1:02 p.m.
Greeting,
FYI, we noticed a -22.1% regression of netperf.Throughput_tps due to commit:
commit: 106b167fed0c54a1f9db472c910979d65fcfd629 ("sched: force update of blocked
load of idle cpus")
git://internal_merge_and_test_tree devel-catchup-201710171950
in testcase: netperf
on test machine: qemu-system-x86_64 -enable-kvm -cpu host -smp 4 -m 5G
with following parameters:
ip: ipv4
runtime: 300s
nr_threads: 25%
cluster: cs-localhost
test: UDP_RR
test-description: Netperf is a benchmark that can be use to measure various aspect of
networking performance.
test-url: http://www.netperf.org/netperf/
Details are as below:
-------------------------------------------------------------------------------------------------->
To reproduce:
git clone https://github.com/intel/lkp-tests.git
cd lkp-tests
bin/lkp qemu -k <bzImage> job-script # job-script is attached in this
email
testcase/path_params/tbox_group/run: netperf/ipv4-300s-25%-cs-localhost-UDP_RR/vm-lkp-a04
6be53520ad8f87dc 106b167fed0c54a1f9db472c91
---------------- --------------------------
50 49 netperf.time.percent_of_cpu_this_job_got
6952 -22% 5414 netperf.Throughput_tps
2060002 -22% 1599840 netperf.time.voluntary_context_switches
15109 39% 21011 vmstat.system.in
28445 -21% 22564 vmstat.system.cs
80.65 4% 83.70 boot-time.boot
215 3% 222 boot-time.idle
netperf.Throughput_tps
7400 +-+------------------------------------------------------------------+
7200 +-+ .+ .+ .+ .+ |
|+ +.+ .++.+ .+++ ++ :.++ ++ ++ : .++ + |
7000 +-+ + ++ ++.+++.+++ + +.+ : :+.++ +. + + |
6800 +-++ ++.+ + + +|
6600 +-+ |
6400 +-+ |
| |
6200 +-+ |
6000 +-+ |
5800 +-+ O |
5600 +-+ O O O O |
O O OOO O O O O O OOO O |
5400 +O+OO O OO OO OO |
5200 +-+------------------------------------------------------------------+
[*] bisect-good sample
[O] bisect-bad sample
Disclaimer:
Results have been estimated based on internal Intel analysis and are provided
for informational purposes only. Any difference in system hardware or software
design or configuration may affect actual performance.
Thanks,
Xiaolong