[openib-general] RE: some questions/comments on gen2 udapl code

[Or Gerlitz]

Arlin 
 
> I can work on rolling up the uat/ucm/uasync processing into one thread.
 
Sounds good, so each udapl process would consume less resources from the system.
 
>We need some help from verbs to make this uDAPL CQ mapping work:
> Either
> support multiple CQ FD's, one per CQ.
> or
> modify call ibv_get_cq_event() and allow the user to supply the specific ibv_cq event to get, not 
> just the first one on the event queue.
 
I think the first question we want to ask - do we need multi-CQ/threaded support at the gen2 udapl? - the udapl
consumers i am aware to - mvapich and another IB commercial mpi open one DTO EVD , how about intel mpi?
 
So before driving changes to libibverbs lets see if there's a need. When the app that needs it comes, we can 
then either suggest a change in libibverbs or implement two flavor of the udapl library one for libdapl.so and libdapl.mt.so
where lidapl.so will assume no multi cq/threads and libdapl.mt.so not assume this.
 
and there's udapltest - which does use multi thread/cq - but should we spend work on only synthetic test - i dont
know - what do you say?
 
Or.