[Mlir-commits] [mlir] [mlir] Expose loadAllAvailableDialects to mlir-opt (PR #86374)

[Matteo Franciolini]

mfrancio wrote:

> This was removed for good reasons: I don't quite understand the motivation for this right now, you'll need to elaborate on the use case.

Thanks for taking a look. I understand the good reasons for removing this. We have however an internal usecase though for which this seems to be useful. Consider for example a unit test that leverages the one shot bufferize pass. The test has tensor, func, and let's say dialect X. When bufferizing, tensor bufferizes with memref ops, and dialect X bufferizes with dialect Y ops. The memref dialect is loaded by the one shot bufferization as part of the dependent dialects, but who is responsible to load Y?

In production this is fine as our pipeline guarantees to load Y before bufferization, but this does not hold for the unit test.

The alternatives to this to have an out of tree implementation of mlir-opt just for our convenience, but i would be happy to avoid it. I would also like to avoid refactoring our bufferization pipeline.

I hope you find this arguments compelling.

https://github.com/llvm/llvm-project/pull/86374