Scoring mechanism for automatically generated test programs

A method of generating tests for a data processing apparatus comprises simulating functional states (for example using an RTL or FPGA simulator) of the data processing apparatus occurring during simulated execution of a population of program instructions formed of a plurality of instances of ordered lists of program instructions, noting the simulated occurrence of predetermined functional states, calculating respective functional state score values for the predetermined functional states in dependence upon the occurrence of the predetermined functional states during the simulated execution of the population, forming a population score value dependent upon a combination of the functional state score values, mutating one or more of the instances to form one or more candidate instances and replacing at least one of the instances of the population with one of the candidate instances to form a candidate population. These stages are then repeated in respect of the candidate population to form a candidate population score value. The candidate population score value is then compared with the population score value. If the candidate score value indicates that simulated execution of the candidate population results in the simulated occurrence of a higher number if predetermined functional states then the population is replaced with the candidate population. These latter steps are repeated until some predetermined termination conditions are met.