diff --git a/llvm/lib/Target/DirectX/CMakeLists.txt b/llvm/lib/Target/DirectX/CMakeLists.txt
--- a/llvm/lib/Target/DirectX/CMakeLists.txt
+++ b/llvm/lib/Target/DirectX/CMakeLists.txt
@@ -9,7 +9,6 @@
 add_llvm_target(DirectXCodeGen
   DirectXSubtarget.cpp
   DirectXTargetMachine.cpp
-  DXILBitcodeWriter.cpp
   DXILPointerType.cpp
   DXILPrepare.cpp
   PointerTypeAnalysis.cpp
diff --git a/llvm/lib/Target/DirectX/DXILPrepare.cpp b/llvm/lib/Target/DirectX/DXILPrepare.cpp
--- a/llvm/lib/Target/DirectX/DXILPrepare.cpp
+++ b/llvm/lib/Target/DirectX/DXILPrepare.cpp
@@ -12,6 +12,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "DirectX.h"
+#include "PointerTypeAnalysis.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/CodeGen/Passes.h"
@@ -25,6 +26,7 @@
 #define DEBUG_TYPE "dxil-prepare"
 
 using namespace llvm;
+using namespace llvm::dxil;
 
 namespace {
 
@@ -80,6 +82,7 @@
 class DXILPrepareModule : public ModulePass {
 public:
   bool runOnModule(Module &M) override {
+    PointerTypeMap PointerTypes = PointerTypeAnalysis::run(M);
     AttributeMask AttrMask;
     for (Attribute::AttrKind I = Attribute::None; I != Attribute::EndAttrKinds;
          I = Attribute::AttrKind(I + 1)) {
@@ -101,6 +104,70 @@
             Value *Zero = ConstantFP::get(In->getType(), -0.0);
             I.replaceAllUsesWith(Builder.CreateFSub(Zero, In));
             I.eraseFromParent();
+            continue;
+          }
+          // Only insert bitcasts if the IR is using opaque pointers
+          if (!M.getContext().hasSetOpaquePointersValue())
+            continue;
+
+          // Emtting NoOp bitcast instructions allows the ValueEnumerator to be
+          // unmodified as it reserves instruction IDs during contruction.
+          if (auto Inst = dyn_cast_or_null<LoadInst>(&I)) {
+            // Omit bitcasts if we only have one type
+            auto It = PointerTypes.find(Inst->getPointerOperand());
+            if (It != PointerTypes.end()) {
+              if (cast<TypedPointerType>(It->second)->getElementType() ==
+                  Inst->getType())
+                continue;
+            }
+            // Insert bitcasts at the top of the block...
+            Builder.SetInsertPoint(Inst->getParent(),
+                                   Inst->getParent()->begin());
+            Value *NoOpBitcast = Builder.Insert(CastInst::Create(
+                Instruction::BitCast, Inst->getPointerOperand(),
+                Inst->getPointerOperandType()));
+            Builder.SetInsertPoint(Inst);
+            Inst->replaceAllUsesWith(
+                Builder.CreateLoad(Inst->getType(), NoOpBitcast));
+            Inst->eraseFromParent();
+            continue;
+          }
+          if (auto Inst = dyn_cast_or_null<StoreInst>(&I)) {
+            // Omit bitcasts if we only have one type
+            auto It = PointerTypes.find(Inst->getPointerOperand());
+            if (It != PointerTypes.end()) {
+              if (cast<TypedPointerType>(It->second)->getElementType() ==
+                  Inst->getValueOperand()->getType())
+                continue;
+            }
+            // Insert bitcasts at the top of the block...
+            Builder.SetInsertPoint(Inst->getParent(),
+                                   Inst->getParent()->begin());
+            Value *NoOpBitcast = Builder.Insert(CastInst::Create(
+                Instruction::BitCast, Inst->getPointerOperand(),
+                Inst->getPointerOperandType()));
+            Builder.SetInsertPoint(Inst);
+            Inst->replaceAllUsesWith(
+                Builder.CreateStore(Inst->getValueOperand(), NoOpBitcast));
+            Inst->eraseFromParent();
+            continue;
+          }
+          if (auto Inst = dyn_cast_or_null<GetElementPtrInst>(&I)) {
+            // Omit bitcasts if we only have one type
+            auto It = PointerTypes.find(Inst->getPointerOperand());
+            if (It != PointerTypes.end()) {
+              if (cast<TypedPointerType>(It->second)->getElementType() ==
+                  Inst->getResultElementType())
+                continue;
+            }
+            // Insert bitcasts at the top of the block...
+            Builder.SetInsertPoint(Inst->getParent(),
+                                   Inst->getParent()->begin());
+            Value *NoOpBitcast = Builder.Insert(CastInst::Create(
+                Instruction::BitCast, Inst->getPointerOperand(),
+                Inst->getPointerOperandType()));
+            Inst->setOperand(0, NoOpBitcast);
+            continue;
           }
         }
       }
diff --git a/llvm/test/CodeGen/DirectX/conflicting-bitcast-insert.ll b/llvm/test/CodeGen/DirectX/conflicting-bitcast-insert.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/conflicting-bitcast-insert.ll
@@ -0,0 +1,14 @@
+; RUN: llc --filetype=asm %s -o - | FileCheck %s
+target triple = "dxil-unknown-unknown"
+
+define i64 @test(ptr %p) {
+  store i32 0, ptr %p
+  %v = load i64, ptr %p
+  ret i64 %v
+}
+
+; CHECK: define i64 @test(ptr %p) {
+; CHECK-NEXT: %1 = bitcast ptr %p to ptr
+; CHECK-NEXT: %2 = bitcast ptr %p to ptr
+; CHECK-NEXT: store i32 0, ptr %2, align 4
+; CHECK-NEXT: %3 = load i64, ptr %1, align 8
diff --git a/llvm/test/CodeGen/DirectX/omit-bitcast-insert.ll b/llvm/test/CodeGen/DirectX/omit-bitcast-insert.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/omit-bitcast-insert.ll
@@ -0,0 +1,32 @@
+; RUN: llc --filetype=asm %s -o - | FileCheck %s
+target triple = "dxil-unknown-unknown"
+
+define i64 @test(ptr %p) {
+  %v = load i64, ptr %p
+  ret i64 %v
+}
+
+; CHECK: define i64 @test(ptr %p) {
+; CHECK-NEXT: %v = load i64, ptr %p, align 8
+; CHECK-NEXT: ret i64 %v
+
+define i64 @test2(ptr %p) {
+  store i64 0, ptr %p
+  %v = load i64, ptr %p
+  ret i64 %v
+}
+
+; CHECK: define i64 @test2(ptr %p) {
+; CHECK-NEXT: store i64 0, ptr %p
+; CHECK-NEXT: %v = load i64, ptr %p, align 8
+; CHECK-NEXT: ret i64 %v
+
+define i32 @test3(ptr %0)  {
+  %2 = getelementptr i32, ptr %0, i32 4
+  %3 = load i32, ptr %2
+  ret i32 %3
+}
+
+; CHECK: define i32 @test3(ptr %0)  {
+; CHECK-NEXT: %2 = getelementptr i32, ptr %0, i32 4
+; CHECK-NEXT: %3 = load i32, ptr %2