mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Kuan-Wei Chiu <visitorckw@gmail.com>
To: Yury Norov <yury.norov@gmail.com>
Cc: linux-kernel@vger.kernel.org,
	Rasmus Villemoes <linux@rasmusvillemoes.dk>,
	Andrew Morton <akpm@linux-foundation.org>,
	Chin-Chun Chen <n26122115@gs.ncku.edu.tw>,
	Ching-Chun Huang <jserv@ccns.ncku.edu.tw>
Subject: Re: [PATCH 3/4] bitops: squeeze even more out of fns()
Date: Fri, 3 May 2024 10:19:10 +0800	[thread overview]
Message-ID: <ZjRJnvig6EDAaJ5K@visitorckw-System-Product-Name> (raw)
In-Reply-To: <20240502233204.2255158-4-yury.norov@gmail.com>

+Cc Chin-Chun Chen & Ching-Chun (Jim) Huang

On Thu, May 02, 2024 at 04:32:03PM -0700, Yury Norov wrote:
> The function clears N-1 first set bits to find the N'th one with:
> 
> 	while (word && n--)
> 		word &= word - 1;
> 
> In the worst case, it would take 63 iterations.
> 
> Instead of linear walk through the set bits, we can do a binary search
> by using hweight(). This would work even better on platforms supporting
> hardware-assisted hweight() - pretty much every modern arch.
> 
Chin-Chun once proposed a method similar to binary search combined with
hamming weight and discussed it privately with me and Jim. However,
Chin-Chun found that binary search would actually impair performance
when n is small. Since we are unsure about the typical range of n in
our actual workload, we have not yet proposed any relevant patches. If
considering only the overall benchmark results, this patch looks good
to me.

Regards,
Kuan-Wei

> On my Ryzen 9 5900X, the test_fns() benchmark runs binary fns() twice
> faster, comparing to linear one.
> 
> The fns8() returns 64 to make sure that in case of no bit found, the
> return value will be greater than the bit capacity of arguments of all
> fnsXX() functions up to fns64().
> 
> Signed-off-by: Yury Norov <yury.norov@gmail.com>
> ---
>  include/linux/bitops.h | 42 +++++++++++++++++++++++++++++++++++++-----
>  1 file changed, 37 insertions(+), 5 deletions(-)
> 
> diff --git a/include/linux/bitops.h b/include/linux/bitops.h
> index 57ecef354f47..1c4773db56e0 100644
> --- a/include/linux/bitops.h
> +++ b/include/linux/bitops.h
> @@ -247,17 +247,49 @@ static inline unsigned long __ffs64(u64 word)
>  	return __ffs((unsigned long)word);
>  }
>  
> +static inline unsigned long fns8(u8 word, unsigned int n)
> +{
> +	while (word && n--)
> +		word &= word - 1;
> +
> +	return word ? __ffs(word) : 64;
> +}
> +
> +static inline unsigned long fns16(u16 word, unsigned int n)
> +{
> +	unsigned int w = hweight8((u8)word);
> +
> +	return n < w ? fns8((u8)word, n) : 8 + fns8((u8)(word >> 8), n - w);
> +}
> +
> +static inline unsigned long fns32(u32 word, unsigned int n)
> +{
> +	unsigned int w = hweight16((u16)word);
> +
> +	return n < w ? fns16((u16)word, n) : 16 + fns16((u16)(word >> 16), n - w);
> +}
> +
> +static inline unsigned long fns64(u64 word, unsigned int n)
> +{
> +	unsigned int w = hweight32((u32)word);
> +
> +	return n < w ? fns32((u32)word, n) : 32 + fns32((u32)(word >> 32), n - w);
> +}
> +
>  /**
>   * fns - find N'th set bit in a word
>   * @word: The word to search
> - * @n: Bit to find
> + * @n: Bit to find, counting from 0
> + *
> + * Returns N'th set bit. If no such bit found, returns >= BITS_PER_LONG
>   */
>  static inline unsigned long fns(unsigned long word, unsigned int n)
>  {
> -	while (word && n--)
> -		word &= word - 1;
> -
> -	return word ? __ffs(word) : BITS_PER_LONG;
> +#if BITS_PER_LONG == 64
> +	return fns64(word, n);
> +#else
> +	return fns32(word, n);
> +#endif
>  }
>  
>  /**
> -- 
> 2.40.1
> 

  reply	other threads:[~2024-05-03  2:19 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-05-02 23:32 [PATCH 0/4] bitops: optimize fns() for more Yury Norov
2024-05-02 23:32 ` [PATCH 1/4] lib: make test_bitops compilable into the kernel image Yury Norov
2024-05-03  2:00   ` Kuan-Wei Chiu
2024-05-03  2:14     ` Yury Norov
2024-05-03  2:24       ` Kuan-Wei Chiu
2024-05-03 15:13         ` Yury Norov
2024-05-03 15:29           ` Kuan-Wei Chiu
2024-05-02 23:32 ` [PATCH 2/4] bitmap: relax find_nth_bit() limitation on return value Yury Norov
2024-05-02 23:32 ` [PATCH 3/4] bitops: squeeze even more out of fns() Yury Norov
2024-05-03  2:19   ` Kuan-Wei Chiu [this message]
2024-05-03 16:13     ` Yury Norov
2024-05-04  8:53       ` Kuan-Wei Chiu
2024-05-02 23:32 ` [PATCH 4/4] MAINTAINERS: add BITOPS API record Yury Norov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=ZjRJnvig6EDAaJ5K@visitorckw-System-Product-Name \
    --to=visitorckw@gmail.com \
    --cc=akpm@linux-foundation.org \
    --cc=jserv@ccns.ncku.edu.tw \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux@rasmusvillemoes.dk \
    --cc=n26122115@gs.ncku.edu.tw \
    --cc=yury.norov@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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®