[llvm] [LV] Vectorize predicated loop-invariant loads via masked load + broadcast (PR #200361)

[Ramkumar Ramachandra via llvm-commits]

================
@@ -57,7 +58,7 @@ cl::opt<bool> llvm::PreferInLoopReductions(
 
 /// Note: This currently only applies to `llvm.masked.load` and
 /// `llvm.masked.store`. TODO: Extend this to cover other operations as needed.
-static cl::opt<bool> ForceTargetSupportsMaskedMemoryOps(
+cl::opt<bool> llvm::ForceTargetSupportsMaskedMemoryOps(
     "force-target-supports-masked-memory-ops", cl::init(false), cl::Hidden,
     cl::desc("Assume the target supports masked memory operations (used for "
              "testing)."));
----------------
artagnon wrote:

Just enclose it in namespace llvm {}?

https://github.com/llvm/llvm-project/pull/200361