[openib-general] questions and a comment on the perftest package

[Or Gerlitz]

Michael S. Tsirkin wrote:
> Quoting r. Or Gerlitz <ogerlitz at voltaire.com>:
>> oops, i have just noted that read_lat.c practically ignores the tx_depth 
>> param... so stamp[i+1]-stamp[i] is indeed the wall time of the i'th 
>> operation. Anyway, i guess you would be open to get a patch that does 
>> exercise tx_depth in a similar fashion to read_bw.c ?

> Something like this has been on my todo list for a while.

OK, anyway, we will not be able to work on this immediately, but this 
way or another we need a way to measure the latency per op per tx_depth 
and transfer size, so unless you would fix it before, we would need to 
do so...

> However, isn't it the case that just giving tx depth = 1 to rdma_bw we get
> all the necessary deltas?

i was talking about read_lat.c and i want to get the correct delta for 
each possible value of tx_depth, not just for tx_depth=1

> So the right thing to do, IMO, is to take rdma_bw and teach it to report latency
> as well.  We thus will have a single test that measures both BW and latency for
> reads, and have number of in-flight messages as parameter.  With tx_depth = 1
> we'll get ping-pong.

I don't see who does the -pong here, its a client doing rdma read from 
(rdma write to) a server. Your suggestion makes sense, we will look into 
this.

Or.