uthash + this lastest patchset, compiled with -pg -O2. Staging 14k routes in 2048k chunks over about 120 sec. 12 core ubuntu 18 box netlink reports an error on every insert, but all the routes get there, and get out...
killing it shortly after staging things is quite lovely compared to what it was before this branch. ~5 sec of cpu time? (has recvmmsg ever been fixed?) I note (that as mr inliner), I also put in an inline qsort once... but not on this run. I'll go beat this up harder. Each sample counts as 0.01 seconds. % cumulative self self total time seconds seconds calls ms/call ms/call name 38.90 1.98 1.98 51898 0.04 0.05 netlink_read 18.08 2.90 0.92 kernel_route_compare 8.64 3.34 0.44 54528432 0.00 0.00 do_filter 8.25 3.76 0.42 check_xroutes 7.37 4.14 0.38 56484315 0.00 0.00 filter_route 3.93 4.34 0.20 53788418 0.00 0.00 redistribute_filter 3.83 4.53 0.20 112984285 0.00 0.00 martian_prefix 3.54 4.71 0.18 55569652 0.00 0.00 xroute_compare 2.55 4.84 0.13 165057 0.00 0.00 find_xroute_slot 0.79 4.88 0.04 738564 0.00 0.00 output_filter 0.69 4.92 0.04 wait_for_fd 0.59 4.95 0.03 738560 0.00 0.00 really_buffer_update 0.49 4.97 0.03 141157 0.00 0.00 filter_address 0.39 4.99 0.02 326535 0.00 0.00 find_route_slot 0.39 5.01 0.02 233227 0.00 0.00 update_route 0.39 5.03 0.02 1 20.00 20.00 getint 0.20 5.04 0.01 908969 0.00 0.00 timeval_minus_msec 0.20 5.05 0.01 746789 0.00 0.00 start_message _______________________________________________ Babel-users mailing list [email protected] https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/babel-users
