Omprakash,
there are really two lookup implementations in LruCtpMsgCacheP. When I
found the internal one (it comes first), I missed the missing "Cache."
in front and thought it was the published one... Uh.
Anyway, it's good to see that I understood the intention of the code
correctly. The CTP is a really nice thing.
Thanks,
Oliver
Omprakash Gnawali schrieb:
> On Thu, Jul 31, 2008 at 7:01 AM, Oliver Frietsch
> <[EMAIL PROTECTED]> wrote:
>> Hello list,
>>
>> I'm currently working on a special-purpose derivate of CTP and therefore
>> read all the source code to get an overview. One point that is totally
>> unclear to me is in CtpForwardingEngineP (most recent CVS release).
>> It says (line 439):
>>
>> // Once we are here, we have decided to send the packet.
>> if (call SentCache.lookup(qe->msg)) {
>> call CollectionDebug.logEvent(NET_C_FE_DUPLICATE_CACHE_AT_SEND);
>> call SendQueue.dequeue();
>> if (call MessagePool.put(qe->msg) != SUCCESS)
>> call CollectionDebug.logEvent(NET_C_FE_PUT_MSGPOOL_ERR);
>> if (call QEntryPool.put(qe) != SUCCESS)
>> call CollectionDebug.logEvent(NET_C_FE_PUT_QEPOOL_ERR);
>> post sendTask();
>> return;
>> }
>>
>> OK, so it should *stop sending* and drop the next packet, in the case
>> that this packet (i.e. something very similar...) has already been sent
>> before. That's what I think...
>>
>> Unfortunately, my opinion collides with the comment "Once we are here,
>> we have decided to send the packet."
>
> The comment is misleading. As you note, there are more tests done
> after the comment before deciding to send the packet.
>
>
>> and the implementation of the cache
>> itself.
>>
>> As stated in LruCtpMsgCacheP, the result of lookup is: "if key [packet]
>> is in cache returns the index (offset by first), otherwise returns count
>> [of enqueued packets]".
>> So... The packet is dropped *everytime*, except that is is the first one
>> in the sent cache or the cache is empty? What does this position imply
>> for the packet, except that it is the oldest one that I can remember?
>>
>> I'm very sure that I missed something important, as this implementation
>> looks senseless to me ATM. Please help me!
>
> Cache.lookup returns TRUE if a packet with the same signature is in
> the cache. This TRUE/FALSE return by Cache.lookup is different from
> what the internal lookup function in LruCtpMsgCacheP.nc returns. The
> internal lookup function in LruCtpMsgCacheP.nc returns the position of
> a packet in the cache if a similar packet is found, otherwise the size
> of the cache. CTP does not use the internal lookup function
> LruCtpMsgCacheP.nc.
>
> - om_p
_______________________________________________
Tinyos-help mailing list
[email protected]
https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
