shooting sparrows with RPGs and shooting elephants with BB guns is equally foolharded.

There is a reason that Keil '51 only support 32 bit floating point and that reason is the '51 is singularily unsuited for high resolution maths.

Choose a suitable processor and, voila, you will find that the compiler support 64 bit floating point.

you can get an ARM for six bucks or less, so the "cost consideration" argument is totally invalid these days.

If you make your own function you may find that the "speed demon" (AT89C51ED2) you have chosen can do a 64 bit floating point calculation in about an hour or so of processing time.

Erik