> Multiplexing of Northbridge events does not work with the current
> 2.6.30 code base. I was suggesting
> you keep using system-wide but instead use two separate runs.
>
> pfmon --with-header --outfile=test1 --pin-command -u
> -eDRAM_ACCESSES_PAGE:HIT,DRAM_ACCESSES_PAGE:MISS,DRAM_ACCESSES_PAGE
On Mon, Nov 30, 2009 at 4:50 PM, David Nellans wrote:
> Hi Stephane,
>
> I have been working with Manu - thus the follow up in another thread.
>
> Using system-wide monitoring alleviates the problem as I mentioned in
> another email.
>
> Its not clear how we can use two monitoring processes for
Hi Stephane,
I have been working with Manu - thus the follow up in another thread.
Using system-wide monitoring alleviates the problem as I mentioned in
another email.
Its not clear how we can use two monitoring processes for a single
8-thread app without using
system-wide monitoring since each
Hi,
Sorry for the long delay. I missed your email initially.
I am trying this with the 2.6.30 kernel. It does not crash
there but I see a kernel with the way control of Northbridge
events is handled in the presence of multiple event sets.
That does not work.
So I suggest you try with two separat
Hi all,
I have been measuring memory events for the parsec benchmark suite on a dual
socket, quad-core opteron machine with pfmlib version 3.9, kernel pfmon
version 2.82 and kernel version 2.6.29.6 . This is what I use as my
commandline option:
>> pfmon --with-header --outfile=test1 --verbose -u -
