我已将您原始问题的评论线程总结为答案。

以下是您可以尝试的几件事：

1. 增加每次迭代生成的 Arnoldi 向量 ( NCV ) 的数量。以下是DSAUPD的 ARPACK 文档对此的说明：

   At present there is no a-priori analysis to guide the selection of NCV relative to NEV (the number of eigenvalues you're looking for). The only formal requirement is that NCV > NEV. However, it is recommended that NCV >= 2*NEV.

   If many problems of the same type are to be solved, one should experiment with increasing NCV while keeping NEV fixed for a given test problem. This will usually decrease the required number of OP*x operations but it also increases the work and storage required to maintain the orthogonal basis vectors. The optimal "cross-over" with respect to CPU time is problem dependent and must be determined empirically.

2. 重新审视线性运算符的实现。在你的情况下，它应该是这样的：

   (a)$z \leftarrow A v$

   (b)使用预先计算的 M 分解对于。$M w = z$$w$$M$

3. 如果您正在寻找最小的特征值 - 或接近给定实际值的特征值，建议使用ARPACK 的移位和反转模式。shift可以为零（参见他们在此处给出的示例）或任何其他实数/复数值。您的线性运算符变为（对于对称和不定）。$\sigma$$(A - \sigma M)^{-1} M$$M$

4. 如果 SciPy-ARPACK 足够好地解决您的问题，那么 PETSc/SLEPc 可能是矫枉过正。