I'm biased on this topic, as I work for Intel on regular expression matching, so I suppose I am 100% invested in the traditional Von Neumann architecture for this task.
I'd say the Micron Automata processor (which never really shipped from Micron in performant form - there were some "too slow to provide real numbers" units out there) is very interesting. But many of the applications for it in pattern matching as elsewhere are contingent on doing useful parallel work if they are to outcompete an optimized software implementation. So if you have a giant NFA where you need to run 16384 states all the time, the Micron AP looks great. However, if you can manage to find sleazy optimization tricks that allow you to cheat and get down to running the occasional 256-bit NFA over a few bits of your input, the AP isn't competitive.
More of the pattern matching task that we're familiar with (i.e. for network security) looks like the latter case (tricks being possible).
I'm very interested in seeing the tech get out there, but its been spun out into a startup and I don't know what the future looks like for it.
I'd say the Micron Automata processor (which never really shipped from Micron in performant form - there were some "too slow to provide real numbers" units out there) is very interesting. But many of the applications for it in pattern matching as elsewhere are contingent on doing useful parallel work if they are to outcompete an optimized software implementation. So if you have a giant NFA where you need to run 16384 states all the time, the Micron AP looks great. However, if you can manage to find sleazy optimization tricks that allow you to cheat and get down to running the occasional 256-bit NFA over a few bits of your input, the AP isn't competitive.
More of the pattern matching task that we're familiar with (i.e. for network security) looks like the latter case (tricks being possible).
I'm very interested in seeing the tech get out there, but its been spun out into a startup and I don't know what the future looks like for it.