Echelon Neuron C - Manual (Page 87)

Neuron C Reference Guide 67
Conversion of ASCII String to Floating-
Point
Table 27 lists the conversion functions for an ASCII srtring to a floating-point
number.
Table 27. Conversions of ASCII Strings to Floating-Point Numbers
Short
Name Function
from
ascii void fl_from_ascii(const char *arg1, float_type *arg2);
Converts an ASCII string to a floating-point number. The conversion
stops at the first invalid character in the input buffer—there is no error
notification. The acceptable format is the following:
[+/-][xx][.][xxxxx][E/e[+/-]nnn]
Examples:
0, 1, .1, 1.2, 1E3, 1E-3, -1E1
There should be no more than nine significant digits in the mantissa
portion of the number, or else the results are unpredictable. A
significant digit is a digit following any leading zeroes.
Embedded spaces within the number are also not allowed. This routine
uses repeated multiplication and division, and can be time-consuming,
depending on the input data.
Examples:
0.00123456789E4 // is acceptable
123.4567890 // is not acceptable
123 E4 // is not acceptable
The value 123.4567890 is not acceptable because it has 10 significant
digits, and the value 123 E4 is not acceptable because it has an
embedded space.
Floating-Point Performance
Table 28 lists times in milliseconds for the various functions in the floating-point
library. They were measured using a Series 3100 Neuron Chip with a 10 MHz
input clock. These values scale with faster or slower input clocks. The
measurements are maximums and averages over random data logarithmically
distributed in the range 0.001 to 1,000,000.
Table 28. Floating-Point Function Performance
Function Maximum Average
Add 0.56 0.36
Subtract 0.71 0.5