[openib-general] Re: [PATCH] IPoIB splitting CQ, increase both send/recv poll NUM_WC & interval

[Shirley Ma]

Roland Dreier <rdreier at cisco.com> wrote on 04/19/2006 04:36:50 PM:

>     Shirley> Since I haven't found any kernel use 128 bit address, I
>     Shirley> use wr_id to save skb address, DMA mapping and other
>     Shirley> stuffs are saved in skb->cb, which is the private data
>     Shirley> for each protocal layer in skb. Same for rx_ring, so
>     Shirley> rx_buff and tx_buff is not necessary to be used.
> 
> That sounds like a good plan.  Although I guess we want to keep around
> some sort of rx ring if only to free receive buffers when bringing an
> interface down.  But that could become a linked list with the
> list_head in skb->cb I guess.

Yes, a list is needed for gracefully shutdown.

> Does all this matter for performance?
> 
>  - R.

It helps the performance about 10% for the touch netperf/netserver test,
then hit driver errors. I notice that the send is faster than before. Let 
me 
send you the patch tomorrow, maybe you have the hints to identify the 
problem.

Thanks
Shirley Ma
IBM Linux Technology Center
15300 SW Koll Parkway
Beaverton, OR 97006-6063
Phone(Fax): (503) 578-7638
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/general/attachments/20060419/010e8e79/attachment.html>