On Thu, Oct 9, 2008 at 4:32 PM, Al Chu <[EMAIL PROTECTED]> wrote:
> Hey Sasha, Hal, all,
>
> Now I'm thinking this:
>
> if user specifies port 255 - they are expecting AllPortSelect.  So error
> out.
>
> if user specifies '-a' - Do whatever it takes to query/reset on all
> ports.  If it means looping do that.  Aggregate counters appropriately
> for a single output.
>
> if user specifies --loop-ports - loop through the ports one by one no
> matter what.  This includes outputting port counters for each port.
>
> Sound like a good idea?

Sounds good to me too.

-- Hal

> Al
>
> On Thu, 2008-10-09 at 21:39 +0200, Sasha Khapyorsky wrote:
>> On 09:31 Thu 09 Oct     , Al Chu wrote:
>> >
>> > I just re-read this sentence, now I get what you're asking.  I made --
>> > loop_ports only loop if AllPortSelect isn't supported.  So if the
>> > switches already support AllPortSelect, the number of resets should be
>> > the same.
>>
>> OTOH perfquery is diagnostic tool. Reset with AllPortSelect may not work
>> as expected or an user may have another reason to loop over ports
>> regardless to AllPortSelect support.
>>
>> Sasha
> --
> Albert Chu
> [EMAIL PROTECTED]
> Computer Scientist
> High Performance Systems Division
> Lawrence Livermore National Laboratory
>
>
_______________________________________________
general mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general

To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general

Reply via email to