diff --git a/llvm/lib/CodeGen/MachineOutliner.cpp b/llvm/lib/CodeGen/MachineOutliner.cpp
--- a/llvm/lib/CodeGen/MachineOutliner.cpp
+++ b/llvm/lib/CodeGen/MachineOutliner.cpp
@@ -59,7 +59,9 @@
 #include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/Statistic.h"
 #include "llvm/ADT/Twine.h"
+#include "llvm/Analysis/BlockFrequencyInfo.h"
 #include "llvm/Analysis/OptimizationRemarkEmitter.h"
+#include "llvm/Analysis/ProfileSummaryInfo.h"
 #include "llvm/CodeGen/LivePhysRegs.h"
 #include "llvm/CodeGen/MachineModuleInfo.h"
 #include "llvm/CodeGen/MachineOptimizationRemarkEmitter.h"
@@ -368,6 +370,16 @@
   void getAnalysisUsage(AnalysisUsage &AU) const override {
     AU.addRequired<MachineModuleInfoWrapperPass>();
     AU.addPreserved<MachineModuleInfoWrapperPass>();
+    if (!RunOnAllFunctions) {
+      // FIXME: This breaks some tests.
+      // TODO: As it is now, this code turns on the machine outliner for
+      // everyone (even those that don't target AArch64) that provides profile
+      // data unless they supply `-enable-machine-outliner=never`. The
+      // alternative is to add another option like
+      // `-enable-machine-outliner=profile_guided`.
+      AU.addRequired<BlockFrequencyInfoWrapperPass>();
+      AU.addRequired<ProfileSummaryInfoWrapperPass>();
+    }
     AU.setPreservesAll();
     ModulePass::getAnalysisUsage(AU);
   }
@@ -871,6 +883,10 @@
 
 void MachineOutliner::populateMapper(InstructionMapper &Mapper, Module &M,
                                      MachineModuleInfo &MMI) {
+  ProfileSummaryInfo *PSI = nullptr;
+  if (auto *Wrapper = getAnalysisIfAvailable<ProfileSummaryInfoWrapperPass>())
+    PSI = &Wrapper->getPSI();
+
   // Build instruction mappings for each function in the module. Start by
   // iterating over each Function in M.
   for (Function &F : M) {
@@ -891,7 +907,12 @@
 
     const TargetInstrInfo *TII = MF->getSubtarget().getInstrInfo();
 
-    if (!RunOnAllFunctions && !TII->shouldOutlineFromFunctionByDefault(*MF))
+    BlockFrequencyInfo *BFI = nullptr;
+    if (PSI && PSI->hasInstrumentationProfile() && F.hasProfileData())
+      BFI = &getAnalysis<BlockFrequencyInfoWrapperPass>(F).getBFI();
+
+    if (!RunOnAllFunctions && !TII->shouldOutlineFromFunctionByDefault(*MF) &&
+        !BFI)
       continue;
 
     // We have a MachineFunction. Ask the target if it's suitable for outlining.
@@ -917,6 +938,13 @@
       if (MBB.hasAddressTaken())
         continue;
 
+      // TODO: We could either avoid outlining hot blocks, or only consider
+      // outlining cold blocks.
+      // If profile data is available to the machine outliner, then do not
+      // outline hot blocks.
+      if (BFI && PSI->isHotBlock(MBB.getBasicBlock(), BFI))
+        continue;
+
       // MBB is suitable for outlining. Map it to a list of unsigneds.
       Mapper.convertToUnsignedVec(MBB, *TII);
     }
diff --git a/llvm/test/CodeGen/AArch64/machine-outliner-profile.ll b/llvm/test/CodeGen/AArch64/machine-outliner-profile.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/machine-outliner-profile.ll
@@ -0,0 +1,83 @@
+; RUN: llc -mtriple=aarch64-linux-gnu < %s | FileCheck %s
+
+declare void @z(i32, i32, i32, i32)
+
+; CHECK-LABEL: cold:
+define void @cold() !prof !20 {
+entry:
+; CHECK: [[OUTLINED:OUTLINED_FUNCTION_[0-9]+]]
+  tail call void @z(i32 1, i32 2, i32 3, i32 4)
+  ret void
+; CHECK: .cfi_endproc
+}
+
+; CHECK-LABEL: hot:
+define void @hot() !prof !21 {
+entry:
+; CHECK-NOT: [[OUTLINED]]
+  tail call void @z(i32 1, i32 2, i32 3, i32 4)
+  ret void
+; CHECK: .cfi_endproc
+}
+
+; CHECK-LABEL: minsize_cold:
+define void @minsize_cold() optsize minsize !prof !20 {
+entry:
+; CHECK: [[OUTLINED]]
+  tail call void @z(i32 1, i32 2, i32 3, i32 4)
+  ret void
+; CHECK: .cfi_endproc
+}
+
+; CHECK-LABEL: minsize_hot:
+define void @minsize_hot() optsize minsize !prof !21 {
+entry:
+; CHECK-NOT: [[OUTLINED]]
+  tail call void @z(i32 1, i32 2, i32 3, i32 4)
+  ret void
+; CHECK: .cfi_endproc
+}
+
+; CHECK-LABEL: no_profile:
+define void @no_profile() {
+entry:
+; CHECK-NOT: [[OUTLINED]]
+  tail call void @z(i32 1, i32 2, i32 3, i32 4)
+  ret void
+; CHECK: .cfi_endproc
+}
+
+; CHECK-LABEL: minsize_no_profile:
+define void @minsize_no_profile() optsize minsize {
+entry:
+; CHECK: [[OUTLINED]]
+  tail call void @z(i32 1, i32 2, i32 3, i32 4)
+  ret void
+; CHECK: .cfi_endproc
+}
+
+; CHECK: [[OUTLINED]]:
+; CHECK-SAME: // @{{.*}} Tail Call
+; CHECK:      mov     w0, #1
+; CHECK-NEXT: mov     w1, #2
+; CHECK-NEXT: mov     w2, #3
+; CHECK-NEXT: mov     w3, #4
+; CHECK-NEXT: b       z
+
+!llvm.module.flags = !{!1}
+!20 = !{!"function_entry_count", i64 0}
+!21 = !{!"function_entry_count", i64 100}
+
+!1 = !{i32 1, !"ProfileSummary", !2}
+!2 = !{!3, !4, !5, !6, !7, !8, !9, !10}
+!3 = !{!"ProfileFormat", !"InstrProf"}
+!4 = !{!"TotalCount", i64 200}
+!5 = !{!"MaxCount", i64 100}
+!6 = !{!"MaxInternalCount", i64 0}
+!7 = !{!"MaxFunctionCount", i64 100}
+!8 = !{!"NumCounts", i64 4}
+!9 = !{!"NumFunctions", i64 4}
+!10 = !{!"DetailedSummary", !11}
+!11 = !{!12, !13}
+!12 = !{i32 990000, i64 10, i32 4}
+!13 = !{i32 999999, i64 10, i32 4}