[llvm] [AMDGPU] Introduce "amdgpu-uniform-intrinsic-combine" pass to combine uniform AMDGPU lane Intrinsics. (PR #116953)

[Sameer Sahasrabuddhe via llvm-commits]

================
@@ -41,39 +41,50 @@ using namespace llvm;
 using namespace llvm::AMDGPU;
 using namespace llvm::PatternMatch;
 
-/// Map for newly created IR instructions and their uniformity.
-using NewUniformityMap = DenseMap<const Value *, bool>;
+/// Tracks uniformity of newly created instructions.
+/// Wraps a ValueMap so we can enforce consistent mark/erase usage.
----------------
ssahasra wrote:

Where's the ValueMap? I don't think you need this new struct at all. Just create an instance of ValueMap and use it directly. The whole point is that it automatically updates itself.

https://github.com/llvm/llvm-project/pull/116953