[openib-general] Re: Re: [PATCH] (fixed) reduce qp locking on cq poll

[Michael S. Tsirkin]

Quoting r. Grant Grundler <iod00d at hp.com>:
> Subject: Re: Re: [PATCH] (fixed) reduce qp locking on cq poll
> 
> On Wed, Jan 26, 2005 at 10:13:22PM +0200, Michael S. Tsirkin wrote:
> > Its not the speed of hardware, my problem is the variations.
> 
> Ok. Your "log comment" was entirely misleading. It should read
> 	I can't measure a performance difference becuase of my
> 	test results vary by X%.
> 
> Substitute the right value for X.
> 
> > > Did you see the difference in IPoIB numbers that I posted
> > > earlier where I pegged the test processes to the same CPU
> > > taking interrupts or a different CPU? 
> >
> > No - link?
> 
> Sorry - I don't have the URL - just the original mail.
> I posted this on "Date: Thu, 13 Jan 2005 10:43:34 -0800"
> and you should be able to find this in the archives.
> 
> Here is the table again:
> netperf/irq same, netserver/irq same: 1591 Mbps
> netperf/irq diff, netserver/irq same: 1565 Mbps
> netperf/irq same, netserver/irq diff: 1507 Mbps
> netperf/irq diff, netserver/irq diff: 1510 Mbps
> 
> (1591-1510)/1510 is about 5%.
> If you use taskset on your platforms, run a similar table to the
> above, we could compare and see if ia64 is more sensitive to this
> than other arches.
> 
> hth,
> grant
> 

I finally found some time to do this.
taskset does not seem to help:

swlab155:~ # taskset 1 netperf -f M  -c -C -H 11.4.8.156 -- -m 100000
TCP STREAM TEST to 11.4.8.156
Recv   Send    Send                          Utilization       Service Demand
Socket Socket  Message  Elapsed              Send     Recv     Send    Recv
Size   Size    Size     Time     Throughput  local    remote   local   remote
bytes  bytes   bytes    secs.    MBytes  /s  % S      % S      us/KB   us/KB

 87380  16384 100000    10.00       301.38   44.20    92.95    2.865   6.024
swlab155:~ # taskset 1 netperf -f M  -c -C -H 11.4.8.156 -- -m 100000
TCP STREAM TEST to 11.4.8.156
Recv   Send    Send                          Utilization       Service Demand
Socket Socket  Message  Elapsed              Send     Recv     Send    Recv
Size   Size    Size     Time     Throughput  local    remote   local   remote
bytes  bytes   bytes    secs.    MBytes  /s  % S      % S      us/KB   us/KB

 87380  16384 100000    10.00       290.86   47.80    89.85    3.210   6.033
swlab155:~ # taskset 1 netperf -f M  -c -C -H 11.4.8.156 -- -m 100000
TCP STREAM TEST to 11.4.8.156
Recv   Send    Send                          Utilization       Service Demand
Socket Socket  Message  Elapsed              Send     Recv     Send    Recv
Size   Size    Size     Time     Throughput  local    remote   local   remote
bytes  bytes   bytes    secs.    MBytes  /s  % S      % S      us/KB   us/KB

 87380  16384 100000    10.00       295.07   43.46    90.45    2.876   5.987


Bandwidth went anywhere from 301.38 to 295.07.


-- 
MST - Michael S. Tsirkin