From: Greg Colvin (greg_at_[hidden])
Date: 2000-07-23 17:49:39
Thanks for the good work.
Reading the motivating paper you reference at
http://daisy.uwaterloo.ca/~eddemain/papers/WADS99a/
I see that the authors claim that their data structures can
be easily extended to handle a deque interface. So I wonder
whether it would degrade the speed of your implementation to
provide a deque interface? The authors also say that they
would like to see their ideas incorporated into the standard
C++ library. So I also wonder if we should make contact with
the authors for their input?
Looking at your performance numbers I had to wonder why your
operator[] was so much faster than the deque you tested, and
why your push_back was so much slower. Is it fundamental to
the data structures used, or an accident of implementation?