[Mlir-commits] [mlir] Remove Pure attribute from Linalg::IndexOp. (PR #68894)

llvmlistbot at llvm.org llvmlistbot at llvm.org
Mon Oct 16 21:38:48 PDT 2023


MaheshRavishankar wrote:

> Seems like a load to me, but we don't want to prove that the value is constant during the region execution. I don't think we have anything that easily expresses this though.

Why is it a load. It resolves to induction variable value when the linalg op is lowered to loops.

> 
> All-in-all it does not change much to the problem: we shouldn't "lie" to the system about semantics just to enable generic transformations!

Well, it has no side-effects. So not sure it is a lie.


https://github.com/llvm/llvm-project/pull/68894


More information about the Mlir-commits mailing list