Fix speed regression due to PySequence_Fast in for loop.
authorRobert Bradshaw <robertwb@math.washington.edu>
Thu, 24 Jul 2008 06:02:37 +0000 (23:02 -0700)
committerRobert Bradshaw <robertwb@math.washington.edu>
Thu, 24 Jul 2008 06:02:37 +0000 (23:02 -0700)
commitae9630e51b3f352320863c0a1f5e2dc5a023d1cf
tree516339a3548312705a43e45ee4366a6aeed141af
parent8958817ff5ca8ebc898296b7f02957a3662a3b64
Fix speed regression due to PySequence_Fast in for loop.

PySequence_Fast isn't so fast after all (due to multiple branching),
was a 35% slowdown for looping over lists. Now it handles tuples as
well with only about a 4% slowdown to lists compared to before.
Cython/Compiler/ExprNodes.py