[DFAJumpThreading][NFC] Replace with const reference

llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp

@@ -993,14 +993,14 @@ struct TransformDFA {
     SmallPtrSet<BasicBlock *, 16> BlocksToClean;
     BlocksToClean.insert_range(successors(SwitchBlock));
 
-    for (ThreadingPath &TPath : SwitchPaths->getThreadingPaths()) {
+    for (const ThreadingPath &TPath : SwitchPaths->getThreadingPaths()) {
       createExitPath(NewDefs, TPath, DuplicateMap, BlocksToClean, &DTU);
       NumPaths++;
     }
 
     // After all paths are cloned, now update the last successor of the cloned
     // path so it skips over the switch statement
-    for (ThreadingPath &TPath : SwitchPaths->getThreadingPaths())
+    for (const ThreadingPath &TPath : SwitchPaths->getThreadingPaths())
       updateLastSuccessor(TPath, DuplicateMap, &DTU);
 
     // For each instruction that was cloned and used outside, update its uses
@@ -1017,7 +1017,7 @@ struct TransformDFA {
   /// To remember the correct destination, we have to duplicate blocks
   /// corresponding to each state. Also update the terminating instruction of
   /// the predecessors, and phis in the successor blocks.
-  void createExitPath(DefMap &NewDefs, ThreadingPath &Path,
+  void createExitPath(DefMap &NewDefs, const ThreadingPath &Path,
                       DuplicateBlockMap &DuplicateMap,
                       SmallPtrSet<BasicBlock *, 16> &BlocksToClean,
                       DomTreeUpdater *DTU) {
@@ -1263,7 +1263,7 @@ struct TransformDFA {
   ///
   /// Note that this is an optional step and would have been done in later
   /// optimizations, but it makes the CFG significantly easier to work with.
-  void updateLastSuccessor(ThreadingPath &TPath,
+  void updateLastSuccessor(const ThreadingPath &TPath,
                            DuplicateBlockMap &DuplicateMap,
                            DomTreeUpdater *DTU) {
     APInt NextState = TPath.getExitValue();