Once created, stateful dataflow multigraphs can be transformed for performance, offloading to different architectures, and reducing elements for simplification and further transformations.
As a white-box approach, SDFG transformations can be written externally, and can be applied on graphs programmatically or interactively, through built-in passes, the transformation Python API, or the Visual Studio Code plugin.
Below you can find the different ways transformations are represented in DaCe and how they are used in the framework.
passes transformations simplify auto_optimize