On Tue, Feb 8, 2011 at 7:59 PM, Venkateswararao Jujjuri (JV)
<jvrao@xxxxxxxxxxxxxxxxxx> wrote:
>>
>>> the implementations it should be really easy to shut me up with
>>> comparison data of zc and non-zc for 1, 64, 128, 256, 512, 1024, 2048,
>>> 4192 byte payloads (without caches enabled of course).
>>
>> I think this is good experiment will publish data.
>
> BTW, unless we have bigger msize with differentiating pdu sizes these experiments
> may not make sense.
>
Not sure I agree (at least in the 1-4k scale), we are measuring the
overhead of memcpy versus the overhead of mapping/pinning the
additional sg -- or am I not thinking clearly. Its possible... I have
not had coffee yet. I suppose with your small buffer patch series
there might be some performance differences due to different allocator
behavior, and while I don't think it'll be significant, it may be
worth re-doing the experiment once we have that in place.
Which brings up another question -- I know your team are doing
functional regressions, but are they also dong performance
regressions? Since we are starting into the optimization patches it
may not be a bad idea to track how the changes are impacting
scalability, latency, and throughput (as well as some metric of
resource consumption, but that may be a harder metric to track).
-eric
-eric
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
