[Numpy-discussion] Slicing slower than matrix multiplication?

[denis]

A general question, is there a collection of numpy code snippets as transformed by experts,
short of google site:mail.scipy.org/pipermail/numpy-discussion or the like ?

And a subquestion, does anyone have a list of algebraic identities for .T vstack etc ?
for a real example, to transform
     dots = np.vstack([ dot( x[j:j+4] .T, imatrix ) .T  for j in ...])

cheers
   -- denis