[llvm] [AMDGPU] Gate runtime unroll of LDS loops instead of overriding it (PR #222853)

[via llvm-commits]

================
@@ -51,9 +51,10 @@ static cl::opt<unsigned> UnrollThresholdIf(
   cl::init(200), cl::Hidden);
 
 static cl::opt<bool> UnrollRuntimeLocal(
-  "amdgpu-unroll-runtime-local",
-  cl::desc("Allow runtime unroll for AMDGPU if local memory used in a loop"),
-  cl::init(true), cl::Hidden);
+    "amdgpu-unroll-runtime-local",
+    cl::desc("Gate runtime unroll for AMDGPU by this knob if local memory used "
+             "in a loop"),
+    cl::init(true), cl::Hidden);
----------------
xgxanq wrote:

The old desc was inaccurate. Since I've changed the logic, I updated the desc.

https://github.com/llvm/llvm-project/pull/222853