Papers like this have been written for older GPU architectures and it is disappointing to even have the need for them. That such a paper was presented at a manu

Papers like this have been written for older GPU architectures and it is disappointing to even have the need for them. That such a paper was presented at a manufacturer's conference is icing on the irony cake. My underdeveloped tinfoil hat speaks that they want to acknowledge the reverse engineering methods involved and are secretly hoping that they get applied to the competition, too, so they can extract the competitor's secrets from a public source instead of ending up in a reverse engineering grey area.