[llvm] [LV] Use getFixedValue instead of getKnownMinValue when appropriate (PR #143526)

[David Sherwood via llvm-commits]

================
@@ -5342,6 +5342,10 @@ LoopVectorizationCostModel::getUniformMemOpCost(Instruction *I,
   StoreInst *SI = cast<StoreInst>(I);
 
   bool IsLoopInvariantStoreValue = Legal->isInvariant(SI->getValueOperand());
+  // TODO: We have tests that request the cost of extracting element
+  // VF.getKnownMinValue() - 1 from a scalable vector. This is actually
+  // meaningless, given what we actually want is the last lane and is likely
+  // to be more expensive.
----------------
david-arm wrote:

I did update the comment. See second commit in the PR:

```
  -// TODO: We have tests that request the cost of extracting element
  -// VF.getKnownMinValue() - 1 from a scalable vector. This is actually
  -// meaningless, given what we actually want is the last lane and is likely
  -// to be more expensive.
  +// TODO: We have existing tests that request the cost of extracting element
  +// VF.getKnownMinValue() - 1 from a scalable vector. This does not represent
  +// the actual generated code, which involves extracting the last element of
  +// a scalable vector where the lane to extract is unknown at compile time.
```

I'm not sure what you were expecting the updated comment to contain? You said `Might be good to adjust the comment a bit, we have tests to exercise this code, but might be clearer to just explain the problematic code path?` and I thought my updated comment was explaining the problematic code path? Which is that asking for the cost of extracting lane `VF.getKnownMinValue() - 1` does not represent the actual generated code, which for SVE at least would be whilelo + lastb instructions.

https://github.com/llvm/llvm-project/pull/143526