lkml.org 
[lkml]   [2015]   [Feb]   [2]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH] lib/int_sqrt.c: Optimize square root function
From
Date
On Mon, 2015-02-02 at 11:00 -0800, Linus Torvalds wrote:
> Hmm. I don't disagree, but would like some more feedback.
>
> Davidlohr - you were the person to touch this function last (commit
> 30493cc9dddb: "lib/int_sqrt.c: optimize square root algorithm"), and
> you did so for performance reasons. And in fact, when you did that,
> you removed that initial loop:
>
> - one = 1UL << (BITS_PER_LONG - 2);
> - while (one > op)
> - one >>= 2;
>
> but I'm not sure that was actually all that conscious, I think the
> real optimization was the changes inside the loop to make the final
> real loop faster and simpler.

I missed that. And yes, the real optimization should be in the loop.

>
> Also, you had performance numbers, so presumably a test harness for it
> all. It probably depends a lot on the actual distribution of argument
> values, of course, but it would be good to accompany the patch with
> actual real numbers like lasty time.

Aha. In my case I recall I ran a usersapce program using each function
from 1 to a million, and throwing perf at it for 10 times.

> (I'm also not entirely sure what uses int_sqrt() that ends up being so
> performance-critical, so it would be good to document that too, since
> that probably also matters for the "what's the normal argument range"
> question..)

It's not a big deal afaik.

Thanks,
Davidlohr



\
 
 \ /
  Last update: 2015-02-03 06:01    [W:1.904 / U:1.600 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site