Many new domains for genetic programming require evolved programs to be
executed for longer amounts of time.  For these applications it is likely
that some test cases optimally require more computation cycles than others.
Therefore, programs must dynamically allocate cycles among test cases in
order to use computation time efficiently.  To elicit the strategic
allocation of computation time, we impose an {\it aggregate computation
time ceiling} that applies over a series of fitness cases.  This exerts
time pressure on evolved programs, with the effect that resulting programs
dynamically allocate computation time, opportunistically spending less time
per test case when possible, with minimal damage to domain performance.
This technique is in principle extensible to resources other than
computation time such as memory or fuel.  We introduce the game {\it
Tetris} as a test problem for this technique.