An example of data flow analysis in MLIR
Recently, a dataflow analysis framework was merged into MLIR. In this talk I will explain how to use the framework, using the stack arrays pass in Flang as an example.

    
Extensible and Composable Dataflow Analysis in MLIR
Dataflow analysis is a core component of any compiler infrastructure. LLVM has the Attributor, and Clang has static analyzers. In this talk, we discuss our addition of a general-purpose dataflow analysis framework to MLIR designed according to MLIR’s core principles and needs: extensible, composable, and debuggable. We present our mathematical formulation of transparently composing dataflow analyses and how it can be used to build libraries of analyses both in-tree and out-of-tree.