Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!mnetor!uunet!seismo!husc6!necntc!ames!ucbcad!ucbvax!ucbarpa.Berkeley.EDU!cuccia From: cuccia@ucbarpa.Berkeley.EDU (Nick Cuccia) Newsgroups: comp.os.minix Subject: Floating Point routines Message-ID: <19676@ucbvax.BERKELEY.EDU> Date: Sun, 12-Jul-87 02:40:52 EDT Article-I.D.: ucbvax.19676 Posted: Sun Jul 12 02:40:52 1987 Date-Received: Mon, 13-Jul-87 00:38:29 EDT References: <114@jc3b21.UUCP> <1205@botter.cs.vu.nl> Sender: usenet@ucbvax.BERKELEY.EDU Reply-To: cuccia@ucbarpa.Berkeley.EDU.UUCP (Nick Cuccia) Organization: University of California, Berkeley Lines: 28 In article <1205@botter.cs.vu.nl> ast@cs.vu.nl (Andy Tanenbaum) writes: >1) No floats. I agree, they would be nice, but I haven't personally used > a number > 32767 in years. Is anyone working on floats? >Andy Tanenbaum (ast@cs.vu.nl) I'm currently working on versions of /usr/lib/libm.a (math functions) and /usr/lib/libmp.a (multiple precision integer functions) to be distributed with GNU, and will post sources when I'm finished. My development environments are several VAXen (750 running 4.2BSD, 8600 running 4.3BSD) and a Sun 3/280 running SunOS v3.2, and will port the routines to MINIX when final tuning is finished. This brings me to another question: one of the niceties of the C compiler for the Sun is the '-f' flag, through which the user can tell the compiler what kind of acceleration hardware, if any, is installed in the workstation or server. For those working on the C compiler: how difficult would it be to install such a flag, whose args would be the following: flag meaning ---- ------- -fsingle do not promote floats to doubles -f8087 compile code to be used with 8087 processor -fswitch generate code that determines whether or not an accelerator chip is installed at runtime. --Nick