Xref: news-dnh.mv.net comp.os.msdos.djgpp:4099 Path: news-dnh.mv.net!mv!news.sprintlink.net!newsfeed.internetmci.com!tank.news.pipex.net!pipex!oleane!in2p3.fr!univ-lyon1.fr!jussieu.fr!unilim.fr!cict.fr!irit.fr!challier From: challier AT news DOT irit DOT fr (Jean-Marc CHALLIER) Newsgroups: comp.os.msdos.djgpp Subject: Re: DJGPP: Inline assembly? Date: 3 Jan 1996 11:55:33 GMT Organization: IRIT-UPS, Toulouse, France Lines: 27 Message-ID: <4cdqrl$ie@orphee.irit.fr> References: <4apphs$las AT micro DOT internexus DOT net> <4b65md$ol0 AT orphee DOT irit DOT fr> <4b9n8n$93q AT stealth DOT mindspring DOT com> NNTP-Posting-Host: lucie.irit.fr To: djgpp AT delorie DOT com DJ-Gateway: from newsgroup comp.os.msdos.djgpp Thomas Sames (tsames AT atl DOT mindspring DOT com) wrote: : You can also use an offset from the stack to obtain the passed : parameter to the function. This is faster for a 386 but for the 486 : and pentuim, use the local variable. True but be careful about that : the compiler may "push" a few registers on the stack before the function code itself, thus destroying your knowledge of "what's-where" relative to ESP. If you meant EBP, it's quite more safe, but your program won't be able to compile both with and with out the "-fomit-frame-pointer" option. I know it's a bit buggy (specially with C++), but for for small functions that can't be inlined (like those that are called via a pointer) it can dramatically improve speed. The point in what I wrote was : Leave to GCC as much work as possible : it's almost the only one who knows how to play in its mud... :-) Hope this helps... Igx, The dreaming drummer... ----------------------------------------------------------------------------- Igx, The dreaming drummer... in real(???) life : Jean-Marc CHALLIER email : challier AT irit DOT fr | J-M CHALLIER, pmail : IR1-17, UPS, 18 rte de Narbonne | 18, Bd Griffoul Dorval 31000 Toulouse - FRANCE | 31400 Toulouse - FRANCE phone : (33) 61-55-63-57 | Tel : (33) 62-16-35-92