4.8 For More Information

Donald Knuth's The Art of Computer Programming, Volume Two: Seminumerical Algorithms , provides an in-depth discussion of floating-point arithmetic and floating-point formats. This book is required reading for someone who wants to fully understand how floating-point arithmetic operates. Also, Intel's documentation on its Pentium processors explains its floating-point formats, exceptional conditions, and other issues related to the use of its FPU. Likewise, the manufacturer's literature for any CPU that supports floating-point arithmetic will explain the specifics for the use of that CPU's floating-point unit.

Those interested in trapping floating-point exceptions from a high-level language will need to check their language vendor's documentation to determine how this is done. Unfortunately, there are few standards around, so most compiler vendors use a proprietary scheme or a CPU- or OS-dependent scheme to trap these exceptions. For general information about the effects of precision and accuracy in floating-point calculations, a good textbook on numerical analysis would be a reasonable place to start.

Finally, The Art of Assembly Language (No Starch Press) contains lots of additional information related to floating-point arithmetic including the implementation ofvarious transcendental and other functions. The UCR Standard Library for 80x86 Assembly Language Programmers (a software package I've developed that is available at http://webster.cs.ucr.edu; check out the 'Assembler Tools' link and look under MASM) contains a full software-based floating-point package in 16-bit 8086 assembly language. The HLA Standard Library includes source code for several FPU support routines, including floating-point I/O and conversion. Check out the Webster website for more details (look under 'HLA' when following the 'Assembler Tools' link).