In the current paper we analyze several methods for generation of loss distribution for credit portfolios. Loss distributions play an important role in pricing of credit derivatives and in credit portfolio optimization. A loss distribution is a function of the number of entities in the portfolio, their credit ratings, the notional amount and recovery of each entity, default probabilities, loss given defaults, and the correlation/dependence structure between entities incorporated in the portfolio. Direct generation of loss distribution may require Monte Carlo simulation which is time consuming and is not effective when applied for credit portfolio optimization. To overcome computational complexity a number of approaches were undertaken based on assumptions imposed on the input parameters, goals of loss distributions generation, and the accepted level of tolerance and computational errors.