With the help of some hardly documented switches in the pcm-pcie.x tool I found out that indeed the majority of my DDIO traffic is a miss.
pcm-pcie.x -e -B There is a huge variance - from 25% to over 90% of misses, but the average ratio of DDIO miss is over 60%. What could cause that? On Sat, Oct 26, 2019 at 6:00 PM Michał Purzyński <michalpurzyns...@gmail.com> wrote: > Yet another post, but I'm almost there and I promise to publish yet > another piece of documentation. > > Correct me if I'm wrong but reading the > > UNC_IIO_DATA_REQ_OF_CPU.MEM_WRITE.PART0 > > perf stat -e unc_iio_data_req_of_cpu.mem_write.part0 -C1 -r 3 sleep 1 > > should tell me what the rate of DDIO misses is (i.e. card wrote to memory > instead of the LLC). > > Now on a system with a dual Xeon Gold 6126 and 2x X710 (one per NUMA node) > this number is huge. Either I'm misinterpreting results or there's a large > number of DDIO misses - pretty much all of my traffic! > > The average traffic per port, at the time of measurement, was around > 250Mbit and the DDIO misses were like 215Mbit. > > Either DDIO does not work or I'm looking at the wrong counter. Either way > let me know. > > -- > Michal Purzynski > > _______________________________________________ E1000-devel mailing list E1000-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/e1000-devel To learn more about Intel Ethernet, visit https://forums.intel.com/s/topic/0TO0P00000018NbWAI/intel-ethernet
