Date: Sun, 07 Jun 2015 17:02:32 +0200 From: Hans Petter Selasky <hps@selasky.org> To: Erich Dollansky <erichsfreebsdlist@alogt.com> Cc: "freebsd-hackers@freebsd.org" <freebsd-hackers@FreeBSD.org> Subject: Re: allow ffs & co. a binary search Message-ID: <55745D08.5000505@selasky.org> In-Reply-To: <20150607195245.62dc191f@B85M-HD3-0.alogt.com> References: <20150607081315.7c0f09fb@B85M-HD3-0.alogt.com> <5573EA5E.40806@selasky.org> <20150607195245.62dc191f@B85M-HD3-0.alogt.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 06/07/15 13:52, Erich Dollansky wrote:
> Hi Hanns,
>
> On Sun, 07 Jun 2015 08:53:18 +0200
> Hans Petter Selasky <hps@selasky.org> wrote:
>
>> On 06/07/15 02:13, Erich Dollansky wrote:
>>>
>>> If you like the binary way I could give you the sources for the
>>> complete family following your style to replace the older functions.
>>>
>> I think this is not the fastest way to do it. You first find the LSB,
>
> I never thought of this as I took the code a long, long time from
> university and expanded it over time from 8 to 64 bits.
>
>> then you do a sumbits, which doesn't have any conditionals IF/ELSE
>> and performs better with the CPU pipeline. I think the software ffs()
>> is only used for platforms which doesn't have a hardware version btw:
>
> The code is not x86 specific.
>>
> This one is the same:
>
>> From my libmbin:
>>
>> int ffs(int value)
>> {
>> int retval = mbin_sumbits32(mbin_lsb32(value) - 1);
>> if (retval == 32)
>> retval = 0;
>> else
>> retval++;
>> return (retval);
>> }
>>
> Where is this? Should this also be in 10.1?
Hi Erich,
It is currently nowhere, just in my private SVN. Would you wrap up
complete patch, and I can put it into Phabricator and let more people
look at it. Does your patch cover both kernel and userspace?
>
> What I saw is that all CPUs except ARM uses the software version.
Kernel too?
--HPS
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?55745D08.5000505>