e-mails. We can also give you access to our code. However, the code we have developed is mainly for our experiments and hence implements a lot of other things which makes it a bit more complex to use and not at all needed for reproducing the problem. Let me know if you want to try our code anyway. Best Regards, Martin On Fri, 2007-03-16 at 11:04 -0700, Joe Polastre wrote: > Hi Martin, > > I've tried your patch, and it actually degrades performance in high > bandwidth utilization cases. Receive is inadvertently called, and > often the same packet is received multiple times due to a race > condition that is caused by adding your patch. > > As for the radio missing SFD, I haven't been able to directly > observe this. If you have a test case that shows this behavior, I can > try it out. > > -Joe > > On 2/22/07, Martin Jacobsson <[EMAIL PROTECTED]> wrote: > > On Mon, 2007-02-19 at 13:28 +0100, Martin Jacobsson wrote: > > > > I don't know how long address decoding takes, but it's 7 bytes from > > > > the SFD to the end of the address byte. That's 56 bits, which on a > > > > 250kbps radio is 224 us. So if there's any series of atomic blocks > > > > that run a bit longer than that (224us + address decode time), then > > > > you might run into an issue like this. I believe that all of the > > > > MSP430 nesC implementations (TinyOS 1.x, Boomerang, TinyOS 2.0) make > > > > all interrupts non-preemptible for programming simplicity, so this > > > > could be a very real problem. > > > > > > Boomerang does not use the address decoding feature in the CC2420 and if > > > it did, I use a broadcast destination address. Anyway, my "hello > > > packets" are very small. They only carry a payload of two bytes. This > > > would mean that SFD is high for 14 bytes (= 448us). Perhaps that is > > > small enough. > > > > > > When I get some time, I will try to see if it is possible to reproduce > > > the error with a large packet size as well. > > > > I have now done some more experiments. It is possible to reproduce the > > error also with a large packet size. Instead of a 2 bytes payload, I > > used 25 bytes and it still happens. The important factor seems to be the > > frequency of packets. The closer it is between two consecutive packets, > > the bigger the chance. > > > > Hence, it is more plausible that the software misses the raising flank > > of SFD and not the lowering flank. > > > > Best Regards, > > Martin Jacobsson > > > > _______________________________________________ > > Tinyos-help mailing list > > Tinyos-help@Millennium.Berkeley.EDU > > https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help > > _______________________________________________ Tinyos-help mailing list Tinyos-help@Millennium.Berkeley.EDU https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
