In this paper, we propose a technique for evaluating and analyzing the information sensitivity of computer algorithms, based on counting the number of basic operations. The existing methods of classical analysis of algorithms allow us to obtain analytically asymptotic estimates, and it is necessary to work with the algorithm itself, and not with software implementation, which limits the practical applicability of methods. Existing practical approaches to measuring the running time of the software implementation of the algorithm yield results that are strongly tied to the hardware and system software of the computational experiment. The evaluation of the complexity of the algorithm's software implementation proposed by the authors allows us to compare algorithms independently of the PC environment, but the dependence on the processor architecture is preserved.