[llvm] [RISCV] Introduce VLOptimizer pass (PR #108640)

[Michael Maitland via llvm-commits]

michaelmaitland wrote:

As I am working on responding to reviews, one thing has become clear. We need better tests for this. I'm not sure what the best way to test things are. It feels like there are two things here:

1. getOperandInfo
2. the actual VL optimization

I wonder if it would be a good idea to add unit tests for `getOperandInfo`. We would go through each operand of each instruction and verify that it had the expected OperandInfo.

Then for the actual VL optimization, I think it would be best to use lit tests to get code coverage of all paths not (not including all paths through getOperandInfo since that would be tested on its own).

Does anyone have different ideas or does this sound okay?

https://github.com/llvm/llvm-project/pull/108640