When converting a negative floating point number to integer or long integer,
using the BCD float library, it loses its sign.
Using the IEEE library it rounds toward zero, retaining the sign.
RIDE BN 746

// Test of float -> int conversion
// compile with BCD floating point

#include 

int Itest;
float Ftest;

// Just to fool the optimiser that it's not a constant
float dummy( float x){ return x;}

main(){
   Ftest = dummy(-3.5);  
// With BCD floating point this loses its sign and gets +3
// With IEEE floating point it gets -3
   Itest = Ftest;   
}