[Mlir-commits] [mlir] [mlir][bufferization] Fix ownership computation of unknown ops (PR #70773)

Martin Erhart llvmlistbot at llvm.org
Wed Nov 1 01:24:54 PDT 2023


maerhart wrote:

> Only when the `to_tensor`+`func.return` gets bufferized, there would be a use-after-free, so this is a problem of the bufferization pass, not the deallocation pass.

Good point! Specifying `to_tensor` like that makes sense to me. Adding what you described here to the documentation would definitely be great to avoid incorrect uses.

https://github.com/llvm/llvm-project/pull/70773


More information about the Mlir-commits mailing list