[Mlir-commits] [mlir] [mlir][vector] Allow multi dim vectors in vector.scatter (PR #132217)

[Kunwar Grover]

Groverkss wrote:

> > Hi @banach-space , this pr as mentioned above depends on #132206 , which is why you may be seeing unrelated changes here. Github doesnt have the best way of having stacked prs, so it's hard to indicate this.
> 
> Please either document such things in the summary, or use https://llvm.org/docs/GitHub.html#using-graphite-for-stacked-pull-requests (I've not used it myself).
> 
> Also, if you are sending multiple dependent PRs to achieve a larger goal, could you create a GitHub issue to summarise and to communicate that? Otherwise it's hard to track what's going on (and to see the bigger picture). Specifically, I assume that these are related:
> 
> * [[mlir][vector] Allow multi dim vectors in vector.scatter #132217](https://github.com/llvm/llvm-project/pull/132217)
> * [[mlir][vector] Allow lowering multi-dim scatters to LLVM #132227](https://github.com/llvm/llvm-project/pull/132227)
> * [[mlir][vector] Decouple unrolling gather and gather to llvm lowering #132206](https://github.com/llvm/llvm-project/pull/132206)
> 
> ? If yes, what's the order and the dependencies? Could this be documented outside this PR?

I added a link to the discourse discussion. All PRs are listed there in order of dependency and how reviewers are expected to review them.

My personal opinion is that as long as it is clear for reviewers on what is happening, why it is happening, and how the changes are reviewed is given, it is enough. The discourse discussion will be updated as I send more prs and a link will be included in further prs.

Let's get back to technical discussion now. I'm happy to create a github issue or use graphite or anything that the reviewers prefer if it will make it easier for them to review it. For now, the discourse discussion link makes it clear.

Thanks for the review so far!

https://github.com/llvm/llvm-project/pull/132217