Date: Thu, 11 Mar 2004 00:13:48 +1100 From: Tim Robbins <tjr@freebsd.org> To: Alfred Perlstein <alfred@freebsd.org> Cc: John Birrell <jb@cimlogic.com.au> Subject: Re: cvs commit: src/lib/libc/stdio _flock_stub.c local.h Message-ID: <20040310131348.GA95975@cat.robbins.dropbear.id.au> In-Reply-To: <20040310035912.GQ56622@elvis.mu.org> References: <200403090245.i292j0a6035728@repoman.freebsd.org> <20040309032248.GA88649@cat.robbins.dropbear.id.au> <20040309143223.Q234@freebsd3.cimlogic.com.au> <20040309035532.GA88825@cat.robbins.dropbear.id.au> <20040309150536.R234@freebsd3.cimlogic.com.au> <20040309043207.GA65153@kanpc.gte.com> <20040310000922.A5035@gamplex.bde.org> <20040310035912.GQ56622@elvis.mu.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, Mar 09, 2004 at 07:59:12PM -0800, Alfred Perlstein wrote: > * Bruce Evans <bde@zeta.org.au> [040309 07:50] wrote: > > On Mon, 8 Mar 2004, Alexander Kabaev wrote: > > > > > On Tue, Mar 09, 2004 at 03:05:36PM +1100, John Birrell wrote: > > > > > > > > I'm not sure that I agree that applications are 'broken' when they > > > > use things that are defined in the header file along with the FILE > > > > structure itself. > > > > It's a historical mistake that FILE is not opaque. > > > > > I would like to see FILE to become transparent to applications and its > > > definition moved to the libc-private header file with the specific > > > purpose of making the hack you mentioned impossible. > > > > This would pessimize even getc_unlocked() and putc_unlocked(). getc() > > and putc() are now extern functions, but the old macro/inline versions > > are still available as getc_unlocked() and putc_unlocked(). Simple > > benchmarks for reading a 100MB file on an Athlon XP1600 overclocked > > show that the function versions are up to 9 times slower: > > > > Time to read the file from a disk cache using read(): 0.17 seconds (sys) > > getc_unlocked() overhead: 0.41 seconds (user) > > getc() overhead: 1.64 seconds (user) > > > > This is with static linkage. Dynamic linkage increases the getc() > > pessimization significantly: > > > > read() time: no significant change > > getc_unlocked() overhead: 0.44 seconds (user) > > getc() overhead: 3.62 seconds (user) > > Hmm, can't we use macros that do this: > > #define getc() (__isthreaded ? old_unlocked_code : getc_unlocked()) > > Where __isthreaded is a global that's set by threading libraries > to 1 and 0 by non-threaded libc, this should get rid of a lot of > the function call overhead. Sounds like a good idea to me. In my testing, this approach was about 5% slower than calling getc_unlocked() directly (due to the conditional jump), but roughly 3 times faster than a call to the getc() function. If there aren't any objections, I think we should implement getc()/putc() this way (and all the other stdio functions that have traditionally had macro equivalents) before 5-stable to try to recoup some of the performance losses caused by the removal of the macros. Tim
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040310131348.GA95975>