float to integer rounding

[Dean Schulze]

    When Zortech implemented IEEE-754 in their Engineering and Science
edition v3.0 they seem to have complied with the ANSI C standard for
implicit conversions and implemented explicit functions to comply with
IEEE-754.  Here is an example program with output:

/*
	File:			roundtest.cpp
	Written by:		Dean Schulze
	Last Modified:	12-21-93

		Tests default rounding mode in Zortech C++.
*/

#include	<fltpnt.h>
#include	<stdio.h>

main()
{
int i,j;
long rounded;
double dbli;

for (j=-5; j<=5; j++)
{
	dbli=j/3.0;
	i=dbli;
	rounded = rndtol(dbli);
	printf("\ndbli = %lf    rounded = %li    i = %i",dbli,rounded,i);
}

return 0;

}


dbli = -1.666667    rounded = -2    i = -1
dbli = -1.333333    rounded = -1    i = -1
dbli = -1.000000    rounded = -1    i = -1
dbli = -0.666667    rounded = -1    i = 0
dbli = -0.333333    rounded = 0     i = 0
dbli = 0.000000     rounded = 0     i = 0
dbli = 0.333333     rounded = 0     i = 0
dbli = 0.666667     rounded = 1     i = 0
dbli = 1.000000     rounded = 1     i = 1
dbli = 1.333333     rounded = 1     i = 1
dbli = 1.666667     rounded = 2     i = 1


    The rndtol() function uses the current rounding mode, which defaults
to round to nearest.  Zortech also provided a function to round to nearest
long which doesn't rely on (and presumably doesn't change) the  current
rounding mode.

    Does an implementation comply with IEEE-754 if the rounding mode is
only used for explicit function calls while the C standard is used for
implicit conversions?

Dean Schulze