The change for this ⚙ D125407 Prepare for inlining of SUM intrinsic is now submitted.
It is currently NOT enabled, we’ll get a patch in for that in a bit. See my recent post as to performance of this change.
I wish to thank everyone that has helped in contributing to this patch, whether it’s by suggesting things in this thread, adding review comments or answering on “How do I go about in MLIR?” on Slack. I’m still learning - which is a good thing! 