Date: Tue, 11 Aug 2015 10:03:27 +0800 From: Erich Dollansky <erichsfreebsdlist@alogt.com> To: Hans Petter Selasky <hps@selasky.org> Cc: Andriy Gapon <avg@FreeBSD.org>, Konstantin Belousov <kostikbel@gmail.com>, "freebsd-hackers@freebsd.org" <freebsd-hackers@FreeBSD.org> Subject: Re: allow ffs & co. a binary search Message-ID: <20150811100327.0d12231f@X220.alogt.com> In-Reply-To: <55C335A7.8020503@selasky.org> References: <20150607081315.7c0f09fb@B85M-HD3-0.alogt.com> <5573EA5E.40806@selasky.org> <20150607195245.62dc191f@B85M-HD3-0.alogt.com> <20150607135453.GH2499@kib.kiev.ua> <558175FA.4040106@FreeBSD.org> <20150617165331.GA2080@kib.kiev.ua> <5582CCF1.8010505@FreeBSD.org> <55C335A7.8020503@selasky.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Hi, On Thu, 06 Aug 2015 12:23:35 +0200 Hans Petter Selasky <hps@selasky.org> wrote: > On 06/18/15 15:51, Andriy Gapon wrote: > > On 17/06/2015 19:53, Konstantin Belousov wrote: > >> AFAIR it was about 'sufficiently smart compiler' and the fact that > >> the functions are not on the hottest paths. > > > > It seems that sufficiently smart compilers still do not exist :-) > > At least as far as compilers that are used for compiling FreeBSD > > are considered. > > > > [Offtopic] my impression is that lately smartness of compilers is > > mostly being improved by various tricks and shortcuts (undefined > > behavior, etc), rather than by recognizing patterns in the C code > > that could be turned into more efficient machine code. > > > Has there been any further thoughts on this topic during the summer? > I was just waiting for a clear result. I do not want to write something which is not accepted at the end. > It sounds fair to me that "Sean Eron Anderson's Bit Twiddling Hacks" > can be used in our code instead of a for() loop for the ffs() > function, given that a 2-clause BSD license is fine by Sean. > Typically the compilers can expand for loops using -O3, but I've > never seen a compiler yet that convert for loops into a simpler > functions where no constants are involved. Why not detect the built-in functions and do the binary search when they are not available and leave the original implementation for all other sizes? > > For sake of readability I would leave the old implementation under > "#if 0" so that readers not familiar with binary tricks can > understand the code. > > Any strong objections? > A clear route would be good. I could do it then. Erich
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20150811100327.0d12231f>