diff --git a/llvm/lib/Target/ARM/ARM.h b/llvm/lib/Target/ARM/ARM.h
--- a/llvm/lib/Target/ARM/ARM.h
+++ b/llvm/lib/Target/ARM/ARM.h
@@ -50,6 +50,7 @@
 FunctionPass *createARMOptimizeBarriersPass();
 FunctionPass *createThumb2SizeReductionPass(
     std::function<bool(const Function &)> Ftor = nullptr);
+FunctionPass *createMcountLoweringPass();
 InstructionSelector *
 createARMInstructionSelector(const ARMBaseTargetMachine &TM, const ARMSubtarget &STI,
                              const ARMRegisterBankInfo &RBI);
diff --git a/llvm/lib/Target/ARM/ARMTargetMachine.cpp b/llvm/lib/Target/ARM/ARMTargetMachine.cpp
--- a/llvm/lib/Target/ARM/ARMTargetMachine.cpp
+++ b/llvm/lib/Target/ARM/ARMTargetMachine.cpp
@@ -518,6 +518,7 @@
 }
 
 void ARMPassConfig::addPreEmitPass() {
+  addPass(createMcountLoweringPass());
   addPass(createThumb2SizeReductionPass());
 
   // Constant island pass work on unbundled instructions.
diff --git a/llvm/lib/Target/ARM/CMakeLists.txt b/llvm/lib/Target/ARM/CMakeLists.txt
--- a/llvm/lib/Target/ARM/CMakeLists.txt
+++ b/llvm/lib/Target/ARM/CMakeLists.txt
@@ -51,6 +51,7 @@
   ARMTargetMachine.cpp
   ARMTargetObjectFile.cpp
   ARMTargetTransformInfo.cpp
+  CustomCallLoweringPass.cpp
   MLxExpansionPass.cpp
   Thumb1FrameLowering.cpp
   Thumb1InstrInfo.cpp
diff --git a/llvm/lib/Target/ARM/CustomCallLoweringPass.cpp b/llvm/lib/Target/ARM/CustomCallLoweringPass.cpp
new file mode 100644
--- /dev/null
+++ b/llvm/lib/Target/ARM/CustomCallLoweringPass.cpp
@@ -0,0 +1,60 @@
+#include "ARM.h"
+#include "ARMBaseInstrInfo.h"
+#include "ARMSubtarget.h"
+#include "llvm/Pass.h"
+#include "llvm/CodeGen/MachineBasicBlock.h"
+#include "llvm/CodeGen/MachineFunction.h"
+#include "llvm/CodeGen/MachineFunctionPass.h"
+#include "llvm/Support/raw_ostream.h"
+
+#define DEBUG_TYPE "arm-custom-call-lowering"
+
+using namespace llvm;
+
+namespace {
+class McountLoweringPass : public MachineFunctionPass {
+ public:
+  static char ID;
+
+  McountLoweringPass() : MachineFunctionPass(ID) {}
+
+  virtual bool runOnMachineFunction(MachineFunction &MF) {
+
+
+    LLVM_DEBUG(dbgs() << MF.getName() << "\n");
+    for (MachineFunction::iterator FI = MF.begin(), FE = MF.end(); FI != FE; ++FI) {
+      for (MachineBasicBlock::iterator BBI = FI->begin(), BBE = FI->end(); BBI != BBE; ++BBI) {
+        if (BBI->isCall()) {
+          LLVM_DEBUG(dbgs() << *BBI << "\n");
+          for (MachineInstr::const_mop_iterator II = BBI->operands_begin(), IE = BBI->operands_end(); II != IE; ++II )
+            if (II->isGlobal() && II->getGlobal()->getName() == "\01__gnu_mcount_nc") {
+              const ARMSubtarget &ST = MF.getSubtarget<ARMSubtarget>();
+              const ARMBaseInstrInfo &TII = *ST.getInstrInfo();
+              DebugLoc DL;
+              if (ST.isThumb()) {
+              BuildMI(*FI, BBI, DL, TII.get(ARM::tPUSH))
+                  .add(predOps(ARMCC::AL))
+                  .addReg(ARM::LR);
+              } else {
+                BuildMI(*FI, BBI, DL, TII.get(ARM::STMDB_UPD))
+                    .addReg(ARM::SP, RegState::Define)
+                    .addReg(ARM::SP)
+                    .add(predOps(ARMCC::AL))
+                    .addReg(ARM::LR);
+              }
+              return true;
+            }
+        }
+      }
+    }
+    return false;
+  }
+};
+}
+
+FunctionPass *llvm::createMcountLoweringPass() {
+  return new McountLoweringPass();
+}
+
+char McountLoweringPass::ID = 0;
+static RegisterPass <McountLoweringPass> X ("arm-mcount-lowering", "ARM mcount Call Lowering Pass");
diff --git a/llvm/test/CodeGen/ARM/mcount.ll b/llvm/test/CodeGen/ARM/mcount.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/ARM/mcount.ll
@@ -0,0 +1,14 @@
+; RUN: llc -mtriple=armv7a-linux-gnueabihf %s -o - | FileCheck %s --check-prefix=CHECK-ARM
+; RUN: llc -mtriple=thumbv7a-linux-gnueabihf %s -o - | FileCheck %s --check-prefix=CHECK-THUMB
+
+define dso_local i32 @foo(i64) local_unnamed_addr #0 {
+; CHECK-ARM:        stmdb   sp!, {lr}
+; CHECK-ARM-NEXT:   bl      __gnu_mcount_nc
+; CHECK-THUMB:      push    {lr}
+; CHECK-THUMB-NEXT: bl      __gnu_mcount_nc
+  %2 = mul nsw i64 %0, %0
+  %3 = trunc i64 %2 to i32
+  ret i32 %3
+}
+
+attributes #0 = { nofree nounwind "instrument-function-entry-inlined"="\01__gnu_mcount_nc" }