Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/3/84; site osu-cgrg.UUCP
Path: utzoo!watmath!clyde!cbosgd!apr!osu-eddie!osu-cgrg!flip
From: flip@osu-cgrg.UUCP (Flip Phillips)
Newsgroups: net.micro.amiga
Subject: floating point with sun x-compiler
Message-ID: <626@osu-cgrg.UUCP>
Date: Wed, 26-Feb-86 18:33:48 EST
Article-I.D.: osu-cgrg.626
Posted: Wed Feb 26 18:33:48 1986
Date-Received: Fri, 28-Feb-86 07:49:34 EST
Distribution: net
Organization: Computer Graphics Research Group, Ohio State University
Lines: 28

If anyone using the cross compiler system for the sun is having problems
with floating point I/O heres the prob: I think...

	The greenhills compiler compiles EVERY double/float into a FFP
number.  When passing it to routines it doesnt really convert it into 
a double as per the way you do things...  it really puts a null byte up
there with it. So much for calling the SPTieee / SPFieee routines since
they return a 32 bit value.
	The Lattice library for the sun dev. system is compiled with/does
floating point conversion to IEEE numbers... 32 bit floats, 64 bit doubles.
When, say, printf("%f",val) goes, it expects a double for val in iEEE
format.  Fine.  Even if you convert it with the afore mentioned things it
seems that it still gets converted into a FFP type pseudo-double with 
null byte attached.  Bummer.  printf's / scanf's conversion routine pukes.
	So... I came up with a sorta kluge that I'd be willing to share.
Respond to me via mail 'causin I dont want to clutter up the already
cluttered up net.
	If any of you other folk have come up with sumpthin, let me know! I
really would appreciate it.

flip
	


-- 

	Flip Phillips {ucbvax,decvax}!cbosg!osu-cgrg!flip
	Computer Graphics Research Group, The Ohio State University