Re: [PATCH v3] mm: optimize stat output for 11% sys time reduce

[Andrew Morton <akpm@linux-foundation.org>]

On Thu, 22 Jan 2026 19:42:42 +0800 Jianyue Wu <wujianyue000@gmail.com> wrote:

> Replace seq_printf/seq_buf_printf with lightweight helpers to avoid
> printf parsing in memcg stats output.
> 
> Key changes:
> - Add memcg_seq_put_name_val() for seq_file "name value\n" formatting
> - Add memcg_seq_buf_put_name_val() for seq_buf "name value\n" formatting
> - Update __memory_events_show(), swap_events_show(),
>   memory_stat_format(), memory_numa_stat_show(), and related helpers
> - Introduce local variables to improve readability and reduce line length
> 
> Performance:
> - 1M reads of memory.stat+memory.numa_stat
> - Before: real 0m9.663s, user 0m4.840s, sys 0m4.823s
> - After:  real 0m9.051s, user 0m4.775s, sys 0m4.275s (~11.4% sys drop)

So the tl;dr here is "vfprintf() is slow". 

It's quite a large change, although not a complex one.

Do we need to change so much?  Would some subset of these changes
provide most of the benefit?

It does rather uglify things so there's a risk that helpful people will
send "cleanups" which switch back to using *printf*.  Explanatory code
comments would help prevent that but we'd need a lot of them.

I dunno, what do people think?  Does the benefit justify the change?

> Tests:
> - Script:
>   for ((i=1; i<=1000000; i++)); do
>       : > /dev/null < /sys/fs/cgroup/memory.stat
>       : > /dev/null < /sys/fs/cgroup/memory.numa_stat
>   done