Return to BSD News archive
Path: sserve!newshost.anu.edu.au!munnari.oz.au!sgiblab!rpal.rockwell.com!headwall.Stanford.EDU!agate!agate.berkeley.edu!cgd From: cgd@erewhon.CS.Berkeley.EDU (Chris G. Demetriou) Newsgroups: comp.os.386bsd.misc Subject: Re: Shared Library Status ? Date: 8 Mar 94 16:25:10 Organization: Kernel Hackers 'r' Us Lines: 25 Message-ID: <CGD.94Mar8162510@erewhon.CS.Berkeley.EDU> References: <JKH.94Mar5233255@whisker.hubbard.ie> <hastyCM8Buv.26z@netcom.com> <CGD.94Mar7155635@erewhon.CS.Berkeley.EDU> <BY1pv3m.dysonj@delphi.com> NNTP-Posting-Host: erewhon.cs.berkeley.edu In-reply-to: John Dyson's message of Tue, 8 Mar 94 12:12:30 -0500 John Dyson says: =>Chris G. Demetriou <cgd@erewhon.CS.Berkeley.EDU> writes: =>>"DAMN!" i didn't realize it made that little difference. cool. 8-) =>>(i'd actually not benchmarked them, nor even compared them...) => =>The biggest cost of the shared libraries is in CPU at process startup. Both =>USER mode and KERNEL mode. Yes, i know exactly what *causes* shared library slow-down. What i was commenting on was the fact that, even given the CPU overhead, there wasn't much difference in the compile time. Yes, some compiles tend to be "long running," but by many they're looked at as a benchmark of sorts, and they're certainly a real-world performance metric, esp. if you've gotta sit and wait for them to complete. cgd -- chris g. demetriou cgd@cs.berkeley.edu you can eat anything once.