Compatibility of Matlab with IEEE-754 1985 or 2008

Hi,

Is MATLAB compatible with IEEE-754 1985 or 2008 ?

-0 = +0 ? or -0 < +0

Best regards

C. VALPARD

Answer 1

[Steven Lord]

Is MATLAB compatible with IEEE-754 1985 or 2008 ?

There are a few operations for which the IEEE specs state what the answer should be but MATLAB returns something different. The most obvious example is the sqrt function. As stated in the IEEE Compliance section on its documentation page the IEEE spec says sqrt(aNegativeNumber) should return NaN. But since MATLAB has the capability to work with complex numbers, we return a complex answer.

There are a handful of functions that have a similar section in their documentation.

As for your specific questions related to plus and minus zero:

format hex
x = -0

x = 
   8000000000000000

y = 0

y = 
   0000000000000000

Note the differences in the sign bits.

isLT = x < y

isLT = logical
   0

isEQ = x == y

isEQ = logical
   1

-0 and +0 are equal.

isPosInfX = (1./x) > 0

isPosInfX = logical
   0

isPosInfY = (1./y) > 0

isPosInfY = logical
   1

But you can tell the sign bit by computing 1./zero and seeing if you get Inf or -Inf.