Re: [PATCH] kernel/irq/proc: performance: replace seq_printf with seq_put_decimal_ull_width

Re: [PATCH] kernel/irq/proc: performance: replace seq_printf with seq_put_decimal_ull_width

[Thomas Gleixner]

David!

On Sun, Nov 03 2024 at 16:05, David Wang wrote:

$Subject: [PATCH] kernel/irq/proc: performance: ...

That's not a valid subsystem prefix.

> seq_printf is costy, when stress reading /proc/interrupts, profiling indicates
> seq_printf takes about ~47% of show_interrupts samples:

Also please follow the documentation for denoting functions in change
logs:

https://www.kernel.org/doc/html/latest/process/maintainer-tip.html#function-references-in-changelogs

>  arch/alpha/kernel/irq.c     |   8 +--
>  arch/arm/kernel/smp.c       |   4 +-
>  arch/arm64/kernel/smp.c     |   3 +-
>  arch/csky/kernel/smp.c      |   4 +-
>  arch/loongarch/kernel/smp.c |   2 +-
>  arch/parisc/kernel/irq.c    |  34 ++++++------
>  arch/powerpc/kernel/irq.c   |  44 ++++++++--------
>  arch/riscv/kernel/smp.c     |   3 +-
>  arch/sh/kernel/irq.c        |   4 +-
>  arch/sparc/kernel/irq_32.c  |  12 ++---
>  arch/sparc/kernel/irq_64.c  |   4 +-
>  arch/x86/kernel/irq.c       | 100 ++++++++++++++++++------------------
>  arch/xtensa/kernel/irq.c    |   2 +-
>  arch/xtensa/kernel/smp.c    |   4 +-
>  kernel/irq/proc.c           |   6 ++-

There is no dependency on these changes. So please split them up into
seperate patches for core and the individual architectures.

Thanks,

        tglx

Re: [PATCH] kernel/irq/proc: performance: replace seq_printf with seq_put_decimal_ull_width

[David Wang]

Hi,  

At 2024-11-08 22:25:13, "Thomas Gleixner" <tglx@linutronix.de> wrote:
>David!
>
>On Sun, Nov 03 2024 at 16:05, David Wang wrote:
>
>$Subject: [PATCH] kernel/irq/proc: performance: ...
>
>That's not a valid subsystem prefix.

copy that~

>
>> seq_printf is costy, when stress reading /proc/interrupts, profiling indicates
>> seq_printf takes about ~47% of show_interrupts samples:
>
>Also please follow the documentation for denoting functions in change
>logs:
>
>https://www.kernel.org/doc/html/latest/process/maintainer-tip.html#function-references-in-changelogs
>

copy that~

>>  arch/alpha/kernel/irq.c     |   8 +--
>>  arch/arm/kernel/smp.c       |   4 +-
>>  arch/arm64/kernel/smp.c     |   3 +-
>>  arch/csky/kernel/smp.c      |   4 +-
>>  arch/loongarch/kernel/smp.c |   2 +-
>>  arch/parisc/kernel/irq.c    |  34 ++++++------
>>  arch/powerpc/kernel/irq.c   |  44 ++++++++--------
>>  arch/riscv/kernel/smp.c     |   3 +-
>>  arch/sh/kernel/irq.c        |   4 +-
>>  arch/sparc/kernel/irq_32.c  |  12 ++---
>>  arch/sparc/kernel/irq_64.c  |   4 +-
>>  arch/x86/kernel/irq.c       | 100 ++++++++++++++++++------------------
>>  arch/xtensa/kernel/irq.c    |   2 +-
>>  arch/xtensa/kernel/smp.c    |   4 +-
>>  kernel/irq/proc.c           |   6 ++-
>
>There is no dependency on these changes. So please split them up into
>seperate patches for core and the individual architectures.an

Thanks for all the review, I will make a patchset for this.
>
>Thanks,
>
>        tglx


David

Re: [PATCH] kernel/irq/proc: performance: replace seq_printf with seq_put_decimal_ull_width

[patchwork-bot+linux-riscv]

Hello:

This patch was applied to riscv/linux.git (fixes)
by Michael Ellerman <mpe@ellerman.id.au>:

On Sun,  3 Nov 2024 16:05:52 +0800 you wrote:
> seq_printf is costy, when stress reading /proc/interrupts, profiling indicates
> seq_printf takes about ~47% of show_interrupts samples:
> 
>     show_interrupts(94.495% 5166019/5466991)
> 	seq_printf(47.429% 2450210/5166019)
> 	    vsnprintf(89.232% 2186366/2450210)
> 		format_decode(24.005% 524831/2186366)
> 		number(19.488% 426084/2186366)
> 		memcpy_orig(3.739% 81753/2186366)
> 		...
> 	_raw_spin_unlock_irqrestore(26.643% 1376379/5166019)
> 	mtree_load(8.059% 416304/5166019)
> 
> [...]

Here is the summary with links:
  - kernel/irq/proc: performance: replace seq_printf with seq_put_decimal_ull_width
    https://git.kernel.org/riscv/c/5b881c1f8379

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html