[llvm] [DA] Introduce domain for monotonicity (PR #176367)

[Sjoerd Meijer via llvm-commits]

sjoerdmeijer wrote:

This is technically off-topic, but it is related to our discussions and how we have them. We have a nice land very active little loop opt workgroup, which is a massive improvement over some time ago, and we all want the same thing here, which is to get loop optimisations enabled. Therefore, this comment is a request to all "loop optimisation people" involved here (@kasuga-fj , @amehsan , @Meinersbur , @sjoerdmeijer), which is to see if we can get our communication more efficient, let me explain that. This is something that has been on my mind observing some other discussions, and is something I wanted to share.

Before I do that, let me actually first explain my approach here. After observing many discussions, I wanted:
- Fact finding: by requesting a write-up, I tried to get all information in one place so that we could start making a well-informed decision on an approach.
- Interpretation and consensus: after reading all presented facts, my interpretation/opinion was “let’s postpone monotonicity until we really need it”, that was my attempt to reach consensus on a way forward; this is what I am trying to achieve here.

With that out of the way, my request now is to see if we can improve the efficiency of our communication, so that hopefully, let's put it bluntly, we spend less time arguing. I have reached capacity of catching up on discussions and participating in more. 

In this discussion, I have tried to mediate as I am not involved in the implementation of either approaches; this was my intend anyway. But I can give an example from my own work as well to make Delinearization more accurate, here is a quote from this [comment](https://github.com/llvm/llvm-project/pull/175158#discussion_r2718407015):

> Regarding the altenatives, I think emitting runtime assumptions is worth to pursue. Not sure if this will be helpful, but here is my thoughts:
>  DA already has functionality for runtime predicates, but I don’t want to enhance it, at least until the existing bugs are fixed.

That is a very strong statement, and I don't know how to read that. We don't know when all bugs are fixed, so does this now means I am blocked to make progress if I do want to progress this? 

I believe this is fairly representative for other statements I have observed. Why is this put so strongly? Who decides this?  They are just some questions that I have. Therefore, my question to all is: can we try to find more consensus? Hopefully that makes everything easier, and less time-consuming. 

https://github.com/llvm/llvm-project/pull/176367