use_custom_solver.md

@@ -9,9 +9,9 @@ linsol = tbox.Pardiso() # or tbox.Dss(), or tbox.Mumps(), or tbox.SparseLu()
 ```
 If GMRES is used, some parameters must be defined,
 ```python
-linsol = tbox.Gmres(fill, drop, rst, tol)
+linsol = tbox.Gmres(fill, drop, rst, tol, mit)
 ```
-where `fill` is the fill-in factor and `drop` is the drop tolerance of the ILU preconditioner, and where `rst` is the number of restart and `tol` is the tolerance of the GMRES solver. To obtain best performance for flow solutions, use `fill=2`, `drop=1e-6`, `rst=50` and `tol=1e-5`.  
+where `fill` is the fill-in factor and `drop` is the drop tolerance of the ILU preconditioner, and where `rst` is the number of restar, `tol` is the tolerance and  `mit` is the maximum number of iterations of the GMRES solver. To obtain best performance for flow solutions, use `fill=2`, `drop=1e-6`, `rst=50`, `tol=1e-5` and `mit=200`.  
 The second step consists in choosing the nonlinear solution algorithm: Picard or Newton. The Newton solver is implemented for transonic flows and will yield the best performance, while the Picard solver works only with subcritical flows (subsonic only) and is mainly used for testing/debugging purposes. The Picard solver can be initialized using the dart::Picard class as,
 ```python
 solver = dart.Picard(pbl, linsol, rTol=1e-6, aTol=1e-8, mIt=25, nthrds=1, vrb=1)