Hi Nagy
You seem to compare two things which are not compatible.
In first instance by using time you measured the time spent by the program in
user and kernel mode.
While in second instance you are doing sampling, that too in kernel mode only.
Regards,
JK
--- On Sun, 19/4/09, Nagy Mostafa <nagy.most...@gmail.com> wrote:
From: Nagy Mostafa <nagy.most...@gmail.com>
Subject: [perfmon2] Strangeness
To: perfmon2-devel@lists.sourceforge.net
Date: Sunday, 19 April, 2009, 12:35 PM
Hi,
I have a question about something strange I have noticed while
collecting profiles using pfmon. When I run a single cold run of the 2to3
benchmark, the
"time" command (or using --show-time) gives me:
$time /usr/local/bin/python -E -O benchmarks/unladen-swallow/lib/2to3/2to3 -f
all benchmarks/unladen-swallow/lib/2to3 > /dev/null
real 0m19.523s user 0m19.485s sys 0m0.044s
which shows dominating user time (actually system time is negligible). Now when
running pfmon to do
sampling every 10000 UNHALTED_CORE_CYCLES (or INSTRUCTIONS_RETIRED), the
results show dominating kernel time 87%
(see below). I am just wondering why there is such a difference. Am I missing
something ?
pfmon
--follow-all -k --long-smpl-period=100000 --resolve --smpl-per-function
--smpl-outfile=pfmon_smpl /usr/local/bin/python -E -O
benchmarks/unladen-swallow/lib/2to3/2to3 -f all
benchmarks/unladen-swallow/lib/2to3 > /dev/null
# results for [19867:19867<-[19864]] (/usr/local/bin/python)
# total samples : 1529
# total buffer overflows : 1
#
# event00
# counts %self %cum code addr symbol
360 23.54% 23.54% 0xffffffff80360f40 clear_page_c<kernel>
252 16.48% 40.03% 0xffffffff80247160 __do_softirq<kernel>
109 7.13% 47.16% 0xffffffff8028eaf0 get_page_from_freelist<kernel>
65 4.25% 51.41% 0xffffffff802c9f60 __d_lookup<kernel>
56 3.66% 55.07% 0xffffffff80477ef0 _spin_unlock_irqrestore<kernel>
45 2.94% 58.01% 0xffffffff80479bc0 do_page_fault<kernel>
41 2.68% 60.69% 0xffffffff80477dd0 _spin_lock<kernel>
39 2.55% 63.24% 0xffffffff802c0ae0 __link_path_walk<kernel>
36 2.35% 65.60% 0xffffffff80298b80 unmap_vmas<kernel>
36 2.35% 67.95% 0xffffffff8028dc20 free_hot_cold_page<kernel>
22 1.44% 69.39% 0x00007f65c75d0000 UNKNOWN</lib/libc-2.7.so>
20 1.31% 70.70% 0xffffffff802999f0 handle_mm_fault<kernel>
20 1.31% 72.01% 0xffffffff80292030 __pagevec_lru_add_active<kernel>
17 1.11% 73.12% 0xffffffff8024afb0 run_timer_softirq<kernel>
13 0.85% 73.97% 0xffffffff8035a540 _atomic_dec_and_lock<kernel>
13 0.85% 74.82% 0xffffffff8028f390 __alloc_pages<kernel>
13 0.85% 75.67% 0xffffffff8028c5f0 page_to_pfn<kernel>
12 0.78% 76.46% 0xffffffff803619a0 __strncpy_from_user<kernel>
12 0.78% 77.24% 0xffffffff802a1e90 anon_vma_prepare<kernel>
12 0.78% 78.02% 0xffffffff80292860 lru_cache_add_active<kernel>
11 0.72% 78.74% 0xffffffff80361240 copy_user_generic_string<kernel>
11 0.72% 79.46% 0xffffffff802b0a60 kmem_cache_alloc<kernel>
11 0.72% 80.18% 0xffffffff80276a60
cpuset_update_task_memory_state<kernel>
9 0.59% 80.77% 0xffffffff88161a90
ext3_count_free_blocks[ext3]<kernel>
9 0.59% 81.36% 0xffffffff802a7b20 alloc_page_vma<kernel>
9 0.59% 81.95% 0xffffffff802a1080 page_remove_rmap<kernel>
9 0.59% 82.54% 0xffffffff80291e70 release_pages<kernel>
9 0.59% 83.13% 0xffffffff8028c5a0 pfn_to_page<kernel>
9 0.59% 83.71% 0xffffffff8023c090 finish_task_switch<kernel>
8 0.52% 84.24% 0xffffffff8029ce40 find_vma<kernel>
7 0.46% 84.70% 0xffffffff802c1c00 do_path_lookup<kernel>
7 0.46% 85.15% 0xffffffff802a6b20 get_vma_policy<kernel>
7 0.46% 85.61% 0xffffffff8028c350 zone_watermark_ok<kernel>
6 0.39% 86.00% 0xffffffff802be9f0 __follow_mount<kernel>
6 0.39% 86.40% 0xffffffff802be780 permission<kernel>
6 0.39% 86.79% 0xffffffff802be670 generic_permission<kernel>
6 0.39% 87.18% 0xffffffff802965d0 __inc_zone_state<kernel>
6 0.39% 87.57% 0x000000000048fee0
PyEval_EvalFrameEx</usr/local/bin/python>
thanks,
- nagy
-----Inline Attachment Follows-----
------------------------------------------------------------------------------
Stay on top of everything new and different, both inside and
around Java (TM) technology - register by April 22, and save
$200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco.
300 plus technical and hands-on sessions. Register today.
Use priority code J9JMT32. http://p.sf.net/sfu/p
-----Inline Attachment Follows-----
_______________________________________________
perfmon2-devel mailing list
perfmon2-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/perfmon2-devel
Cricket on your mind? Visit the ultimate cricket website. Enter
http://beta.cricket.yahoo.com------------------------------------------------------------------------------
Stay on top of everything new and different, both inside and
around Java (TM) technology - register by April 22, and save
$200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco.
300 plus technical and hands-on sessions. Register today.
Use priority code J9JMT32. http://p.sf.net/sfu/p
_______________________________________________
perfmon2-devel mailing list
perfmon2-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/perfmon2-devel
