Ilja,

>I think the article (I somehow overlooked it) and Martin's comments will be 
>very helpfull.

Please not that the article is back from times of 0MQ version 0.1. The
principle still holds, some details are different. (Btw, I am thinking
of writing mini-blogs about internal architecture of 0mq so that people
hacking it have at least some intro.)

>I had an impression there is one ypipe per inproc endpoint, in other words all
>threads write and read to and from inproc://channel shares one ypipe structure.

ypipe connects exactly two threads.

>This is apparently wrong and I am glad 0mq is sane from multithreading 
>standpoint ;)

It should be. The synchronisation on level of CPUs cache coherency
mechanism is a black magic though. If you are an expert in the area, I
would appreciate if you shared your thoughts about the code.

Martin
_______________________________________________
zeromq-dev mailing list
[email protected]
http://lists.zeromq.org/mailman/listinfo/zeromq-dev

Reply via email to