1 2 # Note that you must do all the following as root (I believe). 3 # Although the program to be profiled can be run by anybody. 4 5 # start the profiler 6 opcontrol --stop ; opcontrol --reset ; opcontrol --callgraph=5 --start 7 8 # now run the program(s) to be profiled 9 10 # stop the profiler and dump results to .. um .. some file somewhere 11 opcontrol --stop ; opcontrol --dump 12 13 # get total cost 14 opreport --merge=tgid -x \ 15 /home/sewardj/VgTRUNK/atomk/Inst/lib/valgrind/helgrind-amd64-linux \ 16 | less 17 18 # produce a flat profile 19 opreport --merge=tgid --symbols -x \ 20 /home/sewardj/VgTRUNK/atomk/Inst/lib/valgrind/helgrind-amd64-linux \ 21 | less 22 23 # produce a profile w/ callgraph 24 opreport --merge=tgid --callgraph -x \ 25 /home/sewardj/VgTRUNK/atomk/Inst/lib/valgrind/helgrind-amd64-linux \ 26 | less 27 28 # print annotated source 29 opannotate --merge=tgid --source --include-file=libhb_core.c | less 30 31 32 #### notes. 33 34 1. on amd64, need to build V with -fno-omit-frame-pointer, else the 35 w/ callgraph profiles are useless. (oprofile doesn't do CFI based 36 stack unwinding, I guess). Add -fno-omit-frame-pointer to 37 AM_CFLAGS_BASE in Makefile.flags.am, and rebuild from clean. 38 39 2. even at the best of times the callgraph profiles seem pretty 40 flaky to me. 41 42 3. Even oprofile 0.9.4 (the latest) on amd64-linux doesn't work 43 for callgraph profiling. There is however a patch that 44 makes it work. See 45 46 http://sourceforge.net/tracker/index.php?func=detail&aid=1685267&group_id=16191&atid=116191 47 48 for details. Even then it sometimes fails at the "opcontrol 49 --dump" phase, complaining that the daemon died (or something like 50 that). But apart from that, it seems usable. 51
