Topic : Pure Assembler Documentation Author : John Kormylo Version : PASM.HYP 1.0 Subject : Documentation/Pure Assembler Nodes : 740 Index Size : 20262 HCP-Version : 3 Compiled on : Atari @charset : atarist @lang : en @default : @help : Help @options : +g -i -s +x +z -t4 @width : 75 View Ref-File[ 68881/2 Floating Point Co-processor ] ... uses eight 96 bit data registers (FPn), a 16 bit control register (FPCR), a 32 bit status register (FPSR), and a 32 bit instruction address register (FPIAR). The FPU supports IEEE Single, Double, Extended and Packed BCD formats. It converts any format to its own internal format whenever it accesses RAM. When the exponent has its max value and the mantissa is zero, that number is interpreted as infinity. If the mantissa is NOT zero, then it is not a number (NAN). A "Signalling NAN" is a NAN with the most significant digit set (or the second most significant digit in extended format). NANs are generated when dividing zero by zero or infinity by infinity. Signalling NANs are never generated by the FPU. The FPU supports four rounding modes: round to nearest, round to zero, round to minus infinity, and round to plus infinity. These are set by the FPCR.