Ege Beysel

Same problem occurs with me as well, with the same `virtme-ng` and kernel/ubuntu versions. `virtme-ng 1.31` fixes it as well.

Hi, I'm also hitting the same issue. Correct me if I'm wrong, but I believe the redundant stack buffer allocations over here are an artifact of (iterative) consumer fusion generating...

Hey, thanks for your responses! > As mentioned in the other issue [#20785](https://github.com/iree-org/iree/issues/20785), we want to get rid of lowering config propagation. It implies that we will switch to `TileRootAndFuseProducerConsumer`....

The example I was talking about above does not happen just now, but rather if we switch the `TileAndDistributeToWorkgroupsUsingForallOpPass` to use the root op instead of the last compute op...

> However, I've been thinking to break this rule in `element->pack` cases because the transposition needs to be efficient. I guess I proposed the idea in some casual chat before,...

> We seem to have a similar issue here: [#20723](https://github.com/iree-org/iree/issues/20723) Btw. I also just took a look at this one, and I think this sort of dispatch is also something...

> For this work, you can add it right before vectorization because it is scoped within mmt4d pipeline for now. We will need the switch for other pipelines in the...

> We'll need it for all the pipelines, IMO. Alright, it's because of the flag that it's scoped to just the mmt4d pipelines. I meant that that pipeline over there...

Just leaving this here so it's easier to follow: https://github.com/iree-org/iree/pull/21181, I believe you're handling the case where we _would_ try and hoist the slice of IR out of the dispatch...

> This is a case where workgroup distribution failed to fuse in a pack: Question here: what would be the reason to end up with such a dispatch where the...