TurboPFor: Bit bituil

lib/include_/bitutil.h

@@ -144,7 +144,8 @@ void bitzdec(unsigned char *in, unsigned n,                     unsigned esize);
 #define INT2FLOAT(ival, scalse, bias) ((ival - bias) / scale)
 
 //------- Lossy floating point transform: pad the trailing mantissa bits with zeros according to the error e (ex. e=0.00001)
-  #ifndef _NFLOAT16
+// must include float.h to use _Float16 (see icapp.c)
+ #ifdef FLT16_MAX 
 _Float16 _fprazor16(_Float16 d, float e, int lg2e);  
 void fprazor16(_Float16 *in, unsigned n, _Float16  *out, float  e);
   #endif