Skip to content

Put sum factorization kernel implementations into functions

The amount of LOC in our generated operators will necessarily bring us into trouble at some point. Alternatively, each realization could spit out a loopy kernel (aka a method on the local operator) and the assembly method only sets up the input for that and calls it. Our current infrastructure with all the information being on a the SumfactKernel node is quite good for that: It can implement a hash, where hash equality means same function.