+1, that’s exactly what you’re seeing… D. From: vpp-dev@lists.fd.io <vpp-dev@lists.fd.io> On Behalf Of Brian Brooks Sent: Wednesday, May 9, 2018 11:00 AM To: Saxena, Nitin <nitin.sax...@cavium.com>; vpp-dev <vpp-dev@lists.fd.io> Subject: Re: [vpp-dev] vlib_main() taking maximum cycle in v1804 release
I think this is due to perf aggregating results from all CPUs, and that vlib_main hotspot is from the main (non-worker) thread. From: vpp-dev@lists.fd.io<mailto:vpp-dev@lists.fd.io> <vpp-dev@lists.fd.io<mailto:vpp-dev@lists.fd.io>> On Behalf Of Saxena, Nitin Sent: Thursday, April 26, 2018 1:46 PM To: vpp-dev <vpp-dev@lists.fd.io<mailto:vpp-dev@lists.fd.io>> Subject: [vpp-dev] vlib_main() taking maximum cycle in v1804 release Hi, While running L3 Fwd test case with VPP v1804 release (released yesterday), perf tool shows vlib_main() taking maximum cycles. This has been observed on Aarch64 (Both on Marvell’s Macchiatobin and Cavium’s ThunderX products) and Broadwell machine as well. Perf top from Broadwell machine ========================= 29.72% libvlib.so.0.0.0 [.] vlib_main 17.33% dpdk_plugin.so [.] i40e_xmit_pkts 7.35% dpdk_plugin.so [.] i40e_recv_scattered_pkts_vec 7.21% libvlibmemory.so.0.0.0 [.] memclnt_queue_callback 6.02% libvnet.so.0.0.0 [.] ip4_rewrite_avx2 5.96% libvnet.so.0.0.0 [.] ip4_lookup_avx2 5.69% libvnet.so.0.0.0 [.] ip4_input_no_checksum_avx2 5.49% dpdk_plugin.so [.] dpdk_input_avx2 3.40% dpdk_plugin.so [.] dpdk_interface_tx_avx2 1.63% libvnet.so.0.0.0 [.] vnet_interface_output_node_avx2 Is this the expected behaviour? I haven’t observed vlib_main() taking max perf cycles in VPP v1801. Thanks, Nitin IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you.
