0.0f != - x
That is, IEEE negative zero is not properly accounted for. In this case, this is handled by the compiler rather than a library so it has to be addressed in the compiler itself.
Attached you'll find the original test program and an excerpt from the generated assembly
Log in to post a comment.