Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watnot!watmath!clyde!rutgers!ames!ucbcad!ucbvax!RELAY.CS.NET!btb%ncoast.UUCP%case.CSNET
From: btb%ncoast.UUCP%case.CSNET@RELAY.CS.NET.UUCP
Newsgroups: mod.computers.vax
Subject: Re: Data Conversion (PE real*4 to Vax real*4)
Message-ID: <540325683.29344@ncoast>
Date: Sat, 14-Feb-87 17:01:10 EST
Article-I.D.: ncoast.540325683.29344
Posted: Sat Feb 14 17:01:10 1987
Date-Received: Sun, 15-Feb-87 23:05:24 EST
References: <870206135928.005@Xhmeia.Caltech.Edu>
Sender: daemon@ucbvax.BERKELEY.EDU
Reply-To: Brad Banko <btb%ncoast.UUCP%case.csnet@RELAY.CS.NET>
Organization: Cleveland Public Access UN*X, Cleveland, OH
Lines: 55
Approved: info-vax@sri-kl.arpa

In article <870206135928.005@Xhmeia.Caltech.Edu> zar@XHMEIA.CALTECH.EDU writes:
>Can someone tell me how to convert a real*4 Perkin-Elmer data to VAX
>real*4. What I need is how the PE real data is stored (what bits have the
>mantissa etc.). This info for the VAX would be handy, although I can just
>as easily look the VAX data storage for real numbers myself (I'm lazy :-) ).
>Many many thanks and blessings to all who respond...
>
>Zar
>ZAR@CITCHEM (Bitnet)
>ZAR@XHMEIA.CALTECH.EDU (Arpa)

Zar,
	I have just finished struggling with this... we have a PET scanner
that is run on some PE 3205's and a 3210 running OS/32, and we wanted to
do some analysis and archiving of the data on a Vax 11/750 running VMS...

	There are two main problems with this conversion:

		1) PE stores bytes in a Real*4 in a different order
		than on the Vax, 4-3-2-1 vs. 3-4-1-2 respectively,
		where 4 is the msbyte, and the bytes are listed in
		order of increasing storage location... and two

		2) PE uses a different floating point format than the
		Vax... the PE format actually has a wider dynamic
		range than the Vax (roughly 10+/-76 vx. 10+/-38), so
		you have to decide how you will handle that... you can
		also use the /g_float option of the vax fortran compiler,
		and interpret the pe real*4 into a vax g_float real*8,
		but that is sort of a pain...  i tried it both ways, and
		finally ended up going with a conversion into normal
		vax real*4 letting the conversion routine do some checking
		for under/overflow...

concurrent (i.e., pe) wasn't much help with this problem, but they tried...
i ended up having to learn all about floating point formats...
	pe uses an ibm-like format with a 7-bit excess 64 HEX-based
	exponent, whereas
	dec uses an 8-bit excess 128 binary-based exponent

if you want, i could send you the subroutine that i wrote to :

		pe_real4_to_vax_real4(x)

actually it is a fortran function... i think i still have the g_float
version around, too, let me know.


-- 
			Brad Banko
			...!decvax!cwruecmp!ncoast!btb
			Cleveland, Ohio

"The only thing we have to fear on this planet is man."
			-- Carl Jung, 1875-1961