perf_event_tests icon indicating copy to clipboard operation
perf_event_tests copied to clipboard

Failures at ivb and haswell with v4.13-rc2/v4.13-rc3

Open zhijianli88 opened this issue 7 years ago • 2 comments

hi @deater , i tested perf_event, below are the failure results 2 failures at ivb

lizhijian@inn:/result/perf_event_tests/not_paranoid_at_all/lkp-ivb-d01/debian-x86_64-2016-08-31.cgz/x86_64-rhel-7.2/gcc-6/520eccdfe187591a51ea9ab4c1a024ae4d0f68d9/4$ grep -e FAILED perf_event_tests 
    Testing fixed counter 0 event...                           FAILED
    Testing multiple event overflow within group...            FAILED

1 failure at haswell

lizhijian@inn:/result/perf_event_tests/not_paranoid_at_all/lkp-hsw-ep4/debian-x86_64-2016-08-31.cgz/x86_64-rhel-7.2/gcc-6/520eccdfe187591a51ea9ab4c1a024ae4d0f68d9/3$ grep -e FAILED perf_event_tests
    Testing OpenMP overflow results...                         FAILED

all above results are tested under root.

are those failures possible related to hardware?

zhijianli88 avatar Aug 02 '17 03:08 zhijianli88

fixed_ctr0 details at ivb platform

root@lkp-ivb-d01 /lkp/benchmarks/perf_event_tests/tests/x86_intel# ./fixed_ctr0
This test checks the intel fixed counter 0
Which is retired instructions.
Linux treats/schedules this the same as in
a general purpose counter even though the results
are not always equivelent.
There is no way to specify which of the two you want

Ideally, all the values should match.

0 1000015, 1 1000016, 2 1000016, 3 1000016, 4 1000016, 
Some results don't match!
0 1000014, 1 1000014, 2 1000014, 3 1000014, 4 1000014, 
0 1000015, 1 1000016, 2 1000016, 3 1000016, 4 1000016, 
Some results don't match!
0 1000014, 1 1000014, 2 1000014, 3 1000014, 4 1000014, 
0 1000014, 1 1000014, 2 1000014, 3 1000014, 4 1000014, 
0 1000015, 1 1000016, 2 1000016, 3 1000016, 4 1000016, 
Some results don't match!
0 1000014, 1 1000014, 2 1000014, 3 1000014, 4 1000014, 
0 1000014, 1 1000014, 2 1000014, 3 1000014, 4 1000014, 
0 1000015, 1 1000016, 2 1000016, 3 1000016, 4 1000016, 
Some results don't match!
0 1000014, 1 1000014, 2 1000014, 3 1000014, 4 1000014, 
0 1000015, 1 1000016, 2 1000016, 3 1000016, 4 1000016, 
Some results don't match!
0 1000014, 1 1000014, 2 1000014, 3 1000014, 4 1000014, 
0 1000014, 1 1000014, 2 1000014, 3 1000014, 4 1000014, 
0 1000015, 1 1000016, 2 1000016, 3 1000016, 4 1000016, 
Some results don't match!
0 1000014, 1 1000014, 2 1000014, 3 1000014, 4 1000014, 
0 1000014, 1 1000014, 2 1000014, 3 1000014, 4 1000014, 
0 1000015, 1 1000016, 2 1000016, 3 1000016, 4 1000016, 
Some results don't match!
0 1000014, 1 1000014, 2 1000014, 3 1000014, 4 1000014, 
0 1000015, 1 1000016, 2 1000016, 3 1000016, 4 1000016, 
Some results don't match!
0 1000014, 1 1000014, 2 1000014, 3 1000014, 4 1000014, 
Testing fixed counter 0 event...                           FAILED

zhijianli88 avatar Aug 03 '17 05:08 zhijianli88

fixtests/overflow/simultaneous_group_overflow details at ivb platform

root@lkp-ivb-d01 /lkp/benchmarks/perf_event_tests# tests/overflow/simultaneous_group_overflow
This tests simultaneous overflow within group.
Testing matrix matrix multiply
	Event perf::instructions with period 1000000
	Matrix multiply sum: s=27665734022509.746094
	fd 3 overflows: 946 (perf::instructions/1000000)
Testing matrix matrix multiply
	Event perf::instructions with period 2000000
	Matrix multiply sum: s=27665734022509.746094
	fd 3 overflows: 473 (perf::instructions/2000000)
Testing matrix matrix multiply
	Event perf::instructions with period 1000000
	Event perf::instructions with period 2000000
	Matrix multiply sum: s=27665734022509.746094
	fd 3 overflows: 946 (perf::instructions/1000000)
	fd 4 overflows: 473 (perf::instructions/2000000)
	Count 0: 946375550
	Count 1: 946378571
Event perf::instructions/1000000 Expected 946 Got 946
Event perf::instructions/2000000 Expected 473 Got 473
Counts should be the same 946375550 946378571
Testing multiple event overflow within group...            FAILED

zhijianli88 avatar Aug 03 '17 05:08 zhijianli88