Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1693032

Re: [PATCH] lib/int_sqrt.c: Optimize square root function

From Peter Zijlstra <peterz@infradead.org>
Newsgroups linux.kernel
Subject Re: [PATCH] lib/int_sqrt.c: Optimize square root function
Date 2017-07-20 17:30 +0200
Message-ID <u5lyO-QM-13@gated-at.bofh.it> (permalink)
References <oL98t-3is-3@gated-at.bofh.it> <oPsPg-7ue-17@gated-at.bofh.it> <oPuxJ-1C4-35@gated-at.bofh.it> <oPuHo-1NW-17@gated-at.bofh.it> <u5hOy-6De-7@gated-at.bofh.it>
Organization linux.* mail to news gateway

Show all headers | View raw

On Thu, Jul 20, 2017 at 01:24:49PM +0200, Peter Zijlstra wrote:
> ~/tmp$ gcc -o sqrt sqrt.c -lm -O2 -DLOOPS=10000000 -DNEW=1 -DFLS=1 -DANSHUL=1  ; perf stat --repeat 10 -e cycles:u -e instructions:u ./sqrt
> 
>  Performance counter stats for './sqrt' (10 runs):
> 
>        328,415,775      cycles:u                                                      ( +-  0.15% )
>      1,138,579,704      instructions:u            #    3.47  insn per cycle           ( +-  0.00% )
> 
>        0.088703205 seconds time elapsed  

> static __always_inline unsigned long fls(unsigned long word)
> {
> 	asm("rep; bsr %1,%0"
> 		: "=r" (word)
> 		: "rm" (word));
> 	return BITS_PER_LONG - 1 - word;
> }

That is actually "lzcnt", if I used the regular fls implementation:


static __always_inline unsigned long __fls(unsigned long word)
{
	asm("bsr %1,%0"
		: "=r" (word)
		: "rm" (word));
	return word;
}

It ends up slightly more expensive:

~/tmp$ gcc -o sqrt sqrt.c -lm -O2 -DLOOPS=10000000 -DNEW=1 -DFLS=1 -DANSHUL=1  ; perf stat --repeat 10 -e cycles:u -e instructions:u ./sqrt

 Performance counter stats for './sqrt' (10 runs):

       384,842,215      cycles:u                                                      ( +-  0.08% )
     1,118,579,712      instructions:u            #    2.91  insn per cycle           ( +-  0.00% )

       0.103018001 seconds time elapsed  


Still loads cheaper than pretty much any other combination.

Back to linux.kernel | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

Re: [PATCH] lib/int_sqrt.c: Optimize square root function Peter Zijlstra <peterz@infradead.org> - 2017-07-20 13:30 +0200
  Re: [PATCH] lib/int_sqrt.c: Optimize square root function Joe Perches <joe@perches.com> - 2017-07-20 14:00 +0200
    Re: [PATCH] lib/int_sqrt.c: Optimize square root function Peter Zijlstra <peterz@infradead.org> - 2017-07-20 16:20 +0200
  Re: [PATCH] lib/int_sqrt.c: Optimize square root function Peter Zijlstra <peterz@infradead.org> - 2017-07-20 17:30 +0200
  Re: [PATCH] lib/int_sqrt.c: Optimize square root function Linus Torvalds <torvalds@linux-foundation.org> - 2017-07-20 20:40 +0200
    Re: [PATCH] lib/int_sqrt.c: Optimize square root function Peter Zijlstra <peterz@infradead.org> - 2017-07-21 00:40 +0200
      Re: [PATCH] lib/int_sqrt.c: Optimize square root function Linus Torvalds <torvalds@linux-foundation.org> - 2017-07-21 01:30 +0200
        Re: [PATCH] lib/int_sqrt.c: Optimize square root function Peter Zijlstra <peterz@infradead.org> - 2017-07-21 13:50 +0200
          Re: [PATCH] lib/int_sqrt.c: Optimize square root function Joe Perches <joe@perches.com> - 2017-07-21 14:20 +0200
            Re: [PATCH] lib/int_sqrt.c: Optimize square root function Peter Zijlstra <peterz@infradead.org> - 2017-07-21 15:30 +0200
              Re: [PATCH] lib/int_sqrt.c: Optimize square root function Peter Zijlstra <peterz@infradead.org> - 2017-07-21 15:40 +0200

csiph-web