Decimal to Binary (ieee 754)

I want to create a code that converts decimal to ieee754. I do not want to use dec2bin so I do not know where to start. I want x to be my decimal number, eb to number of exponent bits and mb for my mantissa. The outputs are the convert value 'q' followed by s, e, m for sign, exponent, mantissa.

 function [q, s, e, m,] = GetRepresentationIEEE754(x, eb, mb)

[Geoff Hayes]

WhatIsMatlab - why are eb and mb variable inputs to your function? Aren't they fixed for IEEE 754 single and double precision formats?

[Walter Roberson]

num2hex() technically does not use dec2bin() :)

[WhatIsMatlab-]

That's a good point. I guess I could just us that. Still not really sure to begin though. But I just want decimal to binary. Yes, they are fixed but I wanted to be able to change them if need be.

[Guillaume]

Note that I wouldn't call my function conv as that name is already used by matlab for 1d convolution.

A better name would be GetRepresentationIEEE754.

[WhatIsMatlab-]

I will note that. I did not know that thank you. I mainly just want to use the ieee 754 format of converting decimal numbers. It does not always have to be 32 bit.

Answer 1

[Walter Roberson]

First you should convert the values to binary. You can do that with num2hex() to get hex that can be converted. Or you can use typecast() to reform the number into one of the unsigned integer data types.

Once you have the number in an unsigned integer data type, you can use bitget() or dec2bin() or de2bi() to convert the number into a binary form, from which you can extract individual groups of bits.

As a practical matter, I would point out that if you use dec2bin() on a uint64() then the number will be converted to double precision before it is converted to binary, and that is going to lose about 13 bits of value in the process (because uint64 are 64 bits and double precision can represent 53 bits.) I would therefore suggest that if you are planning to use dec2bin() that you do not typecast() to uint64 -- typecast to either uint8() or uint32() instead.

But personally I would probably typecast() to uint64, mod() to get the mantissa, bitshift() to remove the mantissa, mod() to get the exponent, bitshift() to get the sign...