Wolfram Research, Inc.
int MLPutFloat(MLINK link, double x) puts the floating-point number x to the MathLink connection specified by link with a precision corresponding to the C type float.
The argument x is typically declared as float in external programs, but must be declared as double in MLPutFloat() itself in order to work even in the absence of C prototypes.
See notes for MLPutReal().
See The Mathematica Book: Section 2.12.5.