Phil,

On Wed, Nov 08, 2006 at 11:00:28AM +0100, Philip J. Mucci wrote:
> Hi Stefane,
> 
> Ok, I was able to get the PMD/NMI working on my P6 laptop...And I
> noticed some trouble with Perfmon2. Since the NMI took PMD and PMC0, my
> enable bit seems to have been stolen. Therefore, even though I could
> program PMC1/PMD1 to count, I could not get any counts due to that
> enable bit being missing. I always got 0. I'm not sure what the expected
> behavior is here...but the library code did the right thing, I just
> always got 0. I can send you the logs if you would like.
> 
Yes. This does not work on P6 because of the single enable bit 
for all counters. It works only on AMD64 and Intel Core 2.
The P4 has other issues I have not yet solved.

On Intel Core2, I will ask Andi to move the NMI to IA32_PMC1 because
even though, each counter has an individual enable bit (if you ingore the
GLOBAL_CTRL register), PEBS requires using IA32_PMC0.


> To test the PAPI support, I hard coded the unavail mask to PMC1/PMD1 and
> that worked, adding the counter in the proper place. (Will, this may run
> on your AMD64 laptop now...)
> 
Yes, it should work on Will's laptop.

> Now, in regards to the PFM API:
> 
> I noticed that there is a set of calls for implemented counters AND a
> set of calls for available counters (using the eventset). Couldn't these
> two me merged? From an application standpoint, all we care about is what
> we can use...is there ever a reason to check 'impl' when using 'avail'?
> 
Are you talking about libpfm or pfm_getinfo-evtsets()?

-- 

-Stephane
_______________________________________________
perfmon mailing list
[email protected]
http://www.hpl.hp.com/hosted/linux/mail-archives/perfmon/

Reply via email to