[OLSR-users] CPU load
aaron
(spam-protected)
Fri Jan 21 02:26:39 CET 2005
On Fri, 21 Jan 2005, ap wrote:
ok, since i did not bother now how to tell gprof how to still do a
gmon.out even if Ctrl-C is pressed, here is what i got using qprof:
Please note that it was running with -d 1 so the *fprintf statements are
probably not important
<<<< olsr.org - 0.4.8 - terminating >>>>
http://www.olsr.org
qprof: /home/funke/olsrd-0.4.8/olsrd: 513 samples, 513 counts
olsrd 2 (
0%)
olsr_hashing 1 (
0%)
if_ifwithaddr 1 (
0%)
lookup_link_status 1 (
0%)
get_neighbor_nexthop 4 (
1%)
get_interface_link_set 3 (
1%)
lookup_link_entry 3 (
1%)
olsr_print_link_set 1 (
0%)
olsr_neighbor_best_link 2 (
0%)
list_get_head 1 (
0%)
list_get_next 2 (
0%)
add_vertex 8 (
2%)
add_edge 35 (
7%)
free_everything 1 (
0%)
extract_best 1 (
0%)
olsr_calculate_lq_routing_table 1 (
0%)
mid_lookup_main_addr 50 (
10%)
mid_lookup_aliases 1 (
0%)
olsr_ip_to_string 2 (
0%)
olsr_printf 1 (
0%)
olsr_find_up_route 3 (
1%)
olsr_free_routing_table 1 (
0%)
olsr_insert_routing_table 1 (
0%)
olsr_calculate_hna_routes 1 (
0%)
libc.so.6 4 (
1%)
libc.so.6(__isinf) 1 (
0%)
libc.so.6(__isnan) 17 (
3%)
libc.so.6(_IO_vfprintf) 97 (
19%)
libc.so.6(__printf_fp) 41 (
8%)
libc.so.6(snprintf) 9 (
2%)
libc.so.6(vsprintf) 2 (
0%)
libc.so.6(vsnprintf) 7 (
1%)
libc.so.6(_IO_file_overflow) 1 (
0%)
libc.so.6(_IO_file_write) 2 (
0%)
libc.so.6(_IO_file_xsputn) 2 (
0%)
libc.so.6(_IO_setb) 3 (
1%)
libc.so.6(_IO_default_xsputn) 33 (
6%)
libc.so.6(__libc_free) 5 (
1%)
libc.so.6(strcpy) 2 (
0%)
libc.so.6(mempcpy) 3 (
1%)
libc.so.6(__write) 72 (
14%)
libc.so.6(ioctl) 2 (
0%)
libc.so.6(recvfrom) 1 (
0%)
libc.so.6(send) 4 (
1%)
libc.so.6(sendto) 2 (
0%)
libc.so.6(socket) 48 (
9%)
libc.so.6(mcount) 16 (
3%)
libc.so.6(inet_ntoa) 7 (
1%)
libpthread.so.0 2 (
0%)
libpthread.so.0(pthread_getspecific) 1 (
0%)
libpthread.so.0(__pthread_once) 1 (
0%)
libpthread.so.0(_pthread_cleanup_pop_restore) 1 (
0%)
this run was done on a i386 (fat SMP machine, where olsrd -d 0 causes 1.9
% CPU usage)
does that help?
ciao,
a.
>>>
>>> If this still doesn't make any difference, we should come up with a
>>> profiling version of olsrd. Has anyone ever used the profiling features
>>> of GCC? Are they supported by the cross-compiler? I guess that this
>>> would give us a good idea in which functions olsrd spends most of its
>>> CPU cycles.
>>>
>> how big would a gprof output become? I could probably NFS mount some share
>> and store it there...
>>
>
> i dont't think the results from the profiling will be different, whether you
> profile it on x86 or on mips. i recommend profiling on x86, it will be much
> easier.
>
> greetings
> andreas
> _______________________________________________
> olsr-users mailing list
> (spam-protected)
> https://www.olsr.org/mailman/listinfo/olsr-users
>
More information about the Olsr-users
mailing list