How to find One's complement of floating point variable ?

Hi,

I used Simulink Bitwise Not operator to find the one's complement of uint8 type variable.

Now I want to find one's complement of float variable (double) using simulink.

Can you please help me on how to perform the task.

Thank you

[James Tursa]

What does a 1's complement of a floating point bit pattern even mean to you? How do you intend to use this result?

[kintali narendra]

I need to convert float variable to binary number and then find one's complement of that, covert it into float variable again.

I am using it for redundancy validation.

Answer 1

[Walter Roberson]

typecast to uint64, take the one's complement, typecast back.

Caution: 1's complement of 0.0 is NaN.

[kintali narendra]

the simulink model has run successfully but the answer is same for every number.

[Walter Roberson]

you used a Convert block which is not appropriate.

https://www.mathworks.com/help/simulink/slref/datatypeconversion.html

You need to use a MATLAB Function Block.

    function out = fcn(u) 
    out = zeros(size(u));
    out = typecast(bitxor(typecast(u, 'uint64'), 0xffffffffffffffffu64), 'double');

[kintali narendra]

Thank you. it worked.

[Paul]

My version (2019a) doesn't support Walter's code. What result should be shown in the Display block when the input is 23?

[Walter Roberson]

    function out = fcn(u) 
    out = zeros(size(u));
    out = typecast(bitxor(typecast(u, 'uint64'), intmax('uint64')), 'double');

[kintali narendra]

Hi Paul, these are my answers.

For 23 it is -0.1953

For 93 it is -0.04834

[Paul]

When the input is 23, do you want the Display block to show -0.1953? In other words, is -0.1953 the correct answer for 23 with respect to what you are trying accomplish?

From what you've described, I thought the goal was to get the decimal representation of the one's complement of the binary representation of 23. Is that what -0.1953 is?

[kintali narendra]

yes, I want decimal representation of the one's complement.

The function which Walter recommened worked.

[Walter Roberson]

@Paul note in particular they wanted the one's complement of the binary representation of double precision numbers, not of an integer data type.