The library is much faster than PCRE (up to 10-13 times, according to Intel's bechmarks). It can be found here
http://intel.github.io/hyperscan/dev-reference/compilation.html#unsupported-constructs
And probably does so by dropping features. PCRE is available everywhere, hyperscan not so much. This would be better as a shard.
Most helpful comment
http://intel.github.io/hyperscan/dev-reference/compilation.html#unsupported-constructs