Time-profile of the DLG event-pipe / sr_waitEvt()

From: "Frode Greger Sivertsen" <frode@netconnect.no>

However, this doesn't affect only solicited sr_putevt() events, but all events passing through the pipe.

The "for I = 1 to 10000: sr_putevt()" completes in 20 ms. It is the read-back that takes time.

I guess you could run a non-sr_putevt() test by doing dx_stop(no-event flag) on an idle voice-channel or similar.

Say, for instance, that all your worker threads should get blocked for a while because of an application error: The event-queue will grow, and the pipe will have its throughput lowered from then on, reducing the overall performance of the app (plus - the memory isn't free'd up).

This is after all one of the central parts of the API, so I find it a little awkward that the implementation is the way it is. (An Exponential-Time algorithm, for one thing..)

But hey - I only commented on this because I was doing some testing for myself, and thought other people maybe would like to know the numbers .. Lets leave this now and focus on more important, happier things.. =)

 - Frode

To reply: mailto:General.26291@pysdiscussext.py.intel.com
To start a new topic: mailto:General@pysdiscussext.py.intel.com
To login: http://members.support.dialogic.com/wbproxy/wbpx.dll/~general
Received on Wed Nov 05 16:38:57 2003