This computationally intensive step is proving a bunch of inequalities originally calculated by floating point numerical routines. It does a software floating

This computationally intensive step is proving a bunch of inequalities originally calculated by floating point numerical routines. It does a software floating point emulation inside the theorem prover with proofs of bounds on the rounding at each step. Software floating point in an inefficient high-level language is slow!