Abstract:
Informally, a language L has speedup if, for any Turing machine (TM) for L, there exists one that is better. Blum showed that there are computable languages that have almost-everywhere speedup. These languages were unnatural in that they were constructed for the sole purpose of having such speedup. We identify an intuitive condition which, like several others in the literature, implies that accepting any coNP-complete language has an infinitely-often (i.o.) superpolynomial speedup. We also exhibit a natural problem which unconditionally has a weaker type of i.o. speedup based upon whether the full input is read. Neither speedup pertains to the worst case.

Changes to previous version:
Drops the claim that there is a p-optimal TM accepting any NP-complete problem.

Abstract:
Informally, a language L has speedup if, for any Turing machine for L, there exists one that is better. Blum showed that there are computable languages that have almost-everywhere speedup. These languages were unnatural in that they were constructed for the sole purpose of having such speedup. We identify a condition apparently only slightly stronger than P!=NP which implies that accepting any coNP-complete language has an infinitely-often (i.o.) superpolynomial speedup and NP!=coNP. We also exhibit a natural problem which unconditionally has a weaker type of i.o. speedup based upon whether the full input is read. Neither speedup pertains to the worst case.