Faster alternative to np.where with pandas dataframe?

Question

I have a dataframe df of the following shape:

         gvkey   fyear      x1     x2  ...        x43        x44       x45       x46
0         1000