Floating-point operations in VDSP (bf561) (3): Flt_min

Flt_min is a constant defined in float.h to represent the minimum value of a single-precision floating-point number.

#define Flt_min 1.1754943508222875e-38f

So where does this value come from?

From the representation of floating point can know that the mantissa is necessarily greater than or equal to 1, to take the minimum, only the index can be set to the minimum value, because the floating point number of the index of 0 to represent a special floating-point number, so the index can only take 1, that is

0 00000001 00000000000000000000000

Sign bit exponent mantissa

The value of a 16-based integer to see it is 0x00 80 00 00.

According to the calculation formula of floating-point number, the value is:

1.0 * 2-126

= 1.1754943508222875079687365372222e-38

This is the calculation result of the calculator.