I don't know why you're being downvoted either, for what seems like a legitimate (if slightly oddly framed) question.
Linear algebra libraries have a ridiculously simple task (by comparison) that can be optimized incredibly well - there seems to be no amount of intellectual effort that isn't worth doing to get 1% more on BLAS given the massive usage of it.
By contrast, Hyperscan has many different subsystems (string search of various kinds, NFA, DFA, various quick SIMD 'lookarounds', different acceleration schemes for NFA and DFA etc. etc). It's enormously complicated already (and a bit unpredictable in terms of the different tradeoffs that can happen) and the optimizations have to be best-effort. We couldn't put in nearly as much optimization into any given hot loop as the linear algebra guys can.
That being said, we've done quite a lot, and many people have picked up useful techniques from our codebase. For example, the string search in Hacker News user burntsushi's regex scanner is an entirely legit (i.e. he freely credits where it comes from) clone of our "Teddy" literal matcher, for example.
Linear algebra libraries have a ridiculously simple task (by comparison) that can be optimized incredibly well - there seems to be no amount of intellectual effort that isn't worth doing to get 1% more on BLAS given the massive usage of it.
By contrast, Hyperscan has many different subsystems (string search of various kinds, NFA, DFA, various quick SIMD 'lookarounds', different acceleration schemes for NFA and DFA etc. etc). It's enormously complicated already (and a bit unpredictable in terms of the different tradeoffs that can happen) and the optimizations have to be best-effort. We couldn't put in nearly as much optimization into any given hot loop as the linear algebra guys can.
That being said, we've done quite a lot, and many people have picked up useful techniques from our codebase. For example, the string search in Hacker News user burntsushi's regex scanner is an entirely legit (i.e. he freely credits where it comes from) clone of our "Teddy" literal matcher, for example.