* Re: [PATCH] vsprintf: uninline simple_strntoull(), reorder arguments
[not found] <82a2af6e-9b6c-4a09-89d7-ca90cc1cdad1@p183>
@ 2023-10-30 8:59 ` Andy Shevchenko
2023-11-01 16:48 ` Petr Mladek
1 sibling, 0 replies; 2+ messages in thread
From: Andy Shevchenko @ 2023-10-30 8:59 UTC (permalink / raw)
To: Alexey Dobriyan
Cc: Petr Mladek, Steven Rostedt, Rasmus Villemoes,
Sergey Senozhatsky, linux-kernel
On Fri, Oct 27, 2023 at 05:13:58PM +0300, Alexey Dobriyan wrote:
> * uninline simple_strntoull(),
> gcc overinlines and this function is not performance critical
>
> * reorder arguments, so that appending INT_MAX as 4th argument
> generates very efficient tail call
>
> Space savings:
>
> add/remove: 1/0 grow/shrink: 0/3 up/down: 27/-179 (-152)
> Function old new delta
> simple_strntoll - 27 +27
> simple_strtoull 15 10 -5
> simple_strtoll 41 7 -34
> vsscanf 1930 1790 -140
Makes sense to me
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
...
> if (is_sign)
> - val.s = simple_strntoll(str,
> - field_width >= 0 ? field_width : INT_MAX,
> - &next, base);
> + val.s = simple_strntoll(str, &next, base,
> + field_width >= 0 ? field_width : INT_MAX);
> else
> - val.u = simple_strntoull(str,
> - field_width >= 0 ? field_width : INT_MAX,
> - &next, base);
> + val.u = simple_strntoull(str, &next, base,
> + field_width >= 0 ? field_width : INT_MAX);
Looking at these, why do we even care about signedness? field_witdh IIRC is 16-bit or less
and if size_t is to big it's still fine. No?
--
With Best Regards,
Andy Shevchenko
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [PATCH] vsprintf: uninline simple_strntoull(), reorder arguments
[not found] <82a2af6e-9b6c-4a09-89d7-ca90cc1cdad1@p183>
2023-10-30 8:59 ` [PATCH] vsprintf: uninline simple_strntoull(), reorder arguments Andy Shevchenko
@ 2023-11-01 16:48 ` Petr Mladek
1 sibling, 0 replies; 2+ messages in thread
From: Petr Mladek @ 2023-11-01 16:48 UTC (permalink / raw)
To: Alexey Dobriyan
Cc: Steven Rostedt, Andy Shevchenko, Rasmus Villemoes,
Sergey Senozhatsky, linux-kernel
On Fri 2023-10-27 17:13:58, Alexey Dobriyan wrote:
> * uninline simple_strntoull(),
> gcc overinlines and this function is not performance critical
>
> * reorder arguments, so that appending INT_MAX as 4th argument
> generates very efficient tail call
>
> Space savings:
>
> add/remove: 1/0 grow/shrink: 0/3 up/down: 27/-179 (-152)
> Function old new delta
> simple_strntoll - 27 +27
> simple_strtoull 15 10 -5
> simple_strtoll 41 7 -34
> vsscanf 1930 1790 -140
>
> Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
I am usually quite skeptical about these micro-optimizations. They might
help only on some architectures with some compiler versions. And
they might make it worse for others.
Well, I could imagine that passing constant via the last parameter
might help in most cases. And the new ordering of the parameters
is fine. So let's get it in.
Reviewed-by: Petr Mladek <pmladek@suse.com>
I have just pushed the patch into printk/linux.git, branch for-6.7.
Best Regards,
Petr
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2023-11-01 16:48 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
[not found] <82a2af6e-9b6c-4a09-89d7-ca90cc1cdad1@p183>
2023-10-30 8:59 ` [PATCH] vsprintf: uninline simple_strntoull(), reorder arguments Andy Shevchenko
2023-11-01 16:48 ` Petr Mladek
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®