The ability to regenerate the CPU assembly or GPU code according to profiling. From there, the Essence® Agent is able to tradeoff memory stalls vs arithmetic operations- or to best manage dependencies from previous results, and more. The Agent generates, auto-tunes, auto-scales, and auto-synchronizes 100%-parallelized code for management of resources [multi-core CPUs and GPUs, memory, storage, I/O, power resources, and other services].

Automates the tuning of controls for visual, audio, physics-simulation, or other sensory and/or calculation based services.

Treats data and code like ingredients to be processed by a unified formula processor. A wide variety of signal transforms can be performed in parallel.

The ability to decide the amount and timing of work done to make better Scheduling estimates for similar future work and code generation. Realtime profiling costs us nearly nothing in performance cycles yet accumulates effective intelligence to estimate future work costs, particularly with varying amounts of data.

The ability to split work apart for many processors/co-processors, potentially creating dependency chains or duplicating necessary overlaps in split data to reach optimal performance. We have also added a scheduler subsystem that partitions spatially for 3D data fields.