Transforming SDFGsΒΆ

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.