w3ai00.f File Reference

Real array to 16 bit packed format. More...

Go to the source code of this file.

Functions/Subroutines
subroutine	q9e3i6 (a, b, n, istat)
	Convert ieee 32 bit task 754 floating point numbers to ibm370 64 bit floating point numbers.
subroutine	q9ei32 (a, b, n, istat)
	Convert IEEE 32 bit task 754 floating point numbers to IBM370 32 bit floating point numbers.
subroutine	w3ai00 (real8, pack, label)
	Converts IEEE floating point numbers to 16 bit packed office note 84 format.

Detailed Description

Real array to 16 bit packed format.

Author

Ralph Jones

Date

1985-07-31

Definition in file w3ai00.f.

Function/Subroutine Documentation

q9e3i6()

subroutine q9e3i6	(	integer(4), dimension(n)	a,
		integer(4), dimension(2,n)	b,
			n,
			istat
	)

Convert ieee 32 bit task 754 floating point numbers to ibm370 64 bit floating point numbers.

Program history log:

• Ralph Jones 1992-08-02
• Ralph Jones 1995-11-15 Add save statement.

Parameters

[in]	A	Real*4 array of IEEE 32 bit floating point numbers.
[in]	N	Number of words to convert to IBM370 64 bit F.P.
[out]	B	Real*8 array of IBM370 64 bit floating point numbers.
[out]	ISTAT	0 All numbers converted. -1 N is less than one. +K K infinity or nan numbers were found.

Note

See IEEE task 754 standard floating point arithmetic for more information about IEEE F.P.

Author

Ralph Jones

Date

1992-08-02

Definition at line 366 of file w3ai00.f.

q9ei32()

subroutine q9ei32	(	integer(4), dimension(*)	a,
		integer(4), dimension(*)	b,
			n,
			istat
	)

Convert IEEE 32 bit task 754 floating point numbers to IBM370 32 bit floating point numbers.

Program history log:

• Ralph Jones 1990-06-04 Convert to sun fortran 1.3.
• Ralph Jones 1990-07-14 Change ishft to lshift or lrshft.
• Ralph Jones 1991-03-28 Change to silicongraphics 3.3 fortran 77.
• Ralph Jones 1992-07-20 Change to ibm aix xl fortran.
• Ralph Jones 1995-11-15 Add save statement.
• Stepen Gilbert 1998-11-18 Specified 4-byte Integer values.

Parameters

[in]	A	- Real*4 array of IEEE 32 bit floating point numbers.
[in]	N	- Number of words to convert to IBM370 32 bit F.P.
[out]	B	- Real*4 array of IBM370 32 bit floating point numbers.
[out]	ISTAT	0: All numbers converted. -1: N is less than one. +K: K infinity or nan numbers were found.

Note

See IEEE task 754 standard floating point arithmetic for more information about IEEE F.P.

Author

Ralph Jones

Date

1990-06-04

Definition at line 270 of file w3ai00.f.

w3ai00()

subroutine w3ai00	(	real, dimension(*)	real8,
		integer(8), dimension(*)	pack,
		integer(8), dimension(6)	label
	)

Converts IEEE floating point numbers to 16 bit packed office note 84 format.

The floating point number are converted to 16 bit signed scaled integers.

Program history log:

• Ralph Jones 1989-10-20 Convert cyber 205 version of w3ai00 to cray.
• Ralph Jones 1990-03-18 Change to use cray integer*2 packer.
• Ralph Jones 1990-10-11 Special version to pack grids larger than 32743 words. Will do old and new version.
• Ralph Jones 1991-02-16 Changes so equivalence of pack and real8 arrays will work.
• Ralph Jones 1993-06-10 Changes for array size (512,512) 262144 words.
• Boi Vuong 1998-03-10 Remove the cdir$ integer=64 directive.
• Stephen Gilbert 1998-11-18 Changed to pack IEEE values for the IBM SP

Parameters

[in]	REAL8	Array of cray floating point numbers.
[in]	LABEL	Six 8-byte integer words. Must have first 8 of 12 32 bit word office note 84 label. word 6 must have in bits 31-00 the number of real words in array real8 if j is greater than 32743. j in bits 15-0 of the 4th id word is set zero.
[out]	PACK	Packed output array of integer words of size 6 + (j+3)/4 , j = no. points in label (from word 4 bits 15-00). Label will be copied to pack words 1-4. Pack will contain the following in words 5-6: word 5 bits 63-48 Number of bytes in whole record. will not be correct if j > 32743. word 5 bits 47-32 Exclusive-or checksum by 16 bit words of whole array pack excluding checksum itself. word 5 bits 31-00 Center value a = mean of max and min values. converted to ibm 32 floating point number. word 6 bits 63-48 Zero. word 6 bits 47-32 16 bit shift value n. the least integer such that abs(x-a)/2**n lt 1 for all x in real8. limited to +-127. word 6 bits 31-00 Number of words in real8 if > 32743, right adjusted if <= 32743 set zero.

Note

Pack and label may be equivalenced. n, the number of points in a grid is now in 32 bit id word 12.

Author

Ralph Jones

Date

1985-07-31

Definition at line 39 of file w3ai00.f.