Closed nicolasvasilache closed 4 years ago
https://github.com/tensorflow/mlir/commit/a1e4422bf34c74b682b5f63465693974d7aa8028 implements the proper machinery.
The place that has to be implemented for AffineForOp is centered around this LoopNestBuilder: https://github.com/tensorflow/mlir/blob/master/lib/Dialect/Linalg/Transforms/LinalgToLoops.cpp#L438
@jsetoain @tetuante
Done in https://reviews.llvm.org/D72094
Once the supporting commit lands (I'll update this message when it does), we can finally implement lowering of Linalg to affine loops and affine load/store.