(j3.2006) decimal floats

[Andy Vaught]

On Sat, 12 Jan 2008, Craig Dedo wrote:

>       Andy
> 
> 	Well, the most widely used chip of all, the Intel 80x87 and its
> successors, has two different radices.  It has both binary floating point
> and Binary Coded Decimal (BCD) in hardware and has the hardware instructions
> to do arithmetic in both binary floating point and BCD.  
> 
> 	I strongly disagree about the suggested motivation for support for
> decimal floating point.  It seems reasonable to conclude that the two
> radices address the needs of different markets.  Thus, instead of being
> contradictory, they very much could be complementary.  

  I'm afraid not.  The 387 can load and store BCD integers (not floats),
but the internal representation is always base two.  A BCD load ends up
being about five times slower than a binary load.

     Andy