Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!uunet!mcsun!inria!bull!bpe2.spix7.depr.bull.fr!marc
From: marc@bpe2.spix7.depr.bull.fr ( Marc PARIGOT )
Newsgroups: comp.sys.apollo
Subject: Porting NCS on Dps7 (study case)
Message-ID: <231@bull.bull.fr>
Date: 20 Dec 89 16:19:37 GMT
Sender: news@bull.bull.fr
Reply-To: marc@bpe2.spix7.depr.bull.fr ( Marc PARIGOT )
Organization: BULL S.A., Paris , France
Lines: 36


	Hello,

I work for Bull (paris) in the team that port Unix on DPS7: Spix7. It is a full
SV.3/X.Open unix system. At present my boss ask me to make some investigation 
in the aim to port NCS on Spix7. He want to know the time to do this job. 
We have ported TCP(UDP)/IP and the bsd socket but I've read somewhere in this 
newsgroup that the difficulty to port NCS is a function of how much of the bsd
socket are ported. I want to know in which way?
I want to know the number of lines NCS represent and perhaps the percent that
is not portable.
And last thing (for the moment) I want to know if it exists a simple and fast
way to convert IBM floatting point format number into IEEE format number
and vice-versa (the format of dps7 floatting point number is the Ibm format):

	-------------------------        ----------------------------
	|s| exp |  mantissa     |  <==>  |   mantissa      | exp  |s|
	-------------------------        ----------------------------
		fibm                             fieee


 fibm = (-1)**s * (16)**(exp1-64) * mantissa1
 fieee = (-1)**s * 2**(exp2-127) * mantissa2

 ibm_to_ieee(exp1, mantissa1, &exp2, &mantissa2);
 ieee_to_ibm(exp2, mantissa2, &exp1, &mantissa1);

		Thanks for your time.

					Marc

Post-scriptum:
	I beg your pardon for my approximative english, I realize that it's a
	kind of pseudo language...

E.mail: Marc.Parigot@depr.bull.fr