Suffix Array Vs. Compressed Suffix Array Vs. Bwt Search

Question

[Rob]

Hi all,

I was wondering if there is a good reference on what the (practically) fastest data structure for exact sequence search is. Specifically, I'm curious about the speed of finding all occurrences (exactly) of a string in a large corpus using backward search in a BWT versus search in a (presumably uncompressed) suffix array. I understand that they can both be performed asymptotically in O(m), where m is the length of the pattern string, with both structures. However, I'm curious if there are significant practical differences in the search speed. Any references or benchmarks would be greatly appreciated.

Thanks,

Rob