[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20170822210043.GB28715@tassilo.jf.intel.com>
Date: Tue, 22 Aug 2017 14:00:43 -0700
From: Andi Kleen <ak@...ux.intel.com>
To: Jiri Olsa <jolsa@...hat.com>
Cc: Andi Kleen <andi@...stfloor.org>, acme@...nel.org,
jolsa@...nel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 05/19] perf, tools: Support weak groups
> also if I put 'cycles' instead of the l2_rqsts.all_code_rd,
> I get clean open but 'not counted' as result.. I wonder
> there's some counter scheduling issue
>
> [root@...va perf]# ./perf stat -v -e '{branches,branch-misses,l1d.replacement,l2_lines_in.all,cycles}:W' -a sleep 1
> Using CPUID GenuineIntel-6-3D
> l1d.replacement -> cpu/umask=0x1,period=2000003,event=0x51/
> l2_lines_in.all -> cpu/umask=0x7,period=100003,event=0xf1/
> branches: 0 4004293853 0
> branch-misses: 0 4004293853 0
> l1d.replacement: 0 4004293853 0
> l2_lines_in.all: 0 4004293853 0
> cycles: 0 4004293853 0
>
> Performance counter stats for 'system wide':
>
> <not counted> branches (0.00%)
> <not counted> branch-misses (0.00%)
> <not counted> l1d.replacement (0.00%)
> <not counted> l2_lines_in.all (0.00%)
> <not counted> cycles (0.00%)
>
> 1.001088589 seconds time elapsed
Cannot reproduce on my system
Does it work without group or a normal group?
./perf stat -v -e '{branches,branch-misses,l1d.replacement,l2_lines_in.all,cycles}:W'
Using CPUID GenuineIntel-6-3E
l1d.replacement -> cpu/umask=0x1,period=2000003,event=0x51/
l2_lines_in.all -> cpu/umask=0x7,period=100003,event=0xf1/
^Cbranches: 105004611 27393959413 27393959413
branch-misses: 5854172 27393959413 27393959413
l1d.replacement: 13935607 27393959413 27393959413
l2_lines_in.all: 24987238 27393959413 27393959413
cycles: 1259015810 27393959413 27393959413
Performance counter stats for 'system wide':
105,004,611 branches
5,854,172 branch-misses # 5.58% of all branches
13,935,607 l1d.replacement
24,987,238 l2_lines_in.all
1,259,015,810 cycles
2.282966609 seconds time elapsed
Powered by blists - more mailing lists