Necessity of PLY's table files

In the past I've found it annoying that PLY creates lex/yacc table files from wherever you happen to run the script.  The same is happening with pycparser, so I investigated if these tables actually speed up execution at all with this _particular_ grammar.

I timed the execution of examples\cdecl.py over 50 runs, and found that removing the tables (https://github.com/Syeberman/pycparser/commit/c2187621698c3352cf4a703431061079b0a78de1) actually made execution _faster_: 44.77 seconds over the 50 runs without tables versus 47.44 with (with tables already written out).  (Win7 Intel Core i7-2600 @ 3.4GHz)

It seems David Beazley himself has also questioned the necessity of these tables, as recently as a year ago:
http://comments.gmane.org/gmane.comp.python.ply/636

The only benefit I see for these tables is to allow Python's optimized mode to be used:
http://www.dabeaz.com/ply/ply.html#ply_nn38
I tested with -O (after cleaning __pycache__) and it works just fine, but -OO does indeed fail.  More investigation would be needed if -OO support was necessary (perhaps there's a way to keep docstrings for certain modules).

[]