Re: [LKP] [fs/locks] fd7732e033: stress-ng.eventfd.ops 21.9% improvement

From: J. Bruce Fields
Date: Tue Dec 11 2018 - 12:05:56 EST


On Tue, Dec 11, 2018 at 05:14:29PM +0800, kernel test robot wrote:
> FYI, we noticed a 21.9% improvement of stress-ng.eventfd.ops due to commit:

So, doesn't sound like a problem, but just looking into it out of
curiosity.....

I love that you do this, but these reports can be hard to figure out
sometimes. The graphs especially could use better labeling.

> commit: fd7732e033e30b3a586923b57e338c859e17858a ("fs/locks: create a tree of dependent requests.")
> https://git.kernel.org/cgit/linux/kernel/git/jlayton/linux.git locks-next
>
> in testcase: stress-ng
> on test machine: 88 threads Intel(R) Xeon(R) CPU E5-2699 v4 @ 2.20GHz with 64G memory
> with following parameters:
>
> nr_threads: 100%
> disk: 1HDD
> testtime: 1s
> class: filesystem
> ucode: 0xb00002e
> cpufreq_governor: performance

Looking at the man page for stress-ng, it doesn't say much about what it
actually does. It does say not to use stress-ng as a benchmark. I
can't tell if it does file locking--the man pages does document some
options which ask for locking (--flock, --locka, --lockf) but I don't
see any evidence in the yaml file or elsewhere that those options are in
use?

These graphs could really use units on the axes. Also:

> stress-ng.fcntl.ops
>
> 120000 +-+----------------------------------------------------------------+
> O O O O O O O O O O O O O O O O O O |
> 110000 +-O O O O O O O |
> 100000 +-+ |
> | |
> 90000 +-+ |
> 80000 +-+ |
> | |
> 70000 +-+ |
> 60000 +-+ |
> | |
> 50000 +-+ |
> 40000 +-+ |
> |. .+.+..+.+. .+.+..+.+. .+.+.+..+. .+. .+.. .+. .+. .+.|
> 30000 +-+----------------------------------------------------------------+
>
>
> stress-ng.fcntl.ops_per_sec
>
> 120000 +-+----------------------------------------------------------------+
> O O O O O O O O O O O O O O O O O O |
> 110000 +-O O O O O O O |
> 100000 +-+ |
> | |
> 90000 +-+ |
> 80000 +-+ |
> | |
> 70000 +-+ |
> 60000 +-+ |
> | |
> 50000 +-+ |
> 40000 +-+ |
> |. .+.+..+.+. .+.+..+.+. .+.+.+..+. .+. .+.. .+. .+. .+.|
> 30000 +-+----------------------------------------------------------------+
>
>
> [*] bisect-good sample
> [O] bisect-bad sample

I see O's, not but any *'s. And what are the +'s?

--b.