GPGPU Programming for Games and Science

CPU Computing 45

// x = quiet NaN

}

else

{

// x = signaling NaN

}

// p a y l o a d = t & NAN

PAYLOAD MASK

}

}

LISTING 2.11: The general decoding of floating-point numbers.

The pseudocode that extracts the sign, biased exponent, and trailing sig-

nificand may be encapsulated, as shown in Listing 2.12. The combination of

parts into a number may also be encapsulated. There is no reason to shift the

sign bit into the lowest-order bit.

void GetEncoding ( binaryN x , UInteger& sign , UInteger& biased ,

UInteger& tr ai lin g )

{

sign = (x . encoding & SIGN

MASK ) ;

b i a s e d = ( x . e n c o d i n g & BIASED

EXPONE NT MASK) >> NUM TRAILING BITS ;

t r a i l i n g = ( x .