Thanks to all who replied! After experimenting a bit I'm now using the Boyer-Moore-Horspool algorithm sent by Kimmo Takkunen and I'm glad to report that the performance of my application has significantly increased.
When I first checked the B-M algorithm before sending my question to the mailing list I obviously didn't look closely enough and missed the fact that I can build the skip table _before_ I enter the time-critical part of my app - which is a huge gain. Thanks again, Edi.
