[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 02 Feb 2015 13:56:27 +0100
From: Rasmus Villemoes <linux@...musvillemoes.dk>
To: "George Spelvin" <linux@...izon.com>
Cc: yury.norov@...il.com, akpm@...ux-foundation.org,
chris@...is-wilson.co.uk, davem@...emloft.net, dborkman@...hat.com,
hannes@...essinduktion.org, klimov.linux@...il.com,
laijs@...fujitsu.com, linux-kernel@...r.kernel.org,
msalter@...hat.com, takahiro.akashi@...aro.org, tgraf@...g.ch,
valentinrothberg@...il.com, y.norov@...sung.com
Subject: Re: [PATCH v2 1/3] lib: find_*_bit reimplementation
On Mon, Feb 02 2015, "George Spelvin" <linux@...izon.com> wrote:
> Rasmus Villemoes <linux@...musvillemoes.dk> wrote:
>> ... and this be part of _find_next_bit? Can find_next_bit not be simply
>> 'return _find_next_bit(addr, size, offset, 1);', and similarly for
>> find_next_zero_bit? Btw., passing true and false for the boolean
>> parameter may be a little clearer.
>
> Looking at the generated code, it would be better to replace the boolean
> parameter with 0ul or ~0ul and XOR with it. The same number of registers,
> and saves a conditional branch.
Nice trick. When I compiled it, gcc inlined _find_next_bit into both its
callers, making the conditional go away completely. That was with gcc
4.7. When I try with 5.0, I do see _find_next_bit being compiled
separately.
With the proposed change, 4.7 also makes find_next{,_zero}_bit wrappers
for _find_next_bit, further reducing the total size, which is a good
thing. And, if some other version decides to still inline it, it
should then know how to optimize the xor with 0ul or ~0ul just as well
as when the conditional was folded away.
Yury, please also incorporate this in the next round.
Rasmus
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists