Should one do this analysis also in two phases, as in, first look for upperBound-lowerBound, as that is a smaller number it may be quicker and more states will be zero, which is good for the first phase, and for the second phase, we can then use the actual zeros from the first phase with again a reduced number of iterations.
I do think the current implementation is also correct, just potentially a bit slower (caching effects may also invalidate anything i said)...
Originally posted by @sjunges in #916 (comment)
Should one do this analysis also in two phases, as in, first look for upperBound-lowerBound, as that is a smaller number it may be quicker and more states will be zero, which is good for the first phase, and for the second phase, we can then use the actual zeros from the first phase with again a reduced number of iterations.
I do think the current implementation is also correct, just potentially a bit slower (caching effects may also invalidate anything i said)...
Originally posted by @sjunges in #916 (comment)