Index: lib/Transforms/IPO/PartialInlining.cpp
===================================================================
--- lib/Transforms/IPO/PartialInlining.cpp
+++ lib/Transforms/IPO/PartialInlining.cpp
@@ -331,7 +331,7 @@
   // Compute the 'InlineCost' of block BB. InlineCost is a proxy used to
   // approximate both the size and runtime cost (Note that in the current
   // inline cost analysis, there is no clear distinction there either).
-  static int computeBBInlineCost(BasicBlock *BB);
+  static int computeBBInlineCost(BasicBlock *BB, bool IsEntryBB = false);
 
   std::unique_ptr<FunctionOutliningInfo> computeOutliningInfo(Function *F);
   std::unique_ptr<FunctionOutliningMultiRegionInfo>
@@ -831,7 +831,7 @@
 // TODO: Ideally  we should share Inliner's InlineCost Analysis code.
 // For now use a simplified version. The returned 'InlineCost' will be used
 // to esimate the size cost as well as runtime cost of the BB.
-int PartialInlinerImpl::computeBBInlineCost(BasicBlock *BB) {
+int PartialInlinerImpl::computeBBInlineCost(BasicBlock *BB, bool IsEntryBB) {
   int InlineCost = 0;
   const DataLayout &DL = BB->getParent()->getParent()->getDataLayout();
   for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I) {
@@ -852,6 +852,11 @@
       break;
     }
 
+    // PHIs in the entry BB to the region will be moved outside the region by
+    // CodeExtractor, so skip them when computing the region cost.
+    if (IsEntryBB && isa<PHINode>(I))
+      continue;
+
     IntrinsicInst *IntrInst = dyn_cast<IntrinsicInst>(I);
     if (IntrInst) {
       if (IntrInst->getIntrinsicID() == Intrinsic::lifetime_start ||
@@ -1185,7 +1190,7 @@
   std::vector<BasicBlock *> ToExtract;
   ToExtract.push_back(ClonedOI->NonReturnBlock);
   OutlinedRegionCost +=
-      PartialInlinerImpl::computeBBInlineCost(ClonedOI->NonReturnBlock);
+      PartialInlinerImpl::computeBBInlineCost(ClonedOI->NonReturnBlock, true);
   for (BasicBlock &BB : *ClonedFunc)
     if (!ToBeInlined(&BB) && &BB != ClonedOI->NonReturnBlock) {
       ToExtract.push_back(&BB);
Index: test/Transforms/CodeExtractor/PartialInlineEntryPHICost.ll
===================================================================
--- /dev/null
+++ test/Transforms/CodeExtractor/PartialInlineEntryPHICost.ll
@@ -0,0 +1,40 @@
+; RUN: opt < %s -partial-inliner -S | FileCheck %s
+; RUN: opt < %s -passes=partial-inliner -S | FileCheck %s
+
+; Check that we do not overcompute the outlined region cost, where the PHIs in
+; the outlined region entry (BB4) are moved outside the region by CodeExtractor.
+
+define i32 @bar(i32 %arg) {
+bb:
+  %tmp = icmp slt i32 %arg, 0
+  br i1 %tmp, label %bb1, label %bb2
+
+bb1:
+  br i1 undef, label %bb4, label %bb2
+
+bb2:                                              ; preds = %bb, %bb1
+  br i1 undef, label %bb4, label %bb5
+
+bb4:                                              ; preds = %bb1, %bb2
+  %xx1 = phi i32 [ 1, %bb1 ], [ 9, %bb2 ]
+  %xx2 = phi i32 [ 1, %bb1 ], [ 9, %bb2 ]
+  %xx3 = phi i32 [ 1, %bb1 ], [ 9, %bb2 ]
+  tail call void (...) @foo() #2
+  br label %bb5
+
+bb5:                                              ; preds = %bb4, %bb2
+  %tmp6 = phi i32 [ 1, %bb2 ], [ 9, %bb4 ]
+  ret i32 %tmp6
+}
+
+declare void @foo(...)
+
+define i32 @dummy_caller(i32 %arg) {
+bb:
+; CHECK-LABEL: @dummy_caller
+; CHECK: br i1
+; CHECK: br i1
+; CHECK: call void @bar.1.
+  %tmp = tail call i32 @bar(i32 %arg)
+  ret i32 %tmp
+}