The testing for D20256 included a table showing the number of cycles consumed in breaking a reservation with first and last pages, only, allocated, with and without that change in place. The first column was the original code, and the second showed the effects of that change. I'm adding a third column that shows the affect of this change.
Pages 0 and 511 allocated, the rest free:
cycles/break: 4233: 4226: 3629
cycles/break: 4236: 4251: 3777
cycles/break: 4218: 4204: 3781
cycles/break: 4248: 4219: 3775
cycles/break: 4241: 4233: 3813
cycles/break: 4195: 4232: 3785
cycles/break: 4251: 4280: 3794
cycles/break: 4193: 4222: 3778
cycles/break: 4281: 4221: 3762
cycles/break: 4240: 4202: 3761