Is there any expectation that floating point operations will always yield the same result on different machines?
In a speed test, I ran 3dDespike on a Sun v1280 (SPARC) and an AMD opteron box. The resulting 4D images are almost exactly the same but not identical. The differences are small and compatible with differences due to floating point operations, but I don't know enough about standard arithmetic to know if that's bad or not.