How do you store a floating point number in x86 asm?

[fliguy84]

Hi guys, I'm currently working on my term project which requires I program a conversion software by using x86 assembly. I've managed to do the project by using just signed integers.

The question is, how do I store floating point number input from the user by just using simple 8086 instructions, without the floating point instruction of 486 and above? I'm using the emu8086 microprocessor emulator.

 

[DaveSimmons]

"A floating point library" is the normal answer. Back in the 6502 days we had to do that for integer multiplication and division too

If you're forced to create your own you can look at how numbers are stored for the IEEE standard used by x86.

Can you get by with fixed-point math, that is a fixed number of decimal digits? The old BCD scheme treated the lower two decimal digits of an integer as .00 - .99 so 123 = 1.23 and 4567 = 45.67.

 

[esun]

Can't offer much help but I can point you to the IEEE 754 spec for floating point representation: http://en.wikipedia.org/wiki/IEEE_754

Although, if you have to implement that yourself, you're in for a world of hurt I'd imagine.

 

[fliguy84]

Dave: Hmm, I haven't thought of using BCD. Thanks for the tips.

Esun: I'm quite familiar with the IEEE method, but I haven't thought of implementing it in a program. Yeah, it seems to take a lot of work. But, I'll try it too.

 

[Modelworks]

float.asm shows one way.
http://www.emu8086.com/dr/asm2...ce_code/float.asm.html

 

[fliguy84]

Originally posted by: Modelworks
float.asm shows one way.
http://www.emu8086.com/dr/asm2...ce_code/float.asm.html

Yeah, I've looked at it before . The example only shows how to output floating point number, not how to store it in the memory. Thanks for mentioning it though.

 

[schenley101]

you really just have to use bcd. it is the easiest without an extra library. It takes more memory though because each digit takes 4bits