Abstract: To perform the generalized sensitivity analysis for nuclear data in a reactor physics design code, KYLIN-Ⅱ, the generalized perturbation theory is adopted and several generalized fix-source equations with the orthogonal definite condition need to be solved when the sensitivity coefficients are figured out. Besides, the paper develops a new approach, CMFD-based generalized fix-source equation solution, to accelerate the convergence. The convergence efficiency of the generalized fixed-source equation is improved by roughly 4.3 times, and the sensitivity coefficients calculated by the GPT accord with those calculated by the direct perturbation theory, which demonstrates the sensitivity analysis ability in KYLIN-Ⅱ.