The Sparse Abstract Machine

We propose the Sparse Abstract Machine (SAM), an intermediate representation and abstract machine model for targeting sparse tensor algebra to reconfigurable and fixed-function spatial dataflow accelerators. SAM defines a streaming abstraction with sparse primitives that encompass a large space of scheduled tensor algebra expressions. SAM dataflow graphs naturally separate tensor formats from algorithms and is expressive enough to incorporate many sparse-iteration and hardware-specific optimizations. We show an automatic compilation technique from a high-level language to SAM and a set of hardware primitives which implement it. We evaluate the generality and extensibility of our sparse abstract machine, explore the performance space of sparse tensor algebra optimizations using SAM, and provide an example implementation of our SAM architecture.