Menu
▾
▴
Re: Multi-Core Output
|
From: Maynard J. <may...@us...> - 2010-08-17 18:08:13
|
Craig Masley wrote: > Hello, > > I recently ran some tests on a quad core machine. I received my output, and > there were two sample counts and two sample percentages. I did some reading > and realized that the duplication was due to having multiple cores. However > my question is why am I only seeing two if I have 4 cores? Craig, do 'cat /proc/cpuinfo' to see the number of processors configured on your system. I'm guessing it must be '2', since that's all the processors oprofile seems to know about. I'm not well versed in Intel architectures, but I suspect there's some configuration setting that is limiting the online processors. Another thing to check is /sys/devices/system/cpu. How many 'cpu<x>' entries do you see there? If you see more than two, then something is offline'ing some of your processors. Do 'cat /sys/devices/system/cpu/cpu<x>/online' to see the online status of each. By the way, from your output below, I can see you're running with either "--separate=cpu" or "--separate=all". Unless you have a particular need to see the profiling data split out per cpu, I wouldn't recommend that setting. If you don't separate by cpu, then you'll get one aggregate sample column and one aggregate percentage column. > > And excerpt from the output can be seen below. > > Second question is what does the no-vmlinux symbol represent? Before starting oprofile, you're required to either explicitly point at a vmlinux file using the '--vmlinux' option (for purposes of matching kernel sample addresses with symbols) or to say you don't care about kernel symbols, using the '--no-vmlinux' option. If you choose the latter, all kernel samples are simply dumped into a "no-vmlinux" bucket and no attempt is made to do address-to-symbol resolution. -Maynard > > Thank you, > > -Craig > ---------------------------------------------------------------------------------------------------------------------------------------------- > > CPU: Core 2, speed 1995.01 MHz (estimated) > Counted CPU_CLK_UNHALTED events (Clock cycles when not halted) with a unit > mask of 0x00 (Unhalted core cycles) count 100000 > Counted CPU_CLK_UNHALTED events (Clock cycles when not halted) with a unit > mask of 0x00 (Unhalted core cycles) count 90000 > samples % samples % image name app > name symbol name > ------------------------------------------------------------------------------- > 4081393 26.6121 0 0 no-vmlinux > no-vmlinux /no-vmlinux > 4081393 100.000 0 0 no-vmlinux > no-vmlinux /no-vmlinux [self] > ------------------------------------------------------------------------------- > 967582 6.3090 866641 0.9194 libpthread-2.5.so > libpthread-2.5.so pthread_mutex_lock > 967582 100.000 866641 100.000 libpthread-2.5.so > libpthread-2.5.so pthread_mutex_lock [self] > ------------------------------------------------------------------------------- > 862616 5.6245 698762 0.7413 libpthread-2.5.so > libpthread-2.5.so pthread_mutex_lock > 862616 100.000 698762 100.000 libpthread-2.5.so > libpthread-2.5.so pthread_mutex_lock [self] > ------------------------------------------------------------------------------- > 671114 4.3759 509441 0.5405 iperf > iperf /vz/root/1000/usr/bin/iperf > 671114 100.000 509441 100.000 iperf > iperf /vz/root/1000/usr/bin/iperf [self] > ------------------------------------------------------------------------------- > 239 0.3268 204 0.0069 libc-2.5.so > libc-2.5.so > __default_morecore > 72895 99.6732 2967914 99.9931 libc-2.5.so > libc-2.5.so > __libc_start_main > 657403 4.2865 2968765 3.1496 libc-2.5.so > libc-2.5.so > __read_nocancel > 657403 100.000 2968765 100.000 libc-2.5.so > libc-2.5.so > __read_nocancel [self] > ------------------------------------------------------------------------------- > 590521 3.8504 3353858 3.5581 vmlinux > vmlinux system_call > 590521 100.000 3353858 100.000 vmlinux > vmlinux system_call [self] > ------------------------------------------------------------------------------- > 505871 3.2985 397910 0.4221 libpthread-2.5.so > libpthread-2.5.so __pthread_mutex_unlock_usercnt > 505871 100.000 397910 100.000 libpthread-2.5.so > libpthread-2.5.so __pthread_mutex_unlock_usercnt [self] > ------------------------------------------------------------------------------- > 452129 2.9480 369596 0.3921 libpthread-2.5.so > libpthread-2.5.so __pthread_mutex_unlock_usercnt > 452129 100.000 369596 100.000 libpthread-2.5.so > libpthread-2.5.so __pthread_mutex_unlock_usercnt [self] > ------------------------------------------------------------------------------- > 451611 2.9447 3256947 3.4553 pktm.ko > pktm.ko xbuf_maxbytes > 451611 100.000 3256947 100.000 pktm.ko > pktm.ko xbuf_maxbytes [self] > ------------------------------------------------------------------------------- > 414926 2.7055 13970437 14.8212 pktd.ko > pktd.ko conduit_slab_free > 414926 100.000 13970437 100.000 pktd.ko > pktd.ko conduit_slab_free [self] > ------------------------------------------------------------------------------- > 370588 2.4164 316116 0.3354 iperf > iperf reporter_spawn > 370588 100.000 316116 100.000 iperf > iperf reporter_spawn [self] > ------------------------------------------------------------------------------- > 341811 2.2287 265764 0.2819 vmlinux > vmlinux schedule > 341811 100.000 265764 99.9996 vmlinux > vmlinux schedule [self] > 0 0 1 3.8e-04 vmlinux > vmlinux main_timer_handler > ------------------------------------------------------------------------------- > > > > > ------------------------------------------------------------------------------ > This SF.net email is sponsored by > > Make an app they can't live without > Enter the BlackBerry Developer Challenge > http://p.sf.net/sfu/RIM-dev2dev > > > > _______________________________________________ > oprofile-list mailing list > opr...@li... > https://lists.sourceforge.net/lists/listinfo/oprofile-list |