Re: [RESEND][PATCH v8 0/7] Preparatory changes for Proxy Execution v8

[K Prateek Nayak]

Hello John,

On 2/29/2024 11:49 AM, John Stultz wrote:
> On Wed, Feb 28, 2024 at 9:37 AM 'K Prateek Nayak' via kernel-team
> <kernel-team@xxxxxxxxxxx> wrote:
>> I got a chance to test the whole of v8 patches on the same dual socket
>> 3rd Generation EPYC system:
>>
>> tl;dr
>>
>> - There is a slight regression in hackbench but instead of the 10x
>>   blowup seen previously, it is only around 5% with overloaded case
>>   not regressing at all.
>>
>> - A small but consistent (~2-3%) regression is seen in tbench and
>>   netperf.
> 
> Once again, thank you so much for your testing and reporting of the
> data! I really appreciate it!
> 
> Do you mind sharing exactly how you're running the benchmarks? (I'd
> like to try to reproduce these locally (though my machine is much
> smaller).
> 
> I'm guessing the hackbench one is the same command you shared earlier with v6?

Yup it is same as earlier. I'll list all the commands down below:

o Hackbench

	perf bench sched messaging -p -t -l 100000 -g <# of groups>

o Old schbench
  git://git.kernel.org/pub/scm/linux/kernel/git/mason/schbench.git
  at commit e4aa540 ("Make sure rps isn't zero in auto_rps mode.")

	schbench -m 2 -t <# workers> -r 30

  (I should probably upgrade this to the latest! Let me get on it)

o tbench (https://www.samba.org/ftp/tridge/dbench/dbench-4.0.tar.gz)

	nohup tbench_srv 0 &
	tbench -c client.txt -t 60 <# clients> 127.0.0.1

o Stream (https://www.cs.virginia.edu/stream/FTP/Code/)

	export ARRAY_SIZE=128000000; # 4 * Local L3 size
	gcc -DSTREAM_ARRAY_SIZE=$ARRAY_SIZE -DNTIMES=<Loops internally> -fopenmp -O2 stream.c -o stream
	export OMP_NUM_THREADS=16; # Number of CCX on my machine
	./stream;

o netperf

	netserver -L 127.0.0.1
	for i in `seq 0 1 <num clients>`;
	do
		netperf -H 127.0.0.1 -t TCP_RR -l 100 -- -r 100 -k REQUEST_SIZE,RESPONSE_SIZE,ELAPSED_TIME,THROUGHPUT,THROUGHPUT_UNITS,MIN_LATENCY,MEAN_LATENCY,P50_LATENCY,P90_LATENCY,P99_LATENCY,MAX_LATENCY,STDDEV_LATENCY&
	done
	wait;

o Unixbench (from mmtest)

	./run-mmtests.sh --no-monitor --config configs/config-workload-unixbench
--

If you have any other question, please do let me know :)

> 
> thanks
> -john
 
--
Thanks and Regards,
Prateek