Machine Precision
Smallest floating point number u such that 1 + u > 1.
Used in:
- Convergence tests -- has an iteration converged?
- Tolerance tests -- is a matrix numerically singular?
- Error analysis -- how does floating point arithmetic affect results?