[Mlir-commits] [mlir] [MLIR] Remove incorrect loop step in scf parallel loop tiling pass. (PR #203798)

[Prem C]

silent-bytesmith wrote:

> This one I'm not so sure about - can you explain why this is the correct form?

Please correct me if I am wrong. But from my understanding, we are defining the inner loop parallelOp with a given step size (not a normalized one) on line 135 as below 
`auto innerLoop = ParallelOp::create(
      b, op.getLoc(), SmallVector<Value, 2>(newBounds.size(), zero), newBounds,
      op.getStep());`
   
  So the `inner_iv` already accounts for the step size. Hence the final index calculation doesn't need a step size.
  
 - `inner_iv + outer_iv` will give us the corresponding original index instead.

https://github.com/llvm/llvm-project/pull/203798