Single Precision "Overflow" when Nearing Divide By Zero;10E-38 (38274)

SUMMARY

The code example below produces an "OVERFLOW" error at n = 38 for QuickBasic Versions 3.00, 4.00, and 4.00b, and Basic Compiler Versions 6.00 and 6.00b for MS-DOS and OS/2. The overflow occurs as you divide by ever larger numbers and approach the limits of the negative exponent for single precision. The program may overflow at different places in other versions of Basic. To work around the "Overflow" error, use at least one double precision variable or constant in the expression before assigning to the variable.

MORE INFORMATION

The following code example gives an "Overflow" error at n = 38:

FOR n = 1 TO 100
 x# = 1 / 10 ^ n
 PRINT n, x#
NEXT

In the above program, the expression 1/10^n is optimized to use single precision, since the most precise argument in the expression is n, which defaults to single precision.

To avoid the overflow of the negative single precision exponent, change n to double precision (n#). This forces the expression 1/10^n# to be stored in a double precision temporary storage area before being assigned to x#:

FOR n# = 1 TO 100  ' This program runs fine from n#=1 through 100.
 x# = 1 / 10 ^ n#
 PRINT n#, x#
NEXT