Hi,
I use pfmon3.9/libpfm3.9,
I want to collect the number of L2 misses every million instructions.
But I got two errors.
First, LAST_LEVEL_CACHE_MISSES was not counted at all !!! pfmon outputs zero
for the second counting event.
Secondly, when tasks was cloned , pfmon encounters lack of memory error. It
was colored red in the following.
Anybody can help me fix this?
Thanks.
[r...@localhost tests]# pfmon --follow-all --verbose
-eINSTRUCTIONS_RETIRED,LAST_LEVEL_CACHE_MISSES --smpl-show-top=20
--long-smpl-periods=1000000 --
../../../../parsec-2.1/pkgs/kernels/canneal/inst/amd64-linux.gcc/bin/canneal
4 15000 2000 ../../../../parsec-2.1/pkgs/kernels/canneal/inputs/400000.nets
128
measuring at user privilege level ONLY
1 event set(s) defined
long sampling periods(val/mask/seed): 1000000/0x0/00/0x0/0
short sampling periods(val/mask/seed): 1000000/0x0/00/0x0/0
using inst-hist sampling module
unavailable_pmcs=0xfffffffffffcfffc
unavailable_pmds=0xfffffffffff8fffc
[FIXED_CTRL(pmc16)=0xa pmi0=1 en0=0x2 pmi1=1 en1=0x0 pmi2=1 en2=0x0]
INSTRUCTIONS_RETIRED
[FIXED_CTR0(pmd16)]
[PERFEVTSEL0(pmc0)=0x51412e event_sel=0x2e umask=0x41 os=0 usr=1 en=1 int=1
inv=0 edge=0 cnt_mask=0] LAST_LEVEL_CACHE_MISSES
[PMC0(pmd0)]
pmd setup for event set0:
[pmd16 set=0 smpl_pmds=0x1 reset_pmds=0x0]
[pmd16 set=0 ival=0xfffffffffff0bdc0 long_rate=0xfffffffffff0bdc0
short_rate=0xfffffffffff0bdc0 mask=0x0 seed=0 randomize=n]
[pmd0 set=0 ival=0x0 long_rate=0x0 short_rate=0x0 mask=0x0 seed=0
randomize=n]
max=65536 cur=65536
locked_mem_avail=65536 pfm_mem_avail=4152639488 mem_avail=65376 ncpus=0
max_entries=1362 smpl_entries=1362 orig_smpl_entries=0 pgsz=4096 hdrsz=128
entrysz=48 slack=32
sampling buffer #entries=1362 size=65536, max_entry_size=48
vCPU0 -> pCPU0
vCPU1 -> pCPU1
vCPU3 -> pCPU3
vCPU2 -> pCPU2
vCPU4 -> pCPU4
vCPU5 -> pCPU5
vCPU6 -> pCPU6
exec-pattern=*
vCPU7 -> pCPU7
[7182] started task:
../../../../parsec-2.1/pkgs/kernels/canneal/inst/amd64-linux.gcc/bin/canneal
4 15000 2000 ../../../../parsec-2.1/pkgs/kernels/canneal/inputs/400000.nets
128
follow_exec=y follow_vfork=y follow_fork=y follow_pthread=y
[7182] 64-bit binary
[7182] monitoring
/home/hll/parsec-2.1/pkgs/kernels/canneal/inst/amd64-linux.gcc/bin/canneal...
[7182] activating monitoring
measurements started at Sat Jul 17 06:29:05 2010
PARSEC Benchmark Suite Version 2.1
Threadcount: 4
15000 swaps per temperature step
start temperature: 2000
netlist filename:
../../../../parsec-2.1/pkgs/kernels/canneal/inputs/400000.nets
number of temperature steps: 128
locs created
locs assigned
Just saw element: 100000
Just saw element: 200000
Just saw element: 300000
Just saw element: 400000
netlist created. 400000 elements.
[7183] out-of-order creation, stopped
[7182] cloned [7183]
[7183] monitoring
/home/hll/parsec-2.1/pkgs/kernels/canneal/inst/amd64-linux.gcc/bin/canneal...
Not enough memory to create perfmon context for [7183],
check your locked memory resource limit with limit or ulimit
session terminated by unrecoverable error
sending signal 9 to [7182]
sending signal 9 to [7183]
[7183] task exited
[7183] detached
[7182] task exited
[7182] detached
sdesc chain to process:
[7182]
(/home/hll/parsec-2.1/pkgs/kernels/canneal/inst/amd64-linux.gcc/bin/canneal)
[7183]
(/home/hll/parsec-2.1/pkgs/kernels/canneal/inst/amd64-linux.gcc/bin/canneal)
# results for [7182:7182<-[7173]]
(/home/hll/parsec-2.1/pkgs/kernels/canneal/inst/amd64-linux.gcc/bin/canneal)
# total samples : 4370
# total buffer overflows : 3
#
# event00 event01
# counts %self %cum counts %self %cum code addr
368 8.42% 8.42% 0 0.00% 0.00% 0x000000000040557c
256 5.86% 14.28% 0 0.00% 0.00% 0x0000000000405592
188 4.30% 18.58% 0 0.00% 0.00% 0x0000000000405590
146 3.34% 21.92% 0 0.00% 0.00% 0x00007f50f9fc8b64
145 3.32% 25.24% 0 0.00% 0.00% 0x0000000000405586
112 2.56% 27.80% 0 0.00% 0.00% 0x00000034bb277aa2
107 2.45% 30.25% 0 0.00% 0.00% 0x00007f50fa00742d
85 1.95% 32.20% 0 0.00% 0.00% 0x000000000040b600
80 1.83% 34.03% 0 0.00% 0.00% 0x000000000040616d
63 1.44% 35.47% 0 0.00% 0.00% 0x00007f50f9fc97c7
59 1.35% 36.82% 0 0.00% 0.00% 0x000000000040b602
38 0.87% 37.69% 0 0.00% 0.00% 0x00000034bb26f5a5
38 0.87% 38.56% 0 0.00% 0.00% 0x00007f50f9fef5f8
37 0.85% 39.41% 0 0.00% 0.00% 0x00007f50f9fde21c
37 0.85% 40.25% 0 0.00% 0.00% 0x00000000004055d7
36 0.82% 41.08% 0 0.00% 0.00% 0x00000034bb277aa0
35 0.80% 41.88% 0 0.00% 0.00% 0x0000000000407d05
33 0.76% 42.63% 0 0.00% 0.00% 0x00000034bb270c8b
31 0.71% 43.34% 0 0.00% 0.00% 0x00000034bb26f599
29 0.66% 44.00% 0 0.00% 0.00% 0x00000000004060fa
created tasks : 0
maximum tasks : 2
maximum active tasks : 1
measurements completed at Sat Jul 17 06:29:15 2010
------------------------------------------------------------------------------
This SF.net email is sponsored by Sprint
What will you do first with EVO, the first 4G phone?
Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first
_______________________________________________
perfmon2-devel mailing list
perfmon2-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/perfmon2-devel
