Is this an AMD class CPU?
The way the counters work on some platforms is that they count _up_ to
an overflow value, trigger the overflow interrupt, and PMC then is
supposed to calculate what the original value was.
It sounds like maybe that isn't working correctly for a certain class
of counters?
On Tue, 20 Sep 2011 16:12:57 + (UTC), Anton Yuzhaninov wrote:
AY> pmcstat -C -w1 -d -p
AY> INSTR_RETIRED_ANY -p DTLB_MISSES.ANY -p MEM_LOAD_RETIRED.DTLB_MISS -p
AY> ITLB_MISSES.ANY -p MEM_LOAD_RETIRED.L3_MISS
AY>
AY> With -C values should monotonous increase, but in output I see
AY> overflous.
It seems to be, that hwpmc counters overflow is not properly handled.
I want to count TLB misses and run:
pmcstat -C -w1 -d -p
INSTR_RETIRED_ANY -p DTLB_MISSES.ANY -p MEM_LOAD_RETIRED.DTLB_MISS -p
ITLB_MISSES.ANY -p MEM_LOAD_RETIRED.L3_MISS
With -C values should monotonous increase, but in outpu
Chris Brennan wrote:
> Anything gonna be done with this? It has promise but needs some more people
> involved.
The original web page which started this thread was removed from the
FreeBSD.org web site:
http://lists.freebsd.org/pipermail/cvs-all/2011-September/345334.html
For anyone who wants to
