[Mlir-commits] [mlir] [mlir][vector] Disallow vector broadcast along scalable dim (PR #212197)

[llvmlistbot at llvm.org]

================

----------------
Brythzz wrote:

> What about examples like this:
> 
> ```mlir
> %0 = vector.broadcast %arg0 : vector<2xf32> to vector<3x[4]x2xf32>
> ```
> 
> ?

Thank you for your comments.

The pattern would generate 3 `vector.insert` ops in the first lowering step and then fail on the second iteration when trying to broadcast `vector<2xf32>` to `vector<[4]x2xf32>`.

https://github.com/llvm/llvm-project/pull/212197