home *** CD-ROM | disk | FTP | other *** search
- Path: portland.caps.maine.edu!dseama41
- Organization: University of Maine System
- Date: Fri, 29 Mar 1996 21:23:32 EST
- From: Starduster <DSEAMA41@portland.caps.maine.edu>
- Message-ID: <96089.212332DSEAMA41@portland.caps.maine.edu>
- Newsgroups: comp.lang.c
- Subject: Re: Slow Turbo-C++ 3.0 for DOS...
- References: <4jejic$b4v@news.xs4all.nl> <4jf5a4$i0f@ccshst05.cs.uoguelph.ca>
- <4jgdnq$ei9@news.xs4all.nl>
-
- In article <4jgdnq$ei9@news.xs4all.nl>, stijnw@xs4all.nl (Stijn Wolters) says:
- >
- >I used some test-sources, very small programs for generating
- >Mandelbrot & Julia fractals. The sources are identical as much as
- >possible.
- >
- >Greetings,
- >Stijn Wolters (stijnw@xs4all.nl)
- >
- >
- >
- What floating-point options did you use for each? Are you using equivalent
- data types in each language? (remembering that real in Pascal is *not* the
- same IEEE type as double in C, etc.) What memory models did you use? WHich
- compiler optimizations were turned on? Lots of questions remain unanswered
- here. It would be best to post the code itself here and let people compare.
- Remember also that certain "special-purpose" I/O functions in C (e.g., getc(),)
- ,getc(),
- putc(), etc. for *one* character only instead of the more general purpose
- printf(), gets(), puts, etc.) may give Turbo C++ the speed advantage you expect
- . In Turbo Pascal, of course, you simply use readln and writeln (or read and
- write) for most I/O, of course. Using the more general counterparts in C may
- take away its expected speed advantage.
-