Finding an optimal partition for a numerical integration using the trapezoidal rule

When a method of numerical integration is implemented on a computer, it is performed with a limited precision arithmetic. This causes a computing error propagated at the level of each elementary operation and affecting, like the method error, the final computed result.