1: \begin{abstract}

2: In 1986, Dixon and McKee

3: developed a discrete fractional Gr\"{o}nwall inequality

4: [Z. Angew. Math. Mech., 66 (1986), pp. 535--544],

5: which can be seen as a generalization of the classical discrete  Gr\"{o}nwall inequality.

6: However, this generalized discrete Gr\"{o}nwall inequality  and its variant

7: [SIAM J. Numer. Anal., 57 (2019), pp. 1524--1544] have not been widely

8: applied  in the numerical analysis of the time-stepping methods

9: for the time-fractional evolution equations.

10: The main purpose of this paper is to show how to apply the generalized discrete

11: Gr\"{o}nwall inequality to prove the convergence of a class of time-stepping

12: numerical methods

13: for time-fractional nonlinear subdiffusion equations,  including the popular

14: fractional backward difference type methods of order one and two,

15: and the  fractional Crank-Nicolson type methods.

16: We obtain the optimal $L^2$ error estimate in space discretization for multi-dimensional problems.

17: The convergence of the fast time-stepping numerical methods is also proved in a simple manner.

18: The present work  unifies the convergence analysis of several existing  time-stepping schemes.

19: Numerical examples are provided to verify the effectiveness of the present method.

20: \end{abstract}

21: