The asymptotic behavior of four coding schemes is evaluated by the norm logQ/log(1/Pe), where Q is an approximation of the amount of computation for decoding, and Pe is the probability of error. The four schemes considered are: (1) Maximum-likelihood decoding, (2) Convolutional codes with syndrome decoding, (3) BCH codes, (4) Finite state and convolutional codes with sequential decoding. Documentation and a listing are included for a computer program which maximizes the frequently used function Eo(8,r) with respect to r for general discrete memoryless channels.