Inconsistent Results When Evaluating a Typecast and Reshape Operation in MATLAB

When I put a break point on this line and evaluate the section of typecast(reshape(uint32(4295010.2451024512 * (variable - 0)), [], 1), 'uint8') I get a array of [48 41 161 113]. However once i step through this line and check the values of tmArray(145:148) I get a array of [0 41 161 113] or sometime something like [48 40 161 113]. what could this be?

[Steven Lord]

What is the class of the variable named variable? What are that variable's contents?

[Walter Roberson]

I definitely would not hard-code 4295010.2451024512 in that line. It is a "magic number" that provides no information to the reader. At the very least, store the value in a documented variable.

I would also assign the result of the typecast to a tempory variable and then assign the variable into the array, so that the result could be checked. If the result in the temporary variable does not match the result that shows up in the array after the assignment, then that would be something that could be independently investigated.

[David]

Its class is a single and its content is a constant float number

[Walter Roberson]

Please show us num2hex(variable)

I am not getting a result anywhere close to what you are reporting.

I would suggest to you that taking variable - 0 is not a transparent way of converting variable to double precision. Just use that number - double(variable) for clarity.

I do not understand what the reshape() is doing there. The hard-coded constant is a scalar and variable is a scalar, so the result of the multiplication will be a scalar, and taking uint32() of the scalar is going to give a scalar. It does not make sense to me to reshape() a scalar.

Answer 1

[Bruno Luong]

Do NOT enter the value from the screen display of a number

format long,
u1=typecast(uint8([48 41 161 113]),'uint32');
s1=single(u1)/4295010.2451024512;
fprintf('%1.8f\n', s1)

443.86093140

fprintf('%1.25f\n', s1); % NOT 818.1404 as you pretend, so you make many mistake along the way

443.8609313964843750000000000

s2=443.86093139 % This is NOT s1

s2 = 
     4.438609313900000e+02

typecast(reshape(uint32(4295010.2451024512 * (s1 - 0)), [], 1), 'uint8')

ans = 1×4

0 41 161 113

typecast(reshape(uint32(4295010.2451024512 * (s2 - 0)), [], 1), 'uint8')

ans = 1×4

48 41 161 113

s1-s2

ans = single
   6.4843562e-09