Many applications have very specialized computing requirements that can be enabled and integrated with custom acceleration modules. For example, radar processing systems for driver assistance require the computation of many fast Fourier transforms every few microseconds and would overwhelm any programmable core. Devices built to support radar processing need a dedicated accelerator to offload these calculations. Another example is from machine learning, where inference engines can be quite heavy in terms of calculations per second and may also require acceleration to support. Adding custom acceleration offers specialized processing blocks that go beyond the capabilities of standard microprocessor cores, improving system performance in these applications 10 to 100 times without significant cost or power increases.