Dillon

I think both of those changes are expected, and are not necessarily better. The old way of sliding window will always do the entire warmup in one production, while the...

Does that reproducer need the second dimension? It looks like y is almost unrelated (except for the store_at). The one thing that is interesting/unique about this pipeline compared to many...

@jiawen it sounds like you expect this to be a shallow equality? This issue is about a much more subtle issue: when doing a deep equality check, we don't short...

I agree we should probably remove `array_ref::operator==`. Maybe we just don't need to provide a helper for this? If it's just `a.base() == b.base() && a.shape() == b.shape()`. Looks like...

This branch has some failing relevant tests: https://github.com/google/XNNPACK/actions/runs/15284031029/job/42989783184?pr=8497 Some of them are new after the merge, my apologies for the conflict if so. However, one of them was failing before...

We are currently working on deprecating gavgpool and replacing it with static_mean, which is a generalization of gavgpool. I suggest directing your efforts towards implementing rsum and rdsum (which implement...

If you merge master into this, we could at least consider merging this with `Dynamic` defaulted to false (and maybe not Browsable at first), then it might not be such...

I think maybe the best way to go is peel off parts of this into separate branches that can be merged one by one. For example, adding the ability for...

Also, there is quite a bit of stuff in this branch that I don't recognize from last time I previewed this change. IIRC, it looked pretty familiar from the way...

I've started looking at this issue, sorry for the delay. I think the thing that is causing problems here is C25. It's a capacitor at the upper right part of...