Hi, On Mon, Oct 18, 2010 at 5:08 PM, Woomin Hwang <wmhw...@core.kaist.ac.kr> wrote: > Hi all, > > > > I’m working with a system using two AMD Opteron 6172 processors, which have > 6 cores in a package. > > (linux 2.6.29, x86_64, pfmon-3.9) > > I want to know the possibilities of per-core memory access count. > > > > More specifically, with this perfmon2 tool, > > I want to count memory access of each core separately, especially the number > of read and write request. > > > > For example, assume that 6 cores run 6 processes concurrently. > > I want to know the number of memory read and write generated from each > cores. > > If it is possible, I want to get results like following. > > > > [core 1] 13 > > [core 2] 15 > > …. > > > > I used this command > > > > # pfmon --system-wide -e memory_controller_requests:write_requests > --print-interval=1000 --cpu-list=0,7 > > This event is part of a group of so-called Northbridge events (e.g, all HYPERTRANSPORT, L3_*). Those events have special constraints in terms of how they can be programmed.
Even though those events are programmed from each of the 6 individual cores on you processor package, you can only measure 4 instances and only in a certain way. These events are measured in the Northbridge which is a shared resource. The events are effectively programmed there, and the Northbridge can only count 4 events at a time. Furthermore, those events are coming from a shared resource off-core, therefore it is not possible to trace back which core caused the event. You can only count for the entire processor package. And for that, you only need to measure the event from one of the cores. > > But I can’t get any result except this error message. > > > > CPU7 error conflicting monitoring session > > > > I found a thread > > http://sourceforge.net/mailarchive/message.php?msg_id=68c6d53a0910301523n4eca6060k34f7d14a46e2baa%40mail.gmail.com > > > > What I understand is that there is no method to count the number of memory > access for each core separately. > > But it can be possible with the Intel Xeon processors. > > > > My questions are: > > 1) Is it impossible to count the number of memory access generated by > each core? (with Opteron processor) > > 2) Is it possible with Xeon processors? > > 3) If it is unavailable, is there any software method to get such > information? > > 4) If it is available, how can I get such information? > > > > Please help me. > > > > Thank you for your time. > > Woomin Hwang > > ------------------------------------------------------------------------------ > Download new Adobe(R) Flash(R) Builder(TM) 4 > The new Adobe(R) Flex(R) 4 and Flash(R) Builder(TM) 4 (formerly > Flex(R) Builder(TM)) enable the development of rich applications that run > across multiple browsers and platforms. Download your free trials today! > http://p.sf.net/sfu/adobe-dev2dev > _______________________________________________ > perfmon2-devel mailing list > perfmon2-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/perfmon2-devel > > ------------------------------------------------------------------------------ Download new Adobe(R) Flash(R) Builder(TM) 4 The new Adobe(R) Flex(R) 4 and Flash(R) Builder(TM) 4 (formerly Flex(R) Builder(TM)) enable the development of rich applications that run across multiple browsers and platforms. Download your free trials today! http://p.sf.net/sfu/adobe-dev2dev _______________________________________________ perfmon2-devel mailing list perfmon2-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/perfmon2-devel
