Re: [sched/fair] d8fcb81f1a: netperf.Throughput_tps -16.9% regression

[Mel Gorman <mgorman@suse.de>]

On Sun, Nov 01, 2020 at 04:24:52PM +0800, kernel test robot wrote:
> Greeting,
> 
> FYI, we noticed a -16.9% regression of netperf.Throughput_tps due to commit:
> 
> 
> commit: d8fcb81f1acf651a0e50eacecca43d0524984f87 ("sched/fair: Check for idle core in wake_affine")
> https://git.kernel.org/cgit/linux/kernel/git/tip/tip.git sched/core
> 
> 
> in testcase: netperf
> on test machine: 144 threads Intel(R) Xeon(R) Gold 5318H CPU @ 2.50GHz with 128G memory
> with following parameters:
> 
> 	ip: ipv4
> 	runtime: 300s
> 	nr_threads: 25%
> 	cluster: cs-localhost
> 	test: SCTP_RR
> 	cpufreq_governor: performance
> 	ucode: 0x700001c
> 
> test-description: Netperf is a benchmark that can be use to measure various aspect of networking performance.
> test-url: http://www.netperf.org/netperf/
> 
> In addition to that, the commit also has significant impact on the following tests:
> 
> +------------------+------------------------------------------------------------------------+
> | testcase: change | aim7: aim7.jobs-per-min 8.7% improvement                               |
> | test machine     | 144 threads Intel(R) Xeon(R) Gold 5318H CPU @ 2.50GHz with 128G memory |
> | test parameters  | cpufreq_governor=performance                                           |
> |                  | disk=1BRD_48G                                                          |
> |                  | fs=xfs                                                                 |
> |                  | load=600                                                               |
> |                  | test=sync_disk_rw                                                      |
> |                  | ucode=0x700001c                                                        |
> +------------------+------------------------------------------------------------------------+
> 
> If you fix the issue, kindly add following tag
> Reported-by: kernel test robot <rong.a.chen@intel.com>
> 

I know I ran tests against this patch in general and gave an Ack but
these cases were not covered by me. For netperf, I neither vary the number
of threads nor did I cover SCTP_RR so this could be an example of where
the patch is a loss. I also had run reaim but not specifically for one
sub-test like this does and the generation of machines used was much
older than Gold 5318H. Grid or no grid, complete coverage is a challenge

-- 
Mel Gorman
SUSE Labs