[Mlir-commits] [mlir] [mlir][vector] Add more tests for ConvertVectorToLLVM (2/n) (PR #102203)

[Hugo Trachino]

nujaa wrote:

Alright, thanks for clarification.
One usecase I find is, if scalable vectors(resp. tiles) have been generated but the machine does not support scalable vectors (resp 2d scalable). Let's consider there are no sve/sme specific ops but e.g. `vector.outerproduct <[4]xf32> <[4]xf32>` . Should lower-to-llvm be able to lower scalable ops to fixed size or compiling without sve/sme/...  flags sets vscale to 1 and shall never build e.g. z registers ?
Mmmh, In any case, if vscale is not supported, there is no loop to be generated. So only remains the case we have 2d scalable vectors but only have SVE. That's quite niche.





> Btw, I really appreciate all the reviews and wanted to say that I'm always happy to return the favour. In particular, I'm reviewing all Vector tests and if you ever feel like taking a look at one of the tests yourself, here's a list that I use to co-ordinate this:
> 
> * https://github.com/orgs/llvm/projects/23
> 
> Sharing mostly so that you are aware :)

Thanks for your kind words. I have to deliver something downstream at present. I have a few features in mind for the short term future. I'll bookmark your status 😉 

https://github.com/llvm/llvm-project/pull/102203