Date: Sat, 14 Sep 1996 16:09:35 +0200 (MET DST) From: Mark Habersack Reply-To: grendel AT ananke DOT amu DOT edu DOT pl To: Zeljko Vrba cc: djgpp AT delorie DOT com Subject: Re: long long type In-Reply-To: <51ba1h$gk8@bagan.srce.hr> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII On 13 Sep 1996, Zeljko Vrba wrote: >The gcc info says that long long type is implemented in software if >the hardware cannot handle 64 bit integers. But Intel coprocessors >CAN handle 64 bit integers, and DJGPP still uses slow library routines >to do addition,substraction etc... Not that easy. True, i386 can handle 64bit integers but only using TWO 32bit registers which is not always desirable and available. Having to cope with a problem of scheduling registers to optimize computations of 64bit integers would mean increasing the compilers size/complexity (I suppose) and also might decrease overall performance of compiled image due to the need to assign EAX and EDX (mostly) to operations on 64 bit integers. But is worth the effort? Are integers of that size used frequently enough to justify creating such a support? > >Even Turbo Pascal knows of Comp type (64 bit integers) and deals with >it as with ordinary 'Real' numbers. > >Why doesn't gcc use Intel 387 instructions to do 64 bit integer arithmetic? Because noone (even today) can assume that everybody out there is using high-tech computers with built-in FPUs. DJGPP and programs compiled with it are supposed to run on ordinary 386SX (even if there will be only one man left using such a machine ;-))) ********************************************************************** So if you ask me how do I feel inside, I could honestly tell you we've been taken on a very long ride. And if my owners let me have free time some day, with all good intention I would probably run away! Clutching the short straw... ******************* http://ananke.amu.edu.pl/~grendel ****************