[Mlir-commits] [mlir] [mlir][docs] Add more examples for the "canonical form" (PR #173667)

[Renato Golin]

rengolin wrote:

> No: there can't be. While you may introduced some prerequisite for the purpose of transformations (like "codegen prepare" in LLVM), these can't be "canonical forms" in MLIR parlance, by definition of the canonical form being the set of patterns included in the canonicalizer. You need to adopt another terminology here.

Ack. For the lack of better terminology, "the set of patterns included in the canonicalizer" is ok for now. We then focus on discussing what set of patterns we want in there. We can call the various valid distinct forms in a particular dialect "normal forms".

https://github.com/llvm/llvm-project/pull/173667