K 10
svn:author
V 6
bostic
K 8
svn:date
V 27
1989-08-15T11:53:45.000000Z
K 24
svn:entry:committed-date
V 27
1989-08-15T11:53:45.000000Z
K 7
svn:log
V 950
date and time created 89/08/15 11:53:45 by bostic
No VAX G_floating support was included in the "float.h" I just sent you.
All constants rely heavily upon a correctly-rounded decimal-binary
conversion routine; at least the following 6 constants *must* be converted
exactly to what the comments say they ought to be:

#define FLT_EPSILON	1.19209290E-07F			/* 2^-23 */
#define FLT_MIN		2.93873588E-39F			/* 2^-128 */
#define FLT_MAX		1.70141173E+38F			/* (1-2^-24)*2^127 */
#define DBL_EPSILON	2.775557561562891351E-17	/* 2^-55 */
#define DBL_MIN		2.938735877055718770E-39	/* 2^-128 */
#define DBL_MAX		1.701411834604692294E+38	/* (1-2^-56)*2^127 */

DBL_MAX/FLT_MAX may generate spurious floating-point overflows if your
decimal-binary conversion routine is deficient; likewise DBL_MIN/FLT_MIN
may underflow inadvertently to zero after conversion.

The constants of type float rely on the ANSI-invented suffix of "F" to
precisely specify their type.


END
