Re: [x86, paravirt] fd6f48529f: aim7.jobs-per-min -26.1% regression

[Huang, Ying <ying.huang@intel.com>]

[-- Attachment #1: Type: text/plain, Size: 1665 bytes --]

Peter Zijlstra <peterz@infradead.org> writes:

> On Thu, Dec 01, 2016 at 02:07:29AM +0800, kernel test robot wrote:
>> 
>> Greeting,
>> 
>> FYI, we noticed a -26.1% regression of aim7.jobs-per-min due to commit:
>> 
>> 
>> commit: fd6f48529f4067c6bfbead88d3d7d117d51150f6 ("x86,paravirt: Optimize native pv_lock_ops.vcpu_is_preempted")
>> https://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git locking/core
>> 
>> in testcase: aim7
>> on test machine: qemu-system-x86_64 -enable-kvm -cpu host -smp 4 -m 5G
>> with following parameters:
>> 
>> 	test: link_test
>> 	load: 2000
>> 
>> test-description: AIM7 is a traditional UNIX system level benchmark
>> suite which is used to test and measure the performance of multiuser
>> system.
>> test-url: https://sourceforge.net/projects/aimbench/files/aim-suite7/
>> 
>
> OK, so I'm having a very hard time reproducing this. How can I run this
> link_test stuff when I got aim7 compiled? This is all impenetrable gunk
> to me.
>

Sorry, the reproduce script wasn't attached in the original email, here
it is:

echo "500 32000 128 512" > /proc/sys/kernel/sem
cat > workfile <<EOF
FILESIZE: 1M
POOLSIZE: 10M
10 link_test
EOF
mkdir -p /fs/shm
umount /fs/shm
mount -t tmpfs /dev/shm /fs/shm
echo "/fs/shm" > config

	(
		echo vm-lkp-a05-1
		echo link_test

		echo 1
		echo 2000
		echo 2
		echo 2000
		echo 1
	) | ./multitask -t

>
> Also, do you see the same problem with the upstream commit?
>
>   3cded4179481 ("x86/paravirt: Optimize native pv_lock_ops.vcpu_is_preempted()")

Sure.  I will test this commit and its parent.

Best Regards,
Huang, Ying