Add DXIL Bitcode Writer and DXIL testing

[beanz]

Summary

This change is a big blob of code that isn't easy to break up. It
either comes in all together as a blob, works and has tests, or it
doesn't do anything.

Logically you can think of this patch as three things:
(1) Adding virtual interfaces so the bitcode writer can be overridden
(2) Adding a new bitcode writer implementation for DXIL
(3) Adding some (optional) crazy CMake goop to build the
DirectXShaderCompiler's llvm-dis as dxil-dis for testing

[nikic]

This implementation is not compatible with opaque pointers (https://llvm.org/docs/OpaquePointers.html) and will likely need a significant rework to support them.

[beanz]

In D122082#3394525, @nikic wrote:

This implementation is not compatible with opaque pointers (https://llvm.org/docs/OpaquePointers.html) and will likely need a significant rework to support them.

I don’t disagree, but…

Much of this code will work just fine with opaque pointers, and opaque pointers aren’t fully supported in clang (the only frontend we’re trying to support).

Since this code is going into an experimental backend, does opaque pointer support need to block this from landing, or can that be addressed subsequently?

[pete]

Overall LGTM. I don’t have a strong opinion on opaque pointers so nothing to add there.

llvm/test/tools/dxil-dis/FMA.ll

6 ↗

(On Diff #416726)

Looks like this is actually testing function attributes, not fma. Can you change the test name, or add the check line for an fma if that is the intention?

llvm/tools/dxil-dis/CMakeLists.txt

2

Is dxil-dis only a testing tool, or is it shipped in a distribution too?

If it’s only for tests then I think this is fine. If it’s shipped, then I wonder if the cmake variable should drop “TESTS”. Up to you though.

[efriedma]

See https://discourse.llvm.org/t/opaque-pointers-status-update/60296 for the current status of opaque pointers.

llvm/include/llvm/Bitcode/BitcodeWriter.h

37 ↗

(On Diff #416726)

Instead of returning a unique_ptr to an object that has exactly one function called on it, can you just pass in a callback function that constructs the writer and calls write()?

38 ↗

(On Diff #416726)

Please land the whitespace change here separately.

llvm/lib/Bitcode/Writer/BitcodeWriterPass.cpp

1 ↗

(On Diff #416726)

BitcodeWriterPass is a tiny amount of code... you might as well just fork it. (Normally, I'd be in favor of trying to share more code, but I'd prefer to minimize the changes to target-independent code for now.)

[nikic]

In D122082#3394675, @beanz wrote:

In D122082#3394525, @nikic wrote:

This implementation is not compatible with opaque pointers (https://llvm.org/docs/OpaquePointers.html) and will likely need a significant rework to support them.

I don’t disagree, but…

Much of this code will work just fine with opaque pointers, and opaque pointers aren’t fully supported in clang (the only frontend we’re trying to support).

Since this code is going into an experimental backend, does opaque pointer support need to block this from landing, or can that be addressed subsequently?

Yes, opaque pointer support is a hard requirement for any new code introduced in LLVM. Of course, it's okay if you implement it in a separate patch from this one, it just needs to be part of the patch stack and land at the same time.

[xbolva00]

Yes, opaque pointer support is a hard requirement for any new code introduced in LLVM.

Is this info documented somewhere with "roadmap" or "future plans"? Like telling developers / users "we are going to ditch undef, we are going to switch to opaque ptrs, etc.." . It would be nice addition to have such place with high level plans for future.

[beanz]

In D122082#3394836, @nikic wrote:

Yes, opaque pointer support is a hard requirement for any new code introduced in LLVM.

Not to be difficult, but was this discussed somewhere that I missed? I didn’t see it in any of the status updates or project plans for opaque pointers. I have no objection to doing the work (I’ve already started). I actually don’t even mind not merging this patch until I have it working. The opaque pointer effort will actually be a net win for the larger project I’m working on, so kudos to that :D.

I greatly dislike “you can’t land <x> because of <not done thing>” being a requirement _ever_ in LLVM, but especially in cases where the not completed work doesn’t have a final completion target (as the latest status update on opaque pointers says).

We held up a lot of great work behind the “new” pass manager. For _years_ nobody was allowed to add features to the “legacy” pass manager or even fix infrastructure issues in ways that relied on the pass manager.

Drawing hard lines in the sand like this is fine if there is wide community acceptance (as with any coding standard), but nobody should be able to draw these lines without discussion and agreement.

[tschuett]

In D122082#3395250, @beanz wrote:

In D122082#3394836, @nikic wrote:

Yes, opaque pointer support is a hard requirement for any new code introduced in LLVM.

LLVM tends to document a lot. Why is there no roadmap document? Get rid of old pass manager, move to globalisel, move to opaque pointers, ...

[xbolva00]

In D122082#3395572, @tschuett wrote:

In D122082#3395250, @beanz wrote:

In D122082#3394836, @nikic wrote:

Yes, opaque pointer support is a hard requirement for any new code introduced in LLVM.

LLVM tends to document a lot. Why is there no roadmap document? Get rid of old pass manager, move to globalisel, move to opaque pointers, ...

Can LLVM board create one?

@tstellar @tonic

[nikic]

@xbolva00 Don't think the board has any relation to this, but I do agree that having some central documentation would be great. While many of the ongoing migration efforts are well-documented by themselves, there's currently no central place that collects them. I've created https://reviews.llvm.org/D122122 as a starting point, feedback welcome.

[beanz]

In D122082#3395747, @nikic wrote:

@xbolva00 Don't think the board has any relation to this, but I do agree that having some central documentation would be great. While many of the ongoing migration efforts are well-documented by themselves, there's currently no central place that collects them. I've created https://reviews.llvm.org/D122122 as a starting point, feedback welcome.

@nikic, even that document doesn't say support for opaque pointers is a "hard requirement" for new contributions.

Given that there is no documentation or clear community consensus anywhere that I can find that it should be a hard requirement, and that this code is going into an experimental backend, which (as documented in the support policy: https://llvm.org/docs/SupportPolicy.html) means "you break it, I fix it". Would you revise your "needs revision" feedback?

[nikic]

In D122082#3395250, @beanz wrote:

In D122082#3394836, @nikic wrote:

Yes, opaque pointer support is a hard requirement for any new code introduced in LLVM.

Not to be difficult, but was this discussed somewhere that I missed? I didn’t see it in any of the status updates or project plans for opaque pointers. I have no objection to doing the work (I’ve already started). I actually don’t even mind not merging this patch until I have it working. The opaque pointer effort will actually be a net win for the larger project I’m working on, so kudos to that :D.

I greatly dislike “you can’t land <x> because of <not done thing>” being a requirement _ever_ in LLVM, but especially in cases where the not completed work doesn’t have a final completion target (as the latest status update on opaque pointers says).

We held up a lot of great work behind the “new” pass manager. For _years_ nobody was allowed to add features to the “legacy” pass manager or even fix infrastructure issues in ways that relied on the pass manager.

Drawing hard lines in the sand like this is fine if there is wide community acceptance (as with any coding standard), but nobody should be able to draw these lines without discussion and agreement.

I'm not sure what you want to hear here. The opaque pointers migration was decided many years ago, long before I started working on LLVM. Well-informed reviewers have been rejecting patches violating opaque pointers principles years before LLVM even had a proper opaque pointer type. Of course, enforcement on this has been inconsistent and somewhat dependent on which reviewer you hit.

Now that the opaque pointers migration is close to complete I'm keeping an active eye out to make sure there are no regressions. LLVM itself is essentially clean under opaque pointers, and we're maybe two weeks away from proposing a default switch in Clang. Which is why I'm so surprised that literally the first time I'm seeing push-back against an opaque pointer requirement in a review is now.

I think that for this change in particular it is important to resolve this issue in advance, because it may impact your overall design in a non-trivial way. Your current handling of DXIL is predicated on LLVM IR and DXIL being essentially the same, minus a few cosmetic differences like fneg vs fsub. With opaque pointers, LLVM IR and DXIL will differ in a more substantial way, and I'd naively expect this to have some impact on your implementation approach.

Finally, if you want to make a new pass manager analogy, I'm not telling you that you can't work on the legacy pass manager, I'm only telling you that you should implement functionality for both the legacy and new pass manager at the same time. That seems like a pretty reasonable request to me? In fact, our biggest failures in the NewPM migration were related to cases where this principle was not followed, and changes were made only to either the new or the legacy pass manager. Let's avoid that to the degree it is possible.

[tstellar]

In D122082#3397728, @beanz wrote:

In D122082#3395747, @nikic wrote:

@xbolva00 Don't think the board has any relation to this, but I do agree that having some central documentation would be great. While many of the ongoing migration efforts are well-documented by themselves, there's currently no central place that collects them. I've created https://reviews.llvm.org/D122122 as a starting point, feedback welcome.

@nikic, even that document doesn't say support for opaque pointers is a "hard requirement" for new contributions.

Given that there is no documentation or clear community consensus anywhere that I can find that it should be a hard requirement, and that this code is going into an experimental backend, which (as documented in the support policy: https://llvm.org/docs/SupportPolicy.html) means "you break it, I fix it". Would you revise your "needs revision" feedback?

It would be great if there were more documentation around the migration timelines, but lack of documentation does not mean lack of plans or lack of consensus. I would recommend starting a "Does the DXIL backend need to be compatible with Opaque Pointers" discussion on Discourse just so we can open this question up to a wider audience and get a little more clarity on this issue.

[beanz]

In D122082#3403025, @tstellar wrote:

It would be great if there were more documentation around the migration timelines, but lack of documentation does not mean lack of plans or lack of consensus. I would recommend starting a "Does the DXIL backend need to be compatible with Opaque Pointers" discussion on Discourse just so we can open this question up to a wider audience and get a little more clarity on this issue.

I've posted patches already to get the DIXL emitter working with opaque pointers. It still relies on typed pointers being a thing (not in the IR, but constructable llvm::Types). I'm also working on removing that reliance, and expect to have patches up this week.

[rnk]

I've posted patches already to get the DIXL emitter working with opaque pointers. It still relies on typed pointers being a thing (not in the IR, but constructable llvm::Types). I'm also working on removing that reliance, and expect to have patches up this week.

Glad to hear it. That's a key thing I wanted to bring up. Right now, an opaque pointer is basically just a pointer with a null element type, but I think the long term vision is that we remove the field altogether. I think we're a long way from here to there, but we should plan on that as the end state.

[MaskRay]

llvm/lib/Target/DirectX/DXILBitcodeWriter.cpp

106 ↗

(On Diff #416726)

https://llvm.org/docs/CodingStandards.html#doxygen-use-in-documentation-comments
Don’t duplicate function or class name at the beginning of the comment.

166 ↗

(On Diff #416726)

range-based for

1689 ↗

(On Diff #416726)

The FIXME (copied from BitcodeWriter.cpp) is probably unrelated and should be deleted.

BitcodeWriter.cpp seems to (ab)use 64 inline elements a lot. Does DXIL need 64?
You may consider SmallVector<unsigned>.

1700 ↗

(On Diff #416726)

ranged-based for

1702 ↗

(On Diff #416726)

1705 ↗

(On Diff #416726)

1711 ↗

(On Diff #416726)

Just remove #if 0. If you need comments, use //

1878 ↗

(On Diff #416726)

These llvm_unreachable calls should be assert

2092 ↗

(On Diff #416726)

range-based for

llvm/tools/dxil-dis/CMakeLists.txt

9

LLVM_TARGETS_TO_BUILD is a superset. I think with this change this line wont' need a change when DirectX becomes non-experimental.

40

Any reason cmake/modules/AddLLVM.cmake add_llvm_tool_symlink can't be used?

[beanz]

Updating based on review feedback.

This change forks the required infrastrcuture from the LLVM Bitcode writer into the DirectX backend instead of relying on refactoring of the existing pieces to be reusable.

[beanz]

The last update makes this change a lot bigger because it is forking a bunch of code out of the bit writer library as per review feedback.

[beanz]

@nikic, are you satisfied with the additional patches in the stack which support opaque pointers?

[nikic]

In D122082#3429630, @beanz wrote:

@nikic, are you satisfied with the additional patches in the stack which support opaque pointers?

I've only reviewed the first patch in the stack (D122268) yet, but have two high level questions:

1. How does/will this handle constant expressions?
2. How is dxil-dis supposed to work with opaque pointers?

[beanz]

In D122082#3433065, @nikic wrote:

In D122082#3429630, @beanz wrote:

@nikic, are you satisfied with the additional patches in the stack which support opaque pointers?

I've only reviewed the first patch in the stack (D122268) yet, but have two high level questions:

1. How does/will this handle constant expressions?

There are some gaps in the type analysis (constant expressions being one), but the method used here should apply. By mapping Values to typed pointer types, I can emit the typed pointer type in place of the opaque pointer type in the bitstream.

2. How is dxil-dis supposed to work with opaque pointers?

dxil-dis is LLVM 3.7's llvm-dis. It _can't_ handle opaque pointers, and doesn't need to. The references to the opaque pointer type are replaced with typed pointer types that are encoded matching LLVM 3.7's PointerType.

[nikic]

In D122082#3433503, @beanz wrote:

There are some gaps in the type analysis (constant expressions being one), but the method used here should apply. By mapping Values to typed pointer types, I can emit the typed pointer type in place of the opaque pointer type in the bitstream.

The bit I'm specifically concerned about is that you can't insert bitcasts in the prepare pass, as those would always get folded away. But I guess you can still materialize them directly during bitcode writing.

In D122082#3433503, @beanz wrote:

2. How is dxil-dis supposed to work with opaque pointers?

dxil-dis is LLVM 3.7's llvm-dis. It _can't_ handle opaque pointers, and doesn't need to. The references to the opaque pointer type are replaced with typed pointer types that are encoded matching LLVM 3.7's PointerType.

Oh, sorry, that was confusion on my part. Somehow I thought that dxil-dis is just a rebranded llvm-dis, I didn't realize that it's built from an external repo.

[beanz]

In D122082#3433583, @nikic wrote:

The bit I'm specifically concerned about is that you can't insert bitcasts in the prepare pass, as those would always get folded away. But I guess you can still materialize them directly during bitcode writing.

We only need to insert those because of how instructions are numbered in the bitcode writer. I don't _think_ we will need them in the ConstantExprs, but I may not be fully thinking that through. If we needed to materialize them in the bitwriter we can do that (especially now that we're forking the whole thing).

[beanz]

llvm/tools/dxil-dis/CMakeLists.txt

40

Sorry @MaskRay, I should have acknowledged this comment earlier. add_llvm_tool_symlink uses generator expressions to read values from the underlying target that doesn't work with this terrible external project goop.

[nikic]

Removing me as a blocker here, I think the opaque pointers concern is sufficiently addressed by dependent patches. I haven't looked carefully over the code here, on the assumption that this is just boring "copy of bitcode writing with necessary tweaks for DXIL".

llvm/test/tools/dxil-dis/FMA.ll

6 ↗

(On Diff #416726)

Comment not resolved.

[thakis]

llvm/tools/dxil-dis/CMakeLists.txt

28

Adding a dep to another github repo is pretty unusual, even though it's off by default. Was this part explicitly discussed somewhere?

[beanz]

llvm/tools/dxil-dis/CMakeLists.txt

28

Not explicitly. The dependency is optional and only for testing. It builds a copy of the modified llvm-3.7 llvm-dis tool to verify that the newly emitted Bitcode can be read.

Each part of this is configurable. It can clone and build, or build from a local directory, or just take a prebuilt tool.

llvm/lib/Target/DirectX/CMakeLists.txt

 add_llvm_component_group(DirectX)
 
 set(LLVM_TARGET_DEFINITIONS DirectX.td)
 
 tablegen(LLVM DirectXGenSubtargetInfo.inc -gen-subtarget)
 
 add_public_tablegen_target(DirectXCommonTableGen)
 
 add_llvm_target(DirectXCodeGen
   DirectXSubtarget.cpp
   DirectXTargetMachine.cpp
   DXILPrepare.cpp
 
   LINK_COMPONENTS
-  Bitwriter
   Core
   Support
   DirectXInfo
+  DXILBitWriter
 
   ADD_TO_COMPONENT
   DirectX
 )
 
 add_subdirectory(MCTargetDesc)
 add_subdirectory(TargetInfo)
+add_subdirectory(DXILWriter)

llvm/lib/Target/DirectX/DXILWriter/CMakeLists.txt

+add_llvm_component_library(LLVMDXILBitWriter
+  DXILBitcodeWriter.cpp
+  DXILValueEnumerator.cpp
+  DXILWriterPass.cpp
+
+  DEPENDS
+  intrinsics_gen
+
+  LINK_COMPONENTS
+  Bitwriter
+  Core
+  MC
+  Object
+  Support
+  )

llvm/lib/Target/DirectX/DXILWriter/DXILBitcodeWriter.h

+//===- Bitcode/Writer/DXILBitcodeWriter.cpp - DXIL Bitcode Writer ---------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// Bitcode writer implementation.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/ADT/StringRef.h"
+#include "llvm/IR/ModuleSummaryIndex.h"
+#include "llvm/MC/StringTableBuilder.h"
+#include "llvm/Support/Allocator.h"
+#include "llvm/Support/MemoryBufferRef.h"
+#include <map>
+#include <memory>
+#include <string>
+#include <vector>
+
+namespace llvm {
+
+class BitstreamWriter;
+class Module;
+class raw_ostream;
+
+namespace dxil {
+
+class BitcodeWriter {
+  SmallVectorImpl<char> &Buffer;
+  std::unique_ptr<BitstreamWriter> Stream;
+
+  StringTableBuilder StrtabBuilder{StringTableBuilder::RAW};
+
+  // Owns any strings created by the irsymtab writer until we create the
+  // string table.
+  BumpPtrAllocator Alloc;
+
+  bool WroteStrtab = false, WroteSymtab = false;
+
+  void writeBlob(unsigned Block, unsigned Record, StringRef Blob);
+
+  std::vector<Module *> Mods;
+
+public:
+  /// Create a BitcodeWriter that writes to Buffer.
+  BitcodeWriter(SmallVectorImpl<char> &Buffer, raw_fd_stream *FS = nullptr);
+
+  ~BitcodeWriter();
+
+  /// Attempt to write a symbol table to the bitcode file. This must be called
+  /// at most once after all modules have been written.
+  ///
+  /// A reader does not require a symbol table to interpret a bitcode file;
+  /// the symbol table is needed only to improve link-time performance. So
+  /// this function may decide not to write a symbol table. It may so decide
+  /// if, for example, the target is unregistered or the IR is malformed.
+  void writeSymtab();
+
+  /// Write the bitcode file's string table. This must be called exactly once
+  /// after all modules and the optional symbol table have been written.
+  void writeStrtab();
+
+  /// Copy the string table for another module into this bitcode file. This
+  /// should be called after copying the module itself into the bitcode file.
+  void copyStrtab(StringRef Strtab);
+
+  /// Write the specified module to the buffer specified at construction time.
+  void writeModule(const Module &M);
+};
+
+/// Write the specified module to the specified raw output stream.
+///
+/// For streams where it matters, the given stream should be in "binary"
+/// mode.
+void WriteDXILToFile(const Module &M, raw_ostream &Out);
+
+} // namespace dxil
+
+} // namespace llvm

llvm/lib/Target/DirectX/DXILWriter/DXILBitcodeWriter.cpp

+//===- Bitcode/Writer/DXILBitcodeWriter.cpp - DXIL Bitcode Writer ---------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// Bitcode writer implementation.
+//
+//===----------------------------------------------------------------------===//
+
+#include "DXILBitcodeWriter.h"
+#include "DXILValueEnumerator.h"
+#include "llvm/ADT/Triple.h"
+#include "llvm/Bitcode/BitcodeCommon.h"
+#include "llvm/Bitcode/BitcodeReader.h"
+#include "llvm/Bitcode/LLVMBitCodes.h"
+#include "llvm/Bitstream/BitCodes.h"
+#include "llvm/Bitstream/BitstreamWriter.h"
+#include "llvm/IR/Attributes.h"
+#include "llvm/IR/BasicBlock.h"
+#include "llvm/IR/Comdat.h"
+#include "llvm/IR/Constant.h"
+#include "llvm/IR/Constants.h"
+#include "llvm/IR/DebugInfoMetadata.h"
+#include "llvm/IR/DebugLoc.h"
+#include "llvm/IR/DerivedTypes.h"
+#include "llvm/IR/Function.h"
+#include "llvm/IR/GlobalAlias.h"
+#include "llvm/IR/GlobalIFunc.h"
+#include "llvm/IR/GlobalObject.h"
+#include "llvm/IR/GlobalValue.h"
+#include "llvm/IR/GlobalVariable.h"
+#include "llvm/IR/InlineAsm.h"
+#include "llvm/IR/InstrTypes.h"
+#include "llvm/IR/Instruction.h"
+#include "llvm/IR/Instructions.h"
+#include "llvm/IR/LLVMContext.h"
+#include "llvm/IR/Metadata.h"
+#include "llvm/IR/Module.h"
+#include "llvm/IR/ModuleSummaryIndex.h"
+#include "llvm/IR/Operator.h"
+#include "llvm/IR/Type.h"
+#include "llvm/IR/UseListOrder.h"
+#include "llvm/IR/Value.h"
+#include "llvm/IR/ValueSymbolTable.h"
+#include "llvm/Object/IRSymtab.h"
+#include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/SHA1.h"
+
+namespace llvm {
+namespace dxil {
+
+// Generates an enum to use as an index in the Abbrev array of Metadata record.
+enum MetadataAbbrev : unsigned {
+#define HANDLE_MDNODE_LEAF(CLASS) CLASS##AbbrevID,
+#include "llvm/IR/Metadata.def"
+  LastPlusOne
+};
+
+class DXILBitcodeWriter {
+
+  /// These are manifest constants used by the bitcode writer. They do not need
+  /// to be kept in sync with the reader, but need to be consistent within this
+  /// file.
+  enum {
+    // VALUE_SYMTAB_BLOCK abbrev id's.
+    VST_ENTRY_8_ABBREV = bitc::FIRST_APPLICATION_ABBREV,
+    VST_ENTRY_7_ABBREV,
+    VST_ENTRY_6_ABBREV,
+    VST_BBENTRY_6_ABBREV,
+
+    // CONSTANTS_BLOCK abbrev id's.
+    CONSTANTS_SETTYPE_ABBREV = bitc::FIRST_APPLICATION_ABBREV,
+    CONSTANTS_INTEGER_ABBREV,
+    CONSTANTS_CE_CAST_Abbrev,
+    CONSTANTS_NULL_Abbrev,
+
+    // FUNCTION_BLOCK abbrev id's.
+    FUNCTION_INST_LOAD_ABBREV = bitc::FIRST_APPLICATION_ABBREV,
+    FUNCTION_INST_BINOP_ABBREV,
+    FUNCTION_INST_BINOP_FLAGS_ABBREV,
+    FUNCTION_INST_CAST_ABBREV,
+    FUNCTION_INST_RET_VOID_ABBREV,
+    FUNCTION_INST_RET_VAL_ABBREV,
+    FUNCTION_INST_UNREACHABLE_ABBREV,
+    FUNCTION_INST_GEP_ABBREV,
+  };
+
+  /// The stream created and owned by the client.
+  BitstreamWriter &Stream;
+
+  StringTableBuilder &StrtabBuilder;
+
+  /// The Module to write to bitcode.
+  const Module &M;
+
+  /// Enumerates ids for all values in the module.
+  ValueEnumerator VE;
+
+  /// Map that holds the correspondence between GUIDs in the summary index,
+  /// that came from indirect call profiles, and a value id generated by this
+  /// class to use in the VST and summary block records.
+  std::map<GlobalValue::GUID, unsigned> GUIDToValueIdMap;
+
+  /// Tracks the last value id recorded in the GUIDToValueMap.
+  unsigned GlobalValueId;
+
+  /// Saves the offset of the VSTOffset record that must eventually be
+  /// backpatched with the offset of the actual VST.
+  uint64_t VSTOffsetPlaceholder = 0;
+
+  /// Pointer to the buffer allocated by caller for bitcode writing.
+  const SmallVectorImpl<char> &Buffer;
+
+  /// The start bit of the identification block.
+  uint64_t BitcodeStartBit;
+
+public:
+  /// Constructs a ModuleBitcodeWriter object for the given Module,
+  /// writing to the provided \p Buffer.
+  DXILBitcodeWriter(const Module &M, SmallVectorImpl<char> &Buffer,
+                    StringTableBuilder &StrtabBuilder, BitstreamWriter &Stream)
+      : Stream(Stream), StrtabBuilder(StrtabBuilder), M(M),
+        VE(M, true), Buffer(Buffer),
+        BitcodeStartBit(Stream.GetCurrentBitNo()) {
+    GlobalValueId = VE.getValues().size();
+  }
+
+  /// Emit the current module to the bitstream.
+  void write();
+
+  static uint64_t getAttrKindEncoding(Attribute::AttrKind Kind);
+  static void writeStringRecord(BitstreamWriter &Stream, unsigned Code,
+                                StringRef Str, unsigned AbbrevToUse);
+  static void writeIdentificationBlock(BitstreamWriter &Stream);
+  static void emitSignedInt64(SmallVectorImpl<uint64_t> &Vals, uint64_t V);
+  static void emitWideAPInt(SmallVectorImpl<uint64_t> &Vals, const APInt &A);
+
+  static unsigned getEncodedComdatSelectionKind(const Comdat &C);
+  static unsigned getEncodedLinkage(const GlobalValue::LinkageTypes Linkage);
+  static unsigned getEncodedLinkage(const GlobalValue &GV);
+  static unsigned getEncodedVisibility(const GlobalValue &GV);
+  static unsigned getEncodedThreadLocalMode(const GlobalValue &GV);
+  static unsigned getEncodedDLLStorageClass(const GlobalValue &GV);
+  static unsigned getEncodedCastOpcode(unsigned Opcode);
+  static unsigned getEncodedUnaryOpcode(unsigned Opcode);
+  static unsigned getEncodedBinaryOpcode(unsigned Opcode);
+  static unsigned getEncodedRMWOperation(AtomicRMWInst::BinOp Op);
+  static unsigned getEncodedOrdering(AtomicOrdering Ordering);
+  static uint64_t getOptimizationFlags(const Value *V);
+
+private:
+  void writeModuleVersion();
+  void writePerModuleGlobalValueSummary();
+
+  void writePerModuleFunctionSummaryRecord(SmallVector<uint64_t, 64> &NameVals,
+                                           GlobalValueSummary *Summary,
+                                           unsigned ValueID,
+                                           unsigned FSCallsAbbrev,
+                                           unsigned FSCallsProfileAbbrev,
+                                           const Function &F);
+  void writeModuleLevelReferences(const GlobalVariable &V,
+                                  SmallVector<uint64_t, 64> &NameVals,
+                                  unsigned FSModRefsAbbrev,
+                                  unsigned FSModVTableRefsAbbrev);
+
+  void assignValueId(GlobalValue::GUID ValGUID) {
+    GUIDToValueIdMap[ValGUID] = ++GlobalValueId;
+  }
+
+  unsigned getValueId(GlobalValue::GUID ValGUID) {
+    const auto &VMI = GUIDToValueIdMap.find(ValGUID);
+    // Expect that any GUID value had a value Id assigned by an
+    // earlier call to assignValueId.
+    assert(VMI != GUIDToValueIdMap.end() &&
+           "GUID does not have assigned value Id");
+    return VMI->second;
+  }
+
+  // Helper to get the valueId for the type of value recorded in VI.
+  unsigned getValueId(ValueInfo VI) {
+    if (!VI.haveGVs() || !VI.getValue())
+      return getValueId(VI.getGUID());
+    return VE.getValueID(VI.getValue());
+  }
+
+  std::map<GlobalValue::GUID, unsigned> &valueIds() { return GUIDToValueIdMap; }
+
+  uint64_t bitcodeStartBit() { return BitcodeStartBit; }
+
+  size_t addToStrtab(StringRef Str);
+
+  unsigned createDILocationAbbrev();
+  unsigned createGenericDINodeAbbrev();
+
+  void writeAttributeGroupTable();
+  void writeAttributeTable();
+  void writeTypeTable();
+  void writeComdats();
+  void writeValueSymbolTableForwardDecl();
+  void writeModuleInfo();
+  void writeValueAsMetadata(const ValueAsMetadata *MD,
+                            SmallVectorImpl<uint64_t> &Record);
+  void writeMDTuple(const MDTuple *N, SmallVectorImpl<uint64_t> &Record,
+                    unsigned Abbrev);
+  void writeDILocation(const DILocation *N, SmallVectorImpl<uint64_t> &Record,
+                       unsigned &Abbrev);
+  void writeGenericDINode(const GenericDINode *N,
+                          SmallVectorImpl<uint64_t> &Record, unsigned &Abbrev) {
+    llvm_unreachable("DXIL cannot contain GenericDI Nodes");
+  }
+  void writeDISubrange(const DISubrange *N, SmallVectorImpl<uint64_t> &Record,
+                       unsigned Abbrev);
+  void writeDIGenericSubrange(const DIGenericSubrange *N,
+                              SmallVectorImpl<uint64_t> &Record,
+                              unsigned Abbrev) {
+    llvm_unreachable("DXIL cannot contain DIGenericSubrange Nodes");
+  }
+  void writeDIEnumerator(const DIEnumerator *N,
+                         SmallVectorImpl<uint64_t> &Record, unsigned Abbrev);
+  void writeDIBasicType(const DIBasicType *N, SmallVectorImpl<uint64_t> &Record,
+                        unsigned Abbrev);
+  void writeDIStringType(const DIStringType *N,
+                         SmallVectorImpl<uint64_t> &Record, unsigned Abbrev) {
+    llvm_unreachable("DXIL cannot contain DIStringType Nodes");
+  }
+  void writeDIDerivedType(const DIDerivedType *N,
+                          SmallVectorImpl<uint64_t> &Record, unsigned Abbrev);
+  void writeDICompositeType(const DICompositeType *N,
+                            SmallVectorImpl<uint64_t> &Record, unsigned Abbrev);
+  void writeDISubroutineType(const DISubroutineType *N,
+                             SmallVectorImpl<uint64_t> &Record,
+                             unsigned Abbrev);
+  void writeDIFile(const DIFile *N, SmallVectorImpl<uint64_t> &Record,
+                   unsigned Abbrev);
+  void writeDICompileUnit(const DICompileUnit *N,
+                          SmallVectorImpl<uint64_t> &Record, unsigned Abbrev);
+  void writeDISubprogram(const DISubprogram *N,
+                         SmallVectorImpl<uint64_t> &Record, unsigned Abbrev);
+  void writeDILexicalBlock(const DILexicalBlock *N,
+                           SmallVectorImpl<uint64_t> &Record, unsigned Abbrev);
+  void writeDILexicalBlockFile(const DILexicalBlockFile *N,
+                               SmallVectorImpl<uint64_t> &Record,
+                               unsigned Abbrev);
+  void writeDICommonBlock(const DICommonBlock *N,
+                          SmallVectorImpl<uint64_t> &Record, unsigned Abbrev) {
+    llvm_unreachable("DXIL cannot contain DICommonBlock Nodes");
+  }
+  void writeDINamespace(const DINamespace *N, SmallVectorImpl<uint64_t> &Record,
+                        unsigned Abbrev);
+  void writeDIMacro(const DIMacro *N, SmallVectorImpl<uint64_t> &Record,
+                    unsigned Abbrev) {
+    llvm_unreachable("DXIL cannot contain DIMacro Nodes");
+  }
+  void writeDIMacroFile(const DIMacroFile *N, SmallVectorImpl<uint64_t> &Record,
+                        unsigned Abbrev) {
+    llvm_unreachable("DXIL cannot contain DIMacroFile Nodes");
+  }
+  void writeDIArgList(const DIArgList *N, SmallVectorImpl<uint64_t> &Record,
+                      unsigned Abbrev) {
+    llvm_unreachable("DXIL cannot contain DIArgList Nodes");
+  }
+  void writeDIModule(const DIModule *N, SmallVectorImpl<uint64_t> &Record,
+                     unsigned Abbrev);
+  void writeDITemplateTypeParameter(const DITemplateTypeParameter *N,
+                                    SmallVectorImpl<uint64_t> &Record,
+                                    unsigned Abbrev);
+  void writeDITemplateValueParameter(const DITemplateValueParameter *N,
+                                     SmallVectorImpl<uint64_t> &Record,
+                                     unsigned Abbrev);
+  void writeDIGlobalVariable(const DIGlobalVariable *N,
+                             SmallVectorImpl<uint64_t> &Record,
+                             unsigned Abbrev);
+  void writeDILocalVariable(const DILocalVariable *N,
+                            SmallVectorImpl<uint64_t> &Record, unsigned Abbrev);
+  void writeDILabel(const DILabel *N, SmallVectorImpl<uint64_t> &Record,
+                    unsigned Abbrev) {
+    llvm_unreachable("DXIL cannot contain DILabel Nodes");
+  }
+  void writeDIExpression(const DIExpression *N,
+                         SmallVectorImpl<uint64_t> &Record, unsigned Abbrev);
+  void writeDIGlobalVariableExpression(const DIGlobalVariableExpression *N,
+                                       SmallVectorImpl<uint64_t> &Record,
+                                       unsigned Abbrev) {
+    llvm_unreachable("DXIL cannot contain GlobalVariableExpression Nodes");
+  }
+  void writeDIObjCProperty(const DIObjCProperty *N,
+                           SmallVectorImpl<uint64_t> &Record, unsigned Abbrev);
+  void writeDIImportedEntity(const DIImportedEntity *N,
+                             SmallVectorImpl<uint64_t> &Record,
+                             unsigned Abbrev);
+  unsigned createNamedMetadataAbbrev();
+  void writeNamedMetadata(SmallVectorImpl<uint64_t> &Record);
+  unsigned createMetadataStringsAbbrev();
+  void writeMetadataStrings(ArrayRef<const Metadata *> Strings,
+                            SmallVectorImpl<uint64_t> &Record);
+  void writeMetadataRecords(ArrayRef<const Metadata *> MDs,
+                            SmallVectorImpl<uint64_t> &Record,
+                            std::vector<unsigned> *MDAbbrevs = nullptr,
+                            std::vector<uint64_t> *IndexPos = nullptr);
+  void writeModuleMetadata();
+  void writeFunctionMetadata(const Function &F);
+  void writeFunctionMetadataAttachment(const Function &F);
+  void pushGlobalMetadataAttachment(SmallVectorImpl<uint64_t> &Record,
+                                    const GlobalObject &GO);
+  void writeModuleMetadataKinds();
+  void writeOperandBundleTags();
+  void writeSyncScopeNames();
+  void writeConstants(unsigned FirstVal, unsigned LastVal, bool isGlobal);
+  void writeModuleConstants();
+  bool pushValueAndType(const Value *V, unsigned InstID,
+                        SmallVectorImpl<unsigned> &Vals);
+  void writeOperandBundles(const CallBase &CB, unsigned InstID);
+  void pushValue(const Value *V, unsigned InstID,
+                 SmallVectorImpl<unsigned> &Vals);
+  void pushValueSigned(const Value *V, unsigned InstID,
+                       SmallVectorImpl<uint64_t> &Vals);
+  void writeInstruction(const Instruction &I, unsigned InstID,
+                        SmallVectorImpl<unsigned> &Vals);
+  void writeFunctionLevelValueSymbolTable(const ValueSymbolTable &VST);
+  void writeGlobalValueSymbolTable(
+      DenseMap<const Function *, uint64_t> &FunctionToBitcodeIndex);
+  void writeUseList(UseListOrder &&Order);
+  void writeUseListBlock(const Function *F);
+  void writeFunction(const Function &F);
+  void writeBlockInfo();
+
+  unsigned getEncodedSyncScopeID(SyncScope::ID SSID) { return unsigned(SSID); }
+
+  unsigned getEncodedAlign(MaybeAlign Alignment) { return encode(Alignment); }
+};
+
+} // namespace dxil
+} // namespace llvm
+
+using namespace llvm;
+using namespace llvm::dxil;
+
+////////////////////////////////////////////////////////////////////////////////
+/// Begin dxil::BitcodeWriter Implementation
+////////////////////////////////////////////////////////////////////////////////
+
+dxil::BitcodeWriter::BitcodeWriter(SmallVectorImpl<char> &Buffer, raw_fd_stream *FS)
+    : Buffer(Buffer), Stream(new BitstreamWriter(Buffer, FS, 512)) {
+  // Emit the file header.
+  Stream->Emit((unsigned)'B', 8);
+  Stream->Emit((unsigned)'C', 8);
+  Stream->Emit(0x0, 4);
+  Stream->Emit(0xC, 4);
+  Stream->Emit(0xE, 4);
+  Stream->Emit(0xD, 4);
+}
+
+dxil::BitcodeWriter::~BitcodeWriter() { assert(WroteStrtab); }
+
+/// Write the specified module to the specified output stream.
+void dxil::WriteDXILToFile(const Module &M, raw_ostream &Out) {
+  SmallVector<char, 0> Buffer;
+  Buffer.reserve(256 * 1024);
+
+  // If this is darwin or another generic macho target, reserve space for the
+  // header.
+  Triple TT(M.getTargetTriple());
+  if (TT.isOSDarwin() || TT.isOSBinFormatMachO())
+    Buffer.insert(Buffer.begin(), BWH_HeaderSize, 0);
+
+  BitcodeWriter Writer(Buffer, dyn_cast<raw_fd_stream>(&Out));
+  Writer.writeModule(M);
+  Writer.writeSymtab();
+  Writer.writeStrtab();
+
+  // Write the generated bitstream to "Out".
+  if (!Buffer.empty())
+    Out.write((char *)&Buffer.front(), Buffer.size());
+}
+
+void BitcodeWriter::writeBlob(unsigned Block, unsigned Record, StringRef Blob) {
+  Stream->EnterSubblock(Block, 3);
+
+  auto Abbv = std::make_shared<BitCodeAbbrev>();
+  Abbv->Add(BitCodeAbbrevOp(Record));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Blob));
+  auto AbbrevNo = Stream->EmitAbbrev(std::move(Abbv));
+
+  Stream->EmitRecordWithBlob(AbbrevNo, ArrayRef<uint64_t>{Record}, Blob);
+
+  Stream->ExitBlock();
+}
+
+void BitcodeWriter::writeSymtab() {
+  assert(!WroteStrtab && !WroteSymtab);
+
+  // If any module has module-level inline asm, we will require a registered asm
+  // parser for the target so that we can create an accurate symbol table for
+  // the module.
+  for (Module *M : Mods) {
+    if (M->getModuleInlineAsm().empty())
+      continue;
+  }
+
+  WroteSymtab = true;
+  SmallVector<char, 0> Symtab;
+  // The irsymtab::build function may be unable to create a symbol table if the
+  // module is malformed (e.g. it contains an invalid alias). Writing a symbol
+  // table is not required for correctness, but we still want to be able to
+  // write malformed modules to bitcode files, so swallow the error.
+  if (Error E = irsymtab::build(Mods, Symtab, StrtabBuilder, Alloc)) {
+    consumeError(std::move(E));
+    return;
+  }
+
+  writeBlob(bitc::SYMTAB_BLOCK_ID, bitc::SYMTAB_BLOB,
+            {Symtab.data(), Symtab.size()});
+}
+
+void BitcodeWriter::writeStrtab() {
+  assert(!WroteStrtab);
+
+  std::vector<char> Strtab;
+  StrtabBuilder.finalizeInOrder();
+  Strtab.resize(StrtabBuilder.getSize());
+  StrtabBuilder.write((uint8_t *)Strtab.data());
+
+  writeBlob(bitc::STRTAB_BLOCK_ID, bitc::STRTAB_BLOB,
+            {Strtab.data(), Strtab.size()});
+
+  WroteStrtab = true;
+}
+
+void BitcodeWriter::copyStrtab(StringRef Strtab) {
+  writeBlob(bitc::STRTAB_BLOCK_ID, bitc::STRTAB_BLOB, Strtab);
+  WroteStrtab = true;
+}
+
+void BitcodeWriter::writeModule(const Module &M) {
+  assert(!WroteStrtab);
+
+  // The Mods vector is used by irsymtab::build, which requires non-const
+  // Modules in case it needs to materialize metadata. But the bitcode writer
+  // requires that the module is materialized, so we can cast to non-const here,
+  // after checking that it is in fact materialized.
+  assert(M.isMaterialized());
+  Mods.push_back(const_cast<Module *>(&M));
+
+  DXILBitcodeWriter ModuleWriter(M, Buffer, StrtabBuilder, *Stream);
+  ModuleWriter.write();
+}
+
+////////////////////////////////////////////////////////////////////////////////
+/// Begin dxil::BitcodeWriterBase Implementation
+////////////////////////////////////////////////////////////////////////////////
+
+unsigned DXILBitcodeWriter::getEncodedCastOpcode(unsigned Opcode) {
+  switch (Opcode) {
+  default:
+    llvm_unreachable("Unknown cast instruction!");
+  case Instruction::Trunc:
+    return bitc::CAST_TRUNC;
+  case Instruction::ZExt:
+    return bitc::CAST_ZEXT;
+  case Instruction::SExt:
+    return bitc::CAST_SEXT;
+  case Instruction::FPToUI:
+    return bitc::CAST_FPTOUI;
+  case Instruction::FPToSI:
+    return bitc::CAST_FPTOSI;
+  case Instruction::UIToFP:
+    return bitc::CAST_UITOFP;
+  case Instruction::SIToFP:
+    return bitc::CAST_SITOFP;
+  case Instruction::FPTrunc:
+    return bitc::CAST_FPTRUNC;
+  case Instruction::FPExt:
+    return bitc::CAST_FPEXT;
+  case Instruction::PtrToInt:
+    return bitc::CAST_PTRTOINT;
+  case Instruction::IntToPtr:
+    return bitc::CAST_INTTOPTR;
+  case Instruction::BitCast:
+    return bitc::CAST_BITCAST;
+  case Instruction::AddrSpaceCast:
+    return bitc::CAST_ADDRSPACECAST;
+  }
+}
+
+unsigned DXILBitcodeWriter::getEncodedUnaryOpcode(unsigned Opcode) {
+  switch (Opcode) {
+  default:
+    llvm_unreachable("Unknown binary instruction!");
+  case Instruction::FNeg:
+    return bitc::UNOP_FNEG;
+  }
+}
+
+unsigned DXILBitcodeWriter::getEncodedBinaryOpcode(unsigned Opcode) {
+  switch (Opcode) {
+  default:
+    llvm_unreachable("Unknown binary instruction!");
+  case Instruction::Add:
+  case Instruction::FAdd:
+    return bitc::BINOP_ADD;
+  case Instruction::Sub:
+  case Instruction::FSub:
+    return bitc::BINOP_SUB;
+  case Instruction::Mul:
+  case Instruction::FMul:
+    return bitc::BINOP_MUL;
+  case Instruction::UDiv:
+    return bitc::BINOP_UDIV;
+  case Instruction::FDiv:
+  case Instruction::SDiv:
+    return bitc::BINOP_SDIV;
+  case Instruction::URem:
+    return bitc::BINOP_UREM;
+  case Instruction::FRem:
+  case Instruction::SRem:
+    return bitc::BINOP_SREM;
+  case Instruction::Shl:
+    return bitc::BINOP_SHL;
+  case Instruction::LShr:
+    return bitc::BINOP_LSHR;
+  case Instruction::AShr:
+    return bitc::BINOP_ASHR;
+  case Instruction::And:
+    return bitc::BINOP_AND;
+  case Instruction::Or:
+    return bitc::BINOP_OR;
+  case Instruction::Xor:
+    return bitc::BINOP_XOR;
+  }
+}
+
+unsigned DXILBitcodeWriter::getEncodedRMWOperation(AtomicRMWInst::BinOp Op) {
+  switch (Op) {
+  default:
+    llvm_unreachable("Unknown RMW operation!");
+  case AtomicRMWInst::Xchg:
+    return bitc::RMW_XCHG;
+  case AtomicRMWInst::Add:
+    return bitc::RMW_ADD;
+  case AtomicRMWInst::Sub:
+    return bitc::RMW_SUB;
+  case AtomicRMWInst::And:
+    return bitc::RMW_AND;
+  case AtomicRMWInst::Nand:
+    return bitc::RMW_NAND;
+  case AtomicRMWInst::Or:
+    return bitc::RMW_OR;
+  case AtomicRMWInst::Xor:
+    return bitc::RMW_XOR;
+  case AtomicRMWInst::Max:
+    return bitc::RMW_MAX;
+  case AtomicRMWInst::Min:
+    return bitc::RMW_MIN;
+  case AtomicRMWInst::UMax:
+    return bitc::RMW_UMAX;
+  case AtomicRMWInst::UMin:
+    return bitc::RMW_UMIN;
+  case AtomicRMWInst::FAdd:
+    return bitc::RMW_FADD;
+  case AtomicRMWInst::FSub:
+    return bitc::RMW_FSUB;
+  }
+}
+
+unsigned DXILBitcodeWriter::getEncodedOrdering(AtomicOrdering Ordering) {
+  switch (Ordering) {
+  case AtomicOrdering::NotAtomic:
+    return bitc::ORDERING_NOTATOMIC;
+  case AtomicOrdering::Unordered:
+    return bitc::ORDERING_UNORDERED;
+  case AtomicOrdering::Monotonic:
+    return bitc::ORDERING_MONOTONIC;
+  case AtomicOrdering::Acquire:
+    return bitc::ORDERING_ACQUIRE;
+  case AtomicOrdering::Release:
+    return bitc::ORDERING_RELEASE;
+  case AtomicOrdering::AcquireRelease:
+    return bitc::ORDERING_ACQREL;
+  case AtomicOrdering::SequentiallyConsistent:
+    return bitc::ORDERING_SEQCST;
+  }
+  llvm_unreachable("Invalid ordering");
+}
+
+void DXILBitcodeWriter::writeStringRecord(BitstreamWriter &Stream,
+                                          unsigned Code, StringRef Str,
+                                          unsigned AbbrevToUse) {
+  SmallVector<unsigned, 64> Vals;
+
+  // Code: [strchar x N]
+  for (char C : Str) {
+    if (AbbrevToUse && !BitCodeAbbrevOp::isChar6(C))
+      AbbrevToUse = 0;
+    Vals.push_back(C);
+  }
+
+  // Emit the finished record.
+  Stream.EmitRecord(Code, Vals, AbbrevToUse);
+}
+
+uint64_t DXILBitcodeWriter::getAttrKindEncoding(Attribute::AttrKind Kind) {
+  switch (Kind) {
+  case Attribute::Alignment:
+    return bitc::ATTR_KIND_ALIGNMENT;
+  case Attribute::AllocAlign:
+    return bitc::ATTR_KIND_ALLOC_ALIGN;
+  case Attribute::AllocSize:
+    return bitc::ATTR_KIND_ALLOC_SIZE;
+  case Attribute::AlwaysInline:
+    return bitc::ATTR_KIND_ALWAYS_INLINE;
+  case Attribute::ArgMemOnly:
+    return bitc::ATTR_KIND_ARGMEMONLY;
+  case Attribute::Builtin:
+    return bitc::ATTR_KIND_BUILTIN;
+  case Attribute::ByVal:
+    return bitc::ATTR_KIND_BY_VAL;
+  case Attribute::Convergent:
+    return bitc::ATTR_KIND_CONVERGENT;
+  case Attribute::InAlloca:
+    return bitc::ATTR_KIND_IN_ALLOCA;
+  case Attribute::Cold:
+    return bitc::ATTR_KIND_COLD;
+  case Attribute::DisableSanitizerInstrumentation:
+    return bitc::ATTR_KIND_DISABLE_SANITIZER_INSTRUMENTATION;
+  case Attribute::Hot:
+    return bitc::ATTR_KIND_HOT;
+  case Attribute::ElementType:
+    return bitc::ATTR_KIND_ELEMENTTYPE;
+  case Attribute::InaccessibleMemOnly:
+    return bitc::ATTR_KIND_INACCESSIBLEMEM_ONLY;
+  case Attribute::InaccessibleMemOrArgMemOnly:
+    return bitc::ATTR_KIND_INACCESSIBLEMEM_OR_ARGMEMONLY;
+  case Attribute::InlineHint:
+    return bitc::ATTR_KIND_INLINE_HINT;
+  case Attribute::InReg:
+    return bitc::ATTR_KIND_IN_REG;
+  case Attribute::JumpTable:
+    return bitc::ATTR_KIND_JUMP_TABLE;
+  case Attribute::MinSize:
+    return bitc::ATTR_KIND_MIN_SIZE;
+  case Attribute::Naked:
+    return bitc::ATTR_KIND_NAKED;
+  case Attribute::Nest:
+    return bitc::ATTR_KIND_NEST;
+  case Attribute::NoAlias:
+    return bitc::ATTR_KIND_NO_ALIAS;
+  case Attribute::NoBuiltin:
+    return bitc::ATTR_KIND_NO_BUILTIN;
+  case Attribute::NoCallback:
+    return bitc::ATTR_KIND_NO_CALLBACK;
+  case Attribute::NoCapture:
+    return bitc::ATTR_KIND_NO_CAPTURE;
+  case Attribute::NoDuplicate:
+    return bitc::ATTR_KIND_NO_DUPLICATE;
+  case Attribute::NoFree:
+    return bitc::ATTR_KIND_NOFREE;
+  case Attribute::NoImplicitFloat:
+    return bitc::ATTR_KIND_NO_IMPLICIT_FLOAT;
+  case Attribute::NoInline:
+    return bitc::ATTR_KIND_NO_INLINE;
+  case Attribute::NoRecurse:
+    return bitc::ATTR_KIND_NO_RECURSE;
+  case Attribute::NoMerge:
+    return bitc::ATTR_KIND_NO_MERGE;
+  case Attribute::NonLazyBind:
+    return bitc::ATTR_KIND_NON_LAZY_BIND;
+  case Attribute::NonNull:
+    return bitc::ATTR_KIND_NON_NULL;
+  case Attribute::Dereferenceable:
+    return bitc::ATTR_KIND_DEREFERENCEABLE;
+  case Attribute::DereferenceableOrNull:
+    return bitc::ATTR_KIND_DEREFERENCEABLE_OR_NULL;
+  case Attribute::NoRedZone:
+    return bitc::ATTR_KIND_NO_RED_ZONE;
+  case Attribute::NoReturn:
+    return bitc::ATTR_KIND_NO_RETURN;
+  case Attribute::NoSync:
+    return bitc::ATTR_KIND_NOSYNC;
+  case Attribute::NoCfCheck:
+    return bitc::ATTR_KIND_NOCF_CHECK;
+  case Attribute::NoProfile:
+    return bitc::ATTR_KIND_NO_PROFILE;
+  case Attribute::NoUnwind:
+    return bitc::ATTR_KIND_NO_UNWIND;
+  case Attribute::NoSanitizeBounds:
+    return bitc::ATTR_KIND_NO_SANITIZE_BOUNDS;
+  case Attribute::NoSanitizeCoverage:
+    return bitc::ATTR_KIND_NO_SANITIZE_COVERAGE;
+  case Attribute::NullPointerIsValid:
+    return bitc::ATTR_KIND_NULL_POINTER_IS_VALID;
+  case Attribute::OptForFuzzing:
+    return bitc::ATTR_KIND_OPT_FOR_FUZZING;
+  case Attribute::OptimizeForSize:
+    return bitc::ATTR_KIND_OPTIMIZE_FOR_SIZE;
+  case Attribute::OptimizeNone:
+    return bitc::ATTR_KIND_OPTIMIZE_NONE;
+  case Attribute::ReadNone:
+    return bitc::ATTR_KIND_READ_NONE;
+  case Attribute::ReadOnly:
+    return bitc::ATTR_KIND_READ_ONLY;
+  case Attribute::Returned:
+    return bitc::ATTR_KIND_RETURNED;
+  case Attribute::ReturnsTwice:
+    return bitc::ATTR_KIND_RETURNS_TWICE;
+  case Attribute::SExt:
+    return bitc::ATTR_KIND_S_EXT;
+  case Attribute::Speculatable:
+    return bitc::ATTR_KIND_SPECULATABLE;
+  case Attribute::StackAlignment:
+    return bitc::ATTR_KIND_STACK_ALIGNMENT;
+  case Attribute::StackProtect:
+    return bitc::ATTR_KIND_STACK_PROTECT;
+  case Attribute::StackProtectReq:
+    return bitc::ATTR_KIND_STACK_PROTECT_REQ;
+  case Attribute::StackProtectStrong:
+    return bitc::ATTR_KIND_STACK_PROTECT_STRONG;
+  case Attribute::SafeStack:
+    return bitc::ATTR_KIND_SAFESTACK;
+  case Attribute::ShadowCallStack:
+    return bitc::ATTR_KIND_SHADOWCALLSTACK;
+  case Attribute::StrictFP:
+    return bitc::ATTR_KIND_STRICT_FP;
+  case Attribute::StructRet:
+    return bitc::ATTR_KIND_STRUCT_RET;
+  case Attribute::SanitizeAddress:
+    return bitc::ATTR_KIND_SANITIZE_ADDRESS;
+  case Attribute::SanitizeHWAddress:
+    return bitc::ATTR_KIND_SANITIZE_HWADDRESS;
+  case Attribute::SanitizeThread:
+    return bitc::ATTR_KIND_SANITIZE_THREAD;
+  case Attribute::SanitizeMemory:
+    return bitc::ATTR_KIND_SANITIZE_MEMORY;
+  case Attribute::SpeculativeLoadHardening:
+    return bitc::ATTR_KIND_SPECULATIVE_LOAD_HARDENING;
+  case Attribute::SwiftError:
+    return bitc::ATTR_KIND_SWIFT_ERROR;
+  case Attribute::SwiftSelf:
+    return bitc::ATTR_KIND_SWIFT_SELF;
+  case Attribute::SwiftAsync:
+    return bitc::ATTR_KIND_SWIFT_ASYNC;
+  case Attribute::UWTable:
+    return bitc::ATTR_KIND_UW_TABLE;
+  case Attribute::VScaleRange:
+    return bitc::ATTR_KIND_VSCALE_RANGE;
+  case Attribute::WillReturn:
+    return bitc::ATTR_KIND_WILLRETURN;
+  case Attribute::WriteOnly:
+    return bitc::ATTR_KIND_WRITEONLY;
+  case Attribute::ZExt:
+    return bitc::ATTR_KIND_Z_EXT;
+  case Attribute::ImmArg:
+    return bitc::ATTR_KIND_IMMARG;
+  case Attribute::SanitizeMemTag:
+    return bitc::ATTR_KIND_SANITIZE_MEMTAG;
+  case Attribute::Preallocated:
+    return bitc::ATTR_KIND_PREALLOCATED;
+  case Attribute::NoUndef:
+    return bitc::ATTR_KIND_NOUNDEF;
+  case Attribute::ByRef:
+    return bitc::ATTR_KIND_BYREF;
+  case Attribute::MustProgress:
+    return bitc::ATTR_KIND_MUSTPROGRESS;
+  case Attribute::EndAttrKinds:
+    llvm_unreachable("Can not encode end-attribute kinds marker.");
+  case Attribute::None:
+    llvm_unreachable("Can not encode none-attribute.");
+  case Attribute::EmptyKey:
+  case Attribute::TombstoneKey:
+    llvm_unreachable("Trying to encode EmptyKey/TombstoneKey");
+  }
+
+  llvm_unreachable("Trying to encode unknown attribute");
+}
+
+void DXILBitcodeWriter::emitSignedInt64(SmallVectorImpl<uint64_t> &Vals,
+                                        uint64_t V) {
+  if ((int64_t)V >= 0)
+    Vals.push_back(V << 1);
+  else
+    Vals.push_back((-V << 1) | 1);
+}
+
+void DXILBitcodeWriter::emitWideAPInt(SmallVectorImpl<uint64_t> &Vals,
+                                      const APInt &A) {
+  // We have an arbitrary precision integer value to write whose
+  // bit width is > 64. However, in canonical unsigned integer
+  // format it is likely that the high bits are going to be zero.
+  // So, we only write the number of active words.
+  unsigned NumWords = A.getActiveWords();
+  const uint64_t *RawData = A.getRawData();
+  for (unsigned i = 0; i < NumWords; i++)
+    emitSignedInt64(Vals, RawData[i]);
+}
+
+uint64_t DXILBitcodeWriter::getOptimizationFlags(const Value *V) {
+  uint64_t Flags = 0;
+
+  if (const auto *OBO = dyn_cast<OverflowingBinaryOperator>(V)) {
+    if (OBO->hasNoSignedWrap())
+      Flags |= 1 << bitc::OBO_NO_SIGNED_WRAP;
+    if (OBO->hasNoUnsignedWrap())
+      Flags |= 1 << bitc::OBO_NO_UNSIGNED_WRAP;
+  } else if (const auto *PEO = dyn_cast<PossiblyExactOperator>(V)) {
+    if (PEO->isExact())
+      Flags |= 1 << bitc::PEO_EXACT;
+  } else if (const auto *FPMO = dyn_cast<FPMathOperator>(V)) {
+    if (FPMO->hasAllowReassoc())
+      Flags |= bitc::AllowReassoc;
+    if (FPMO->hasNoNaNs())
+      Flags |= bitc::NoNaNs;
+    if (FPMO->hasNoInfs())
+      Flags |= bitc::NoInfs;
+    if (FPMO->hasNoSignedZeros())
+      Flags |= bitc::NoSignedZeros;
+    if (FPMO->hasAllowReciprocal())
+      Flags |= bitc::AllowReciprocal;
+    if (FPMO->hasAllowContract())
+      Flags |= bitc::AllowContract;
+    if (FPMO->hasApproxFunc())
+      Flags |= bitc::ApproxFunc;
+  }
+
+  return Flags;
+}
+
+unsigned
+DXILBitcodeWriter::getEncodedLinkage(const GlobalValue::LinkageTypes Linkage) {
+  switch (Linkage) {
+  case GlobalValue::ExternalLinkage:
+    return 0;
+  case GlobalValue::WeakAnyLinkage:
+    return 16;
+  case GlobalValue::AppendingLinkage:
+    return 2;
+  case GlobalValue::InternalLinkage:
+    return 3;
+  case GlobalValue::LinkOnceAnyLinkage:
+    return 18;
+  case GlobalValue::ExternalWeakLinkage:
+    return 7;
+  case GlobalValue::CommonLinkage:
+    return 8;
+  case GlobalValue::PrivateLinkage:
+    return 9;
+  case GlobalValue::WeakODRLinkage:
+    return 17;
+  case GlobalValue::LinkOnceODRLinkage:
+    return 19;
+  case GlobalValue::AvailableExternallyLinkage:
+    return 12;
+  }
+  llvm_unreachable("Invalid linkage");
+}
+
+unsigned DXILBitcodeWriter::getEncodedLinkage(const GlobalValue &GV) {
+  return getEncodedLinkage(GV.getLinkage());
+}
+
+unsigned DXILBitcodeWriter::getEncodedVisibility(const GlobalValue &GV) {
+  switch (GV.getVisibility()) {
+  case GlobalValue::DefaultVisibility:   return 0;
+  case GlobalValue::HiddenVisibility:    return 1;
+  case GlobalValue::ProtectedVisibility: return 2;
+  }
+  llvm_unreachable("Invalid visibility");
+}
+
+unsigned DXILBitcodeWriter::getEncodedDLLStorageClass(const GlobalValue &GV) {
+  switch (GV.getDLLStorageClass()) {
+  case GlobalValue::DefaultStorageClass:   return 0;
+  case GlobalValue::DLLImportStorageClass: return 1;
+  case GlobalValue::DLLExportStorageClass: return 2;
+  }
+  llvm_unreachable("Invalid DLL storage class");
+}
+
+unsigned DXILBitcodeWriter::getEncodedThreadLocalMode(const GlobalValue &GV) {
+  switch (GV.getThreadLocalMode()) {
+    case GlobalVariable::NotThreadLocal:         return 0;
+    case GlobalVariable::GeneralDynamicTLSModel: return 1;
+    case GlobalVariable::LocalDynamicTLSModel:   return 2;
+    case GlobalVariable::InitialExecTLSModel:    return 3;
+    case GlobalVariable::LocalExecTLSModel:      return 4;
+  }
+  llvm_unreachable("Invalid TLS model");
+}
+
+unsigned DXILBitcodeWriter::getEncodedComdatSelectionKind(const Comdat &C) {
+  switch (C.getSelectionKind()) {
+  case Comdat::Any:
+    return bitc::COMDAT_SELECTION_KIND_ANY;
+  case Comdat::ExactMatch:
+    return bitc::COMDAT_SELECTION_KIND_EXACT_MATCH;
+  case Comdat::Largest:
+    return bitc::COMDAT_SELECTION_KIND_LARGEST;
+  case Comdat::NoDeduplicate:
+    return bitc::COMDAT_SELECTION_KIND_NO_DUPLICATES;
+  case Comdat::SameSize:
+    return bitc::COMDAT_SELECTION_KIND_SAME_SIZE;
+  }
+  llvm_unreachable("Invalid selection kind");
+}
+
+////////////////////////////////////////////////////////////////////////////////
+/// Begin DXILBitcodeWriter Implementation
+////////////////////////////////////////////////////////////////////////////////
+
+void DXILBitcodeWriter::writeAttributeGroupTable() {
+  const std::vector<ValueEnumerator::IndexAndAttrSet> &AttrGrps =
+      VE.getAttributeGroups();
+  if (AttrGrps.empty())
+    return;
+
+  Stream.EnterSubblock(bitc::PARAMATTR_GROUP_BLOCK_ID, 3);
+
+  SmallVector<uint64_t, 64> Record;
+  for (ValueEnumerator::IndexAndAttrSet Pair : AttrGrps) {
+    unsigned AttrListIndex = Pair.first;
+    AttributeSet AS = Pair.second;
+    Record.push_back(VE.getAttributeGroupID(Pair));
+    Record.push_back(AttrListIndex);
+
+    for (Attribute Attr : AS) {
+      if (Attr.isEnumAttribute()) {
+        uint64_t Val = getAttrKindEncoding(Attr.getKindAsEnum());
+        assert(Val <= bitc::ATTR_KIND_ARGMEMONLY &&
+               "DXIL does not support attributes above ATTR_KIND_ARGMEMONLY");
+        Record.push_back(0);
+        Record.push_back(Val);
+      } else if (Attr.isIntAttribute()) {
+        uint64_t Val = getAttrKindEncoding(Attr.getKindAsEnum());
+        assert(Val <= bitc::ATTR_KIND_ARGMEMONLY &&
+               "DXIL does not support attributes above ATTR_KIND_ARGMEMONLY");
+        Record.push_back(1);
+        Record.push_back(Val);
+        Record.push_back(Attr.getValueAsInt());
+      } else {
+        StringRef Kind = Attr.getKindAsString();
+        StringRef Val = Attr.getValueAsString();
+
+        Record.push_back(Val.empty() ? 3 : 4);
+        Record.append(Kind.begin(), Kind.end());
+        Record.push_back(0);
+        if (!Val.empty()) {
+          Record.append(Val.begin(), Val.end());
+          Record.push_back(0);
+        }
+      }
+    }
+
+    Stream.EmitRecord(bitc::PARAMATTR_GRP_CODE_ENTRY, Record);
+    Record.clear();
+  }
+
+  Stream.ExitBlock();
+}
+
+void DXILBitcodeWriter::writeAttributeTable() {
+  const std::vector<AttributeList> &Attrs = VE.getAttributeLists();
+  if (Attrs.empty())
+    return;
+
+  Stream.EnterSubblock(bitc::PARAMATTR_BLOCK_ID, 3);
+
+  SmallVector<uint64_t, 64> Record;
+  for (unsigned i = 0, e = Attrs.size(); i != e; ++i) {
+    AttributeList AL = Attrs[i];
+    for (unsigned i : AL.indexes()) {
+      AttributeSet AS = AL.getAttributes(i);
+      if (AS.hasAttributes())
+        Record.push_back(VE.getAttributeGroupID({i, AS}));
+    }
+
+    Stream.EmitRecord(bitc::PARAMATTR_CODE_ENTRY, Record);
+    Record.clear();
+  }
+
+  Stream.ExitBlock();
+}
+
+/// WriteTypeTable - Write out the type table for a module.
+void DXILBitcodeWriter::writeTypeTable() {
+  const ValueEnumerator::TypeList &TypeList = VE.getTypes();
+
+  Stream.EnterSubblock(bitc::TYPE_BLOCK_ID_NEW, 4 /*count from # abbrevs */);
+  SmallVector<uint64_t, 64> TypeVals;
+
+  uint64_t NumBits = VE.computeBitsRequiredForTypeIndicies();
+
+  // Abbrev for TYPE_CODE_POINTER.
+  auto Abbv = std::make_shared<BitCodeAbbrev>();
+  Abbv->Add(BitCodeAbbrevOp(bitc::TYPE_CODE_POINTER));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, NumBits));
+  Abbv->Add(BitCodeAbbrevOp(0)); // Addrspace = 0
+  unsigned PtrAbbrev = Stream.EmitAbbrev(std::move(Abbv));
+
+  // Abbrev for TYPE_CODE_FUNCTION.
+  Abbv = std::make_shared<BitCodeAbbrev>();
+  Abbv->Add(BitCodeAbbrevOp(bitc::TYPE_CODE_FUNCTION));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 1)); // isvararg
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, NumBits));
+  unsigned FunctionAbbrev = Stream.EmitAbbrev(std::move(Abbv));
+
+  // Abbrev for TYPE_CODE_STRUCT_ANON.
+  Abbv = std::make_shared<BitCodeAbbrev>();
+  Abbv->Add(BitCodeAbbrevOp(bitc::TYPE_CODE_STRUCT_ANON));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 1)); // ispacked
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, NumBits));
+  unsigned StructAnonAbbrev = Stream.EmitAbbrev(std::move(Abbv));
+
+  // Abbrev for TYPE_CODE_STRUCT_NAME.
+  Abbv = std::make_shared<BitCodeAbbrev>();
+  Abbv->Add(BitCodeAbbrevOp(bitc::TYPE_CODE_STRUCT_NAME));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Char6));
+  unsigned StructNameAbbrev = Stream.EmitAbbrev(std::move(Abbv));
+
+  // Abbrev for TYPE_CODE_STRUCT_NAMED.
+  Abbv = std::make_shared<BitCodeAbbrev>();
+  Abbv->Add(BitCodeAbbrevOp(bitc::TYPE_CODE_STRUCT_NAMED));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 1)); // ispacked
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, NumBits));
+  unsigned StructNamedAbbrev = Stream.EmitAbbrev(std::move(Abbv));
+
+  // Abbrev for TYPE_CODE_ARRAY.
+  Abbv = std::make_shared<BitCodeAbbrev>();
+  Abbv->Add(BitCodeAbbrevOp(bitc::TYPE_CODE_ARRAY));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8)); // size
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, NumBits));
+  unsigned ArrayAbbrev = Stream.EmitAbbrev(std::move(Abbv));
+
+  // Emit an entry count so the reader can reserve space.
+  TypeVals.push_back(TypeList.size());
+  Stream.EmitRecord(bitc::TYPE_CODE_NUMENTRY, TypeVals);
+  TypeVals.clear();
+
+  // Loop over all of the types, emitting each in turn.
+  for (Type *T : TypeList) {
+    int AbbrevToUse = 0;
+    unsigned Code = 0;
+
+    switch (T->getTypeID()) {
+    case Type::BFloatTyID:
+    case Type::X86_AMXTyID:
+    case Type::TokenTyID:
+      llvm_unreachable("These should never be used!!!");
+      break;
+    case Type::VoidTyID:
+      Code = bitc::TYPE_CODE_VOID;
+      break;
+    case Type::HalfTyID:
+      Code = bitc::TYPE_CODE_HALF;
+      break;
+    case Type::FloatTyID:
+      Code = bitc::TYPE_CODE_FLOAT;
+      break;
+    case Type::DoubleTyID:
+      Code = bitc::TYPE_CODE_DOUBLE;
+      break;
+    case Type::X86_FP80TyID:
+      Code = bitc::TYPE_CODE_X86_FP80;
+      break;
+    case Type::FP128TyID:
+      Code = bitc::TYPE_CODE_FP128;
+      break;
+    case Type::PPC_FP128TyID:
+      Code = bitc::TYPE_CODE_PPC_FP128;
+      break;
+    case Type::LabelTyID:
+      Code = bitc::TYPE_CODE_LABEL;
+      break;
+    case Type::MetadataTyID:
+      Code = bitc::TYPE_CODE_METADATA;
+      break;
+    case Type::X86_MMXTyID:
+      Code = bitc::TYPE_CODE_X86_MMX;
+      break;
+    case Type::IntegerTyID:
+      // INTEGER: [width]
+      Code = bitc::TYPE_CODE_INTEGER;
+      TypeVals.push_back(cast<IntegerType>(T)->getBitWidth());
+      break;
+    case Type::PointerTyID: {
+      PointerType *PTy = cast<PointerType>(T);
+      // POINTER: [pointee type, address space]
+      Code = bitc::TYPE_CODE_POINTER;
+      TypeVals.push_back(VE.getTypeID(PTy->getNonOpaquePointerElementType()));
+      unsigned AddressSpace = PTy->getAddressSpace();
+      TypeVals.push_back(AddressSpace);
+      if (AddressSpace == 0)
+        AbbrevToUse = PtrAbbrev;
+      break;
+    }
+    case Type::FunctionTyID: {
+      FunctionType *FT = cast<FunctionType>(T);
+      // FUNCTION: [isvararg, retty, paramty x N]
+      Code = bitc::TYPE_CODE_FUNCTION;
+      TypeVals.push_back(FT->isVarArg());
+      TypeVals.push_back(VE.getTypeID(FT->getReturnType()));
+      for (Type *PTy : FT->params())
+        TypeVals.push_back(VE.getTypeID(PTy));
+      AbbrevToUse = FunctionAbbrev;
+      break;
+    }
+    case Type::StructTyID: {
+      StructType *ST = cast<StructType>(T);
+      // STRUCT: [ispacked, eltty x N]
+      TypeVals.push_back(ST->isPacked());
+      // Output all of the element types.
+      for (Type *ElTy : ST->elements())
+        TypeVals.push_back(VE.getTypeID(ElTy));
+
+      if (ST->isLiteral()) {
+        Code = bitc::TYPE_CODE_STRUCT_ANON;
+        AbbrevToUse = StructAnonAbbrev;
+      } else {
+        if (ST->isOpaque()) {
+          Code = bitc::TYPE_CODE_OPAQUE;
+        } else {
+          Code = bitc::TYPE_CODE_STRUCT_NAMED;
+          AbbrevToUse = StructNamedAbbrev;
+        }
+
+        // Emit the name if it is present.
+        if (!ST->getName().empty())
+          writeStringRecord(Stream, bitc::TYPE_CODE_STRUCT_NAME, ST->getName(),
+                            StructNameAbbrev);
+      }
+      break;
+    }
+    case Type::ArrayTyID: {
+      ArrayType *AT = cast<ArrayType>(T);
+      // ARRAY: [numelts, eltty]
+      Code = bitc::TYPE_CODE_ARRAY;
+      TypeVals.push_back(AT->getNumElements());
+      TypeVals.push_back(VE.getTypeID(AT->getElementType()));
+      AbbrevToUse = ArrayAbbrev;
+      break;
+    }
+    case Type::FixedVectorTyID:
+    case Type::ScalableVectorTyID: {
+      VectorType *VT = cast<VectorType>(T);
+      // VECTOR [numelts, eltty]
+      Code = bitc::TYPE_CODE_VECTOR;
+      TypeVals.push_back(VT->getElementCount().getKnownMinValue());
+      TypeVals.push_back(VE.getTypeID(VT->getElementType()));
+      break;
+    }
+    }
+
+    // Emit the finished record.
+    Stream.EmitRecord(Code, TypeVals, AbbrevToUse);
+    TypeVals.clear();
+  }
+
+  Stream.ExitBlock();
+}
+
+void DXILBitcodeWriter::writeComdats() {
+  SmallVector<uint16_t, 64> Vals;
+  for (const Comdat *C : VE.getComdats()) {
+    // COMDAT: [selection_kind, name]
+    Vals.push_back(getEncodedComdatSelectionKind(*C));
+    size_t Size = C->getName().size();
+    assert(isUInt<16>(Size));
+    Vals.push_back(Size);
+    for (char Chr : C->getName())
+      Vals.push_back((unsigned char)Chr);
+    Stream.EmitRecord(bitc::MODULE_CODE_COMDAT, Vals, /*AbbrevToUse=*/0);
+    Vals.clear();
+  }
+}
+
+void DXILBitcodeWriter::writeValueSymbolTableForwardDecl() {}
+
+/// Emit top-level description of module, including target triple, inline asm,
+/// descriptors for global variables, and function prototype info.
+/// Returns the bit offset to backpatch with the location of the real VST.
+void DXILBitcodeWriter::writeModuleInfo() {
+  // Emit various pieces of data attached to a module.
+  if (!M.getTargetTriple().empty())
+    writeStringRecord(Stream, bitc::MODULE_CODE_TRIPLE, M.getTargetTriple(),
+                      0 /*TODO*/);
+  const std::string &DL = M.getDataLayoutStr();
+  if (!DL.empty())
+    writeStringRecord(Stream, bitc::MODULE_CODE_DATALAYOUT, DL, 0 /*TODO*/);
+  if (!M.getModuleInlineAsm().empty())
+    writeStringRecord(Stream, bitc::MODULE_CODE_ASM, M.getModuleInlineAsm(),
+                      0 /*TODO*/);
+
+  // Emit information about sections and GC, computing how many there are. Also
+  // compute the maximum alignment value.
+  std::map<std::string, unsigned> SectionMap;
+  std::map<std::string, unsigned> GCMap;
+  MaybeAlign MaxAlignment;
+  unsigned MaxGlobalType = 0;
+  const auto UpdateMaxAlignment = [&MaxAlignment](const MaybeAlign A) {
+    if (A)
+      MaxAlignment = !MaxAlignment ? *A : std::max(*MaxAlignment, *A);
+  };
+  for (const GlobalVariable &GV : M.globals()) {
+    UpdateMaxAlignment(GV.getAlign());
+    MaxGlobalType = std::max(MaxGlobalType, VE.getTypeID(GV.getValueType()));
+    if (GV.hasSection()) {
+      // Give section names unique ID's.
+      unsigned &Entry = SectionMap[std::string(GV.getSection())];
+      if (!Entry) {
+        writeStringRecord(Stream, bitc::MODULE_CODE_SECTIONNAME,
+                          GV.getSection(), 0 /*TODO*/);
+        Entry = SectionMap.size();
+      }
+    }
+  }
+  for (const Function &F : M) {
+    UpdateMaxAlignment(F.getAlign());
+    if (F.hasSection()) {
+      // Give section names unique ID's.
+      unsigned &Entry = SectionMap[std::string(F.getSection())];
+      if (!Entry) {
+        writeStringRecord(Stream, bitc::MODULE_CODE_SECTIONNAME, F.getSection(),
+                          0 /*TODO*/);
+        Entry = SectionMap.size();
+      }
+    }
+    if (F.hasGC()) {
+      // Same for GC names.
+      unsigned &Entry = GCMap[F.getGC()];
+      if (!Entry) {
+        writeStringRecord(Stream, bitc::MODULE_CODE_GCNAME, F.getGC(),
+                          0 /*TODO*/);
+        Entry = GCMap.size();
+      }
+    }
+  }
+
+  // Emit abbrev for globals, now that we know # sections and max alignment.
+  unsigned SimpleGVarAbbrev = 0;
+  if (!M.global_empty()) {
+    // Add an abbrev for common globals with no visibility or thread localness.
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::MODULE_CODE_GLOBALVAR));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed,
+                              Log2_32_Ceil(MaxGlobalType + 1)));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 6));   // AddrSpace << 2
+                                                           //| explicitType << 1
+                                                           //| constant
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 6));   // Initializer.
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 5)); // Linkage.
+    if (MaxAlignment == 0)                                 // Alignment.
+      Abbv->Add(BitCodeAbbrevOp(0));
+    else {
+      unsigned MaxEncAlignment = getEncodedAlign(MaxAlignment);
+      Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed,
+                                Log2_32_Ceil(MaxEncAlignment + 1)));
+    }
+    if (SectionMap.empty()) // Section.
+      Abbv->Add(BitCodeAbbrevOp(0));
+    else
+      Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed,
+                                Log2_32_Ceil(SectionMap.size() + 1)));
+    // Don't bother emitting vis + thread local.
+    SimpleGVarAbbrev = Stream.EmitAbbrev(std::move(Abbv));
+  }
+
+  // Emit the global variable information.
+  SmallVector<unsigned, 64> Vals;
+  for (const GlobalVariable &GV : M.globals()) {
+    unsigned AbbrevToUse = 0;
+
+    // GLOBALVAR: [type, isconst, initid,
+    //             linkage, alignment, section, visibility, threadlocal,
+    //             unnamed_addr, externally_initialized, dllstorageclass,
+    //             comdat]
+    Vals.push_back(VE.getTypeID(GV.getValueType()));
+    Vals.push_back(
+        GV.getType()->getAddressSpace() << 2 | 2 |
+        (GV.isConstant() ? 1 : 0)); // HLSL Change - bitwise | was used with
+                                    // unsigned int and bool
+    Vals.push_back(
+        GV.isDeclaration() ? 0 : (VE.getValueID(GV.getInitializer()) + 1));
+    Vals.push_back(getEncodedLinkage(GV));
+    Vals.push_back(getEncodedAlign(GV.getAlign()));
+    Vals.push_back(GV.hasSection() ? SectionMap[std::string(GV.getSection())]
+                                   : 0);
+    if (GV.isThreadLocal() ||
+        GV.getVisibility() != GlobalValue::DefaultVisibility ||
+        GV.getUnnamedAddr() != GlobalValue::UnnamedAddr::None ||
+        GV.isExternallyInitialized() ||
+        GV.getDLLStorageClass() != GlobalValue::DefaultStorageClass ||
+        GV.hasComdat()) {
+      Vals.push_back(getEncodedVisibility(GV));
+      Vals.push_back(getEncodedThreadLocalMode(GV));
+      Vals.push_back(GV.getUnnamedAddr() != GlobalValue::UnnamedAddr::None);
+      Vals.push_back(GV.isExternallyInitialized());
+      Vals.push_back(getEncodedDLLStorageClass(GV));
+      Vals.push_back(GV.hasComdat() ? VE.getComdatID(GV.getComdat()) : 0);
+    } else {
+      AbbrevToUse = SimpleGVarAbbrev;
+    }
+
+    Stream.EmitRecord(bitc::MODULE_CODE_GLOBALVAR, Vals, AbbrevToUse);
+    Vals.clear();
+  }
+
+  // Emit the function proto information.
+  for (const Function &F : M) {
+    // FUNCTION:  [type, callingconv, isproto, linkage, paramattrs, alignment,
+    //             section, visibility, gc, unnamed_addr, prologuedata,
+    //             dllstorageclass, comdat, prefixdata, personalityfn]
+    Vals.push_back(VE.getTypeID(F.getFunctionType()));
+    Vals.push_back(F.getCallingConv());
+    Vals.push_back(F.isDeclaration());
+    Vals.push_back(getEncodedLinkage(F));
+    Vals.push_back(VE.getAttributeListID(F.getAttributes()));
+    Vals.push_back(getEncodedAlign(F.getAlign()));
+    Vals.push_back(F.hasSection() ? SectionMap[std::string(F.getSection())]
+                                  : 0);
+    Vals.push_back(getEncodedVisibility(F));
+    Vals.push_back(F.hasGC() ? GCMap[F.getGC()] : 0);
+    Vals.push_back(F.getUnnamedAddr() != GlobalValue::UnnamedAddr::None);
+    Vals.push_back(
+        F.hasPrologueData() ? (VE.getValueID(F.getPrologueData()) + 1) : 0);
+    Vals.push_back(getEncodedDLLStorageClass(F));
+    Vals.push_back(F.hasComdat() ? VE.getComdatID(F.getComdat()) : 0);
+    Vals.push_back(F.hasPrefixData() ? (VE.getValueID(F.getPrefixData()) + 1)
+                                     : 0);
+    Vals.push_back(
+        F.hasPersonalityFn() ? (VE.getValueID(F.getPersonalityFn()) + 1) : 0);
+
+    unsigned AbbrevToUse = 0;
+    Stream.EmitRecord(bitc::MODULE_CODE_FUNCTION, Vals, AbbrevToUse);
+    Vals.clear();
+  }
+
+  // Emit the alias information.
+  for (const GlobalAlias &A : M.aliases()) {
+    // ALIAS: [alias type, aliasee val#, linkage, visibility]
+    Vals.push_back(VE.getTypeID(A.getValueType()));
+    Vals.push_back(VE.getValueID(A.getAliasee()));
+    Vals.push_back(getEncodedLinkage(A));
+    Vals.push_back(getEncodedVisibility(A));
+    Vals.push_back(getEncodedDLLStorageClass(A));
+    Vals.push_back(getEncodedThreadLocalMode(A));
+    Vals.push_back(A.getUnnamedAddr() != GlobalValue::UnnamedAddr::None);
+    unsigned AbbrevToUse = 0;
+    Stream.EmitRecord(bitc::MODULE_CODE_ALIAS_OLD, Vals, AbbrevToUse);
+    Vals.clear();
+  }
+}
+
+void DXILBitcodeWriter::writeValueAsMetadata(
+    const ValueAsMetadata *MD, SmallVectorImpl<uint64_t> &Record) {
+  // Mimic an MDNode with a value as one operand.
+  Value *V = MD->getValue();
+  Record.push_back(VE.getTypeID(V->getType()));
+  Record.push_back(VE.getValueID(V));
+  Stream.EmitRecord(bitc::METADATA_VALUE, Record, 0);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeMDTuple(const MDTuple *N,
+                                     SmallVectorImpl<uint64_t> &Record,
+                                     unsigned Abbrev) {
+  for (unsigned i = 0, e = N->getNumOperands(); i != e; ++i) {
+    Metadata *MD = N->getOperand(i);
+    assert(!(MD && isa<LocalAsMetadata>(MD)) &&
+           "Unexpected function-local metadata");
+    Record.push_back(VE.getMetadataOrNullID(MD));
+  }
+  Stream.EmitRecord(N->isDistinct() ? bitc::METADATA_DISTINCT_NODE
+                                    : bitc::METADATA_NODE,
+                    Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDILocation(const DILocation *N,
+                                        SmallVectorImpl<uint64_t> &Record,
+                                        unsigned &Abbrev) {
+  if (!Abbrev)
+    Abbrev = createDILocationAbbrev();
+  Record.push_back(N->isDistinct());
+  Record.push_back(N->getLine());
+  Record.push_back(N->getColumn());
+  Record.push_back(VE.getMetadataID(N->getScope()));
+  Record.push_back(VE.getMetadataOrNullID(N->getInlinedAt()));
+
+  Stream.EmitRecord(bitc::METADATA_LOCATION, Record, Abbrev);
+  Record.clear();
+}
+
+static uint64_t rotateSign(APInt Val) {
+  int64_t I = Val.getSExtValue();
+  uint64_t U = I;
+  return I < 0 ? ~(U << 1) : U << 1;
+}
+
+static uint64_t rotateSign(DISubrange::BoundType Val) {
+  return rotateSign(Val.get<ConstantInt *>()->getValue());
+}
+
+void DXILBitcodeWriter::writeDISubrange(const DISubrange *N,
+                                        SmallVectorImpl<uint64_t> &Record,
+                                        unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  Record.push_back(
+      N->getCount().get<ConstantInt *>()->getValue().getSExtValue());
+  Record.push_back(rotateSign(N->getLowerBound()));
+
+  Stream.EmitRecord(bitc::METADATA_SUBRANGE, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDIEnumerator(const DIEnumerator *N,
+                                          SmallVectorImpl<uint64_t> &Record,
+                                          unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  Record.push_back(rotateSign(N->getValue()));
+  Record.push_back(VE.getMetadataOrNullID(N->getRawName()));
+
+  Stream.EmitRecord(bitc::METADATA_ENUMERATOR, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDIBasicType(const DIBasicType *N,
+                                         SmallVectorImpl<uint64_t> &Record,
+                                         unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  Record.push_back(N->getTag());
+  Record.push_back(VE.getMetadataOrNullID(N->getRawName()));
+  Record.push_back(N->getSizeInBits());
+  Record.push_back(N->getAlignInBits());
+  Record.push_back(N->getEncoding());
+
+  Stream.EmitRecord(bitc::METADATA_BASIC_TYPE, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDIDerivedType(const DIDerivedType *N,
+                                           SmallVectorImpl<uint64_t> &Record,
+                                           unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  Record.push_back(N->getTag());
+  Record.push_back(VE.getMetadataOrNullID(N->getRawName()));
+  Record.push_back(VE.getMetadataOrNullID(N->getFile()));
+  Record.push_back(N->getLine());
+  Record.push_back(VE.getMetadataOrNullID(N->getScope()));
+  Record.push_back(VE.getMetadataOrNullID(N->getBaseType()));
+  Record.push_back(N->getSizeInBits());
+  Record.push_back(N->getAlignInBits());
+  Record.push_back(N->getOffsetInBits());
+  Record.push_back(N->getFlags());
+  Record.push_back(VE.getMetadataOrNullID(N->getExtraData()));
+
+  Stream.EmitRecord(bitc::METADATA_DERIVED_TYPE, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDICompositeType(const DICompositeType *N,
+                                             SmallVectorImpl<uint64_t> &Record,
+                                             unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  Record.push_back(N->getTag());
+  Record.push_back(VE.getMetadataOrNullID(N->getRawName()));
+  Record.push_back(VE.getMetadataOrNullID(N->getFile()));
+  Record.push_back(N->getLine());
+  Record.push_back(VE.getMetadataOrNullID(N->getScope()));
+  Record.push_back(VE.getMetadataOrNullID(N->getBaseType()));
+  Record.push_back(N->getSizeInBits());
+  Record.push_back(N->getAlignInBits());
+  Record.push_back(N->getOffsetInBits());
+  Record.push_back(N->getFlags());
+  Record.push_back(VE.getMetadataOrNullID(N->getElements().get()));
+  Record.push_back(N->getRuntimeLang());
+  Record.push_back(VE.getMetadataOrNullID(N->getVTableHolder()));
+  Record.push_back(VE.getMetadataOrNullID(N->getTemplateParams().get()));
+  Record.push_back(VE.getMetadataOrNullID(N->getRawIdentifier()));
+
+  Stream.EmitRecord(bitc::METADATA_COMPOSITE_TYPE, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDISubroutineType(const DISubroutineType *N,
+                                              SmallVectorImpl<uint64_t> &Record,
+                                              unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  Record.push_back(N->getFlags());
+  Record.push_back(VE.getMetadataOrNullID(N->getTypeArray().get()));
+
+  Stream.EmitRecord(bitc::METADATA_SUBROUTINE_TYPE, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDIFile(const DIFile *N,
+                                    SmallVectorImpl<uint64_t> &Record,
+                                    unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  Record.push_back(VE.getMetadataOrNullID(N->getRawFilename()));
+  Record.push_back(VE.getMetadataOrNullID(N->getRawDirectory()));
+
+  Stream.EmitRecord(bitc::METADATA_FILE, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDICompileUnit(const DICompileUnit *N,
+                                           SmallVectorImpl<uint64_t> &Record,
+                                           unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  Record.push_back(N->getSourceLanguage());
+  Record.push_back(VE.getMetadataOrNullID(N->getFile()));
+  Record.push_back(VE.getMetadataOrNullID(N->getRawProducer()));
+  Record.push_back(N->isOptimized());
+  Record.push_back(VE.getMetadataOrNullID(N->getRawFlags()));
+  Record.push_back(N->getRuntimeVersion());
+  Record.push_back(VE.getMetadataOrNullID(N->getRawSplitDebugFilename()));
+  Record.push_back(N->getEmissionKind());
+  Record.push_back(VE.getMetadataOrNullID(N->getEnumTypes().get()));
+  Record.push_back(VE.getMetadataOrNullID(N->getRetainedTypes().get()));
+  Record.push_back(/* subprograms */ 0);
+  Record.push_back(VE.getMetadataOrNullID(N->getGlobalVariables().get()));
+  Record.push_back(VE.getMetadataOrNullID(N->getImportedEntities().get()));
+  Record.push_back(N->getDWOId());
+
+  Stream.EmitRecord(bitc::METADATA_COMPILE_UNIT, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDISubprogram(const DISubprogram *N,
+                                          SmallVectorImpl<uint64_t> &Record,
+                                          unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  Record.push_back(VE.getMetadataOrNullID(N->getScope()));
+  Record.push_back(VE.getMetadataOrNullID(N->getRawName()));
+  Record.push_back(VE.getMetadataOrNullID(N->getRawLinkageName()));
+  Record.push_back(VE.getMetadataOrNullID(N->getFile()));
+  Record.push_back(N->getLine());
+  Record.push_back(VE.getMetadataOrNullID(N->getType()));
+  Record.push_back(N->isLocalToUnit());
+  Record.push_back(N->isDefinition());
+  Record.push_back(N->getScopeLine());
+  Record.push_back(VE.getMetadataOrNullID(N->getContainingType()));
+  Record.push_back(N->getVirtuality());
+  Record.push_back(N->getVirtualIndex());
+  Record.push_back(N->getFlags());
+  Record.push_back(N->isOptimized());
+  Record.push_back(VE.getMetadataOrNullID(N->getRawUnit()));
+  Record.push_back(VE.getMetadataOrNullID(N->getTemplateParams().get()));
+  Record.push_back(VE.getMetadataOrNullID(N->getDeclaration()));
+  Record.push_back(VE.getMetadataOrNullID(N->getRetainedNodes().get()));
+
+  Stream.EmitRecord(bitc::METADATA_SUBPROGRAM, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDILexicalBlock(const DILexicalBlock *N,
+                                            SmallVectorImpl<uint64_t> &Record,
+                                            unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  Record.push_back(VE.getMetadataOrNullID(N->getScope()));
+  Record.push_back(VE.getMetadataOrNullID(N->getFile()));
+  Record.push_back(N->getLine());
+  Record.push_back(N->getColumn());
+
+  Stream.EmitRecord(bitc::METADATA_LEXICAL_BLOCK, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDILexicalBlockFile(
+    const DILexicalBlockFile *N, SmallVectorImpl<uint64_t> &Record,
+    unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  Record.push_back(VE.getMetadataOrNullID(N->getScope()));
+  Record.push_back(VE.getMetadataOrNullID(N->getFile()));
+  Record.push_back(N->getDiscriminator());
+
+  Stream.EmitRecord(bitc::METADATA_LEXICAL_BLOCK_FILE, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDINamespace(const DINamespace *N,
+                                         SmallVectorImpl<uint64_t> &Record,
+                                         unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  Record.push_back(VE.getMetadataOrNullID(N->getScope()));
+  Record.push_back(VE.getMetadataOrNullID(N->getFile()));
+  Record.push_back(VE.getMetadataOrNullID(N->getRawName()));
+  Record.push_back(/* line number */ 0);
+
+  Stream.EmitRecord(bitc::METADATA_NAMESPACE, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDIModule(const DIModule *N,
+                                      SmallVectorImpl<uint64_t> &Record,
+                                      unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  for (auto &I : N->operands())
+    Record.push_back(VE.getMetadataOrNullID(I));
+
+  Stream.EmitRecord(bitc::METADATA_MODULE, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDITemplateTypeParameter(
+    const DITemplateTypeParameter *N, SmallVectorImpl<uint64_t> &Record,
+    unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  Record.push_back(VE.getMetadataOrNullID(N->getRawName()));
+  Record.push_back(VE.getMetadataOrNullID(N->getType()));
+
+  Stream.EmitRecord(bitc::METADATA_TEMPLATE_TYPE, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDITemplateValueParameter(
+    const DITemplateValueParameter *N, SmallVectorImpl<uint64_t> &Record,
+    unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  Record.push_back(N->getTag());
+  Record.push_back(VE.getMetadataOrNullID(N->getRawName()));
+  Record.push_back(VE.getMetadataOrNullID(N->getType()));
+  Record.push_back(VE.getMetadataOrNullID(N->getValue()));
+
+  Stream.EmitRecord(bitc::METADATA_TEMPLATE_VALUE, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDIGlobalVariable(const DIGlobalVariable *N,
+                                              SmallVectorImpl<uint64_t> &Record,
+                                              unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  Record.push_back(VE.getMetadataOrNullID(N->getScope()));
+  Record.push_back(VE.getMetadataOrNullID(N->getRawName()));
+  Record.push_back(VE.getMetadataOrNullID(N->getRawLinkageName()));
+  Record.push_back(VE.getMetadataOrNullID(N->getFile()));
+  Record.push_back(N->getLine());
+  Record.push_back(VE.getMetadataOrNullID(N->getType()));
+  Record.push_back(N->isLocalToUnit());
+  Record.push_back(N->isDefinition());
+  Record.push_back(/* N->getRawVariable() */ 0);
+  Record.push_back(VE.getMetadataOrNullID(N->getStaticDataMemberDeclaration()));
+
+  Stream.EmitRecord(bitc::METADATA_GLOBAL_VAR, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDILocalVariable(const DILocalVariable *N,
+                                             SmallVectorImpl<uint64_t> &Record,
+                                             unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  Record.push_back(N->getTag());
+  Record.push_back(VE.getMetadataOrNullID(N->getScope()));
+  Record.push_back(VE.getMetadataOrNullID(N->getRawName()));
+  Record.push_back(VE.getMetadataOrNullID(N->getFile()));
+  Record.push_back(N->getLine());
+  Record.push_back(VE.getMetadataOrNullID(N->getType()));
+  Record.push_back(N->getArg());
+  Record.push_back(N->getFlags());
+
+  Stream.EmitRecord(bitc::METADATA_LOCAL_VAR, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDIExpression(const DIExpression *N,
+                                          SmallVectorImpl<uint64_t> &Record,
+                                          unsigned Abbrev) {
+  Record.reserve(N->getElements().size() + 1);
+
+  Record.push_back(N->isDistinct());
+  Record.append(N->elements_begin(), N->elements_end());
+
+  Stream.EmitRecord(bitc::METADATA_EXPRESSION, Record, Abbrev);
+  Record.clear();
+}
+
+void DXILBitcodeWriter::writeDIObjCProperty(const DIObjCProperty *N,
+                                            SmallVectorImpl<uint64_t> &Record,
+                                            unsigned Abbrev) {
+  llvm_unreachable("DXIL does not support objc!!!");
+}
+
+void DXILBitcodeWriter::writeDIImportedEntity(const DIImportedEntity *N,
+                                              SmallVectorImpl<uint64_t> &Record,
+                                              unsigned Abbrev) {
+  Record.push_back(N->isDistinct());
+  Record.push_back(N->getTag());
+  Record.push_back(VE.getMetadataOrNullID(N->getScope()));
+  Record.push_back(VE.getMetadataOrNullID(N->getEntity()));
+  Record.push_back(N->getLine());
+  Record.push_back(VE.getMetadataOrNullID(N->getRawName()));
+
+  Stream.EmitRecord(bitc::METADATA_IMPORTED_ENTITY, Record, Abbrev);
+  Record.clear();
+}
+
+unsigned DXILBitcodeWriter::createDILocationAbbrev() {
+  // Abbrev for METADATA_LOCATION.
+  //
+  // Assume the column is usually under 128, and always output the inlined-at
+  // location (it's never more expensive than building an array size 1).
+  std::shared_ptr<BitCodeAbbrev> Abbv = std::make_shared<BitCodeAbbrev>();
+  Abbv->Add(BitCodeAbbrevOp(bitc::METADATA_LOCATION));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 1));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 6));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 6));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 6));
+  return Stream.EmitAbbrev(std::move(Abbv));
+}
+
+unsigned DXILBitcodeWriter::createGenericDINodeAbbrev() {
+  // Abbrev for METADATA_GENERIC_DEBUG.
+  //
+  // Assume the column is usually under 128, and always output the inlined-at
+  // location (it's never more expensive than building an array size 1).
+  std::shared_ptr<BitCodeAbbrev> Abbv = std::make_shared<BitCodeAbbrev>();
+  Abbv->Add(BitCodeAbbrevOp(bitc::METADATA_GENERIC_DEBUG));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 1));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 6));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 1));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 6));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 6));
+  return Stream.EmitAbbrev(std::move(Abbv));
+}
+
+void DXILBitcodeWriter::writeMetadataRecords(ArrayRef<const Metadata *> MDs,
+                                             SmallVectorImpl<uint64_t> &Record,
+                                             std::vector<unsigned> *MDAbbrevs,
+                                             std::vector<uint64_t> *IndexPos) {
+  if (MDs.empty())
+    return;
+
+    // Initialize MDNode abbreviations.
+#define HANDLE_MDNODE_LEAF(CLASS) unsigned CLASS##Abbrev = 0;
+#include "llvm/IR/Metadata.def"
+
+  for (const Metadata *MD : MDs) {
+    if (IndexPos)
+      IndexPos->push_back(Stream.GetCurrentBitNo());
+    if (const MDNode *N = dyn_cast<MDNode>(MD)) {
+      assert(N->isResolved() && "Expected forward references to be resolved");
+
+      switch (N->getMetadataID()) {
+      default:
+        llvm_unreachable("Invalid MDNode subclass");
+#define HANDLE_MDNODE_LEAF(CLASS)                                              \
+  case Metadata::CLASS##Kind:                                                  \
+    if (MDAbbrevs)                                                             \
+      write##CLASS(cast<CLASS>(N), Record,                                     \
+                   (*MDAbbrevs)[MetadataAbbrev::CLASS##AbbrevID]);             \
+    else                                                                       \
+      write##CLASS(cast<CLASS>(N), Record, CLASS##Abbrev);                     \
+    continue;
+#include "llvm/IR/Metadata.def"
+      }
+    }
+    writeValueAsMetadata(cast<ValueAsMetadata>(MD), Record);
+  }
+}
+
+unsigned DXILBitcodeWriter::createMetadataStringsAbbrev() {
+  auto Abbv = std::make_shared<BitCodeAbbrev>();
+  Abbv->Add(BitCodeAbbrevOp(bitc::METADATA_STRING_OLD));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 8));
+  return Stream.EmitAbbrev(std::move(Abbv));
+}
+
+void DXILBitcodeWriter::writeMetadataStrings(
+    ArrayRef<const Metadata *> Strings, SmallVectorImpl<uint64_t> &Record) {
+  for (const Metadata *MD : Strings) {
+    const MDString *MDS = cast<MDString>(MD);
+    // Code: [strchar x N]
+    Record.append(MDS->bytes_begin(), MDS->bytes_end());
+
+    // Emit the finished record.
+    Stream.EmitRecord(bitc::METADATA_STRING_OLD, Record,
+                      createMetadataStringsAbbrev());
+    Record.clear();
+  }
+}
+
+void DXILBitcodeWriter::writeModuleMetadata() {
+  if (!VE.hasMDs() && M.named_metadata_empty())
+    return;
+
+  Stream.EnterSubblock(bitc::METADATA_BLOCK_ID, 5);
+
+  // Emit all abbrevs upfront, so that the reader can jump in the middle of the
+  // block and load any metadata.
+  std::vector<unsigned> MDAbbrevs;
+
+  MDAbbrevs.resize(MetadataAbbrev::LastPlusOne);
+  MDAbbrevs[MetadataAbbrev::DILocationAbbrevID] = createDILocationAbbrev();
+  MDAbbrevs[MetadataAbbrev::GenericDINodeAbbrevID] =
+      createGenericDINodeAbbrev();
+
+  unsigned NameAbbrev = 0;
+  if (!M.named_metadata_empty()) {
+    // Abbrev for METADATA_NAME.
+    std::shared_ptr<BitCodeAbbrev> Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::METADATA_NAME));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 8));
+    NameAbbrev = Stream.EmitAbbrev(std::move(Abbv));
+  }
+
+  SmallVector<uint64_t, 64> Record;
+  writeMetadataStrings(VE.getMDStrings(), Record);
+
+  std::vector<uint64_t> IndexPos;
+  IndexPos.reserve(VE.getNonMDStrings().size());
+  writeMetadataRecords(VE.getNonMDStrings(), Record, &MDAbbrevs, &IndexPos);
+
+  // Write named metadata.
+  for (const NamedMDNode &NMD : M.named_metadata()) {
+    // Write name.
+    StringRef Str = NMD.getName();
+    Record.append(Str.bytes_begin(), Str.bytes_end());
+    Stream.EmitRecord(bitc::METADATA_NAME, Record, NameAbbrev);
+    Record.clear();
+
+    // Write named metadata operands.
+    for (const MDNode *N : NMD.operands())
+      Record.push_back(VE.getMetadataID(N));
+    Stream.EmitRecord(bitc::METADATA_NAMED_NODE, Record, 0);
+    Record.clear();
+  }
+
+  Stream.ExitBlock();
+}
+
+void DXILBitcodeWriter::writeFunctionMetadata(const Function &F) {
+  if (!VE.hasMDs())
+    return;
+
+  Stream.EnterSubblock(bitc::METADATA_BLOCK_ID, 4);
+  SmallVector<uint64_t, 64> Record;
+  writeMetadataStrings(VE.getMDStrings(), Record);
+  writeMetadataRecords(VE.getNonMDStrings(), Record);
+  Stream.ExitBlock();
+}
+
+void DXILBitcodeWriter::writeFunctionMetadataAttachment(const Function &F) {
+  Stream.EnterSubblock(bitc::METADATA_ATTACHMENT_ID, 3);
+
+  SmallVector<uint64_t, 64> Record;
+
+  // Write metadata attachments
+  // METADATA_ATTACHMENT - [m x [value, [n x [id, mdnode]]]
+  SmallVector<std::pair<unsigned, MDNode *>, 4> MDs;
+  F.getAllMetadata(MDs);
+  if (!MDs.empty()) {
+    for (const auto &I : MDs) {
+      Record.push_back(I.first);
+      Record.push_back(VE.getMetadataID(I.second));
+    }
+    Stream.EmitRecord(bitc::METADATA_ATTACHMENT, Record, 0);
+    Record.clear();
+  }
+
+  for (const BasicBlock &BB : F)
+    for (const Instruction &I : BB) {
+      MDs.clear();
+      I.getAllMetadataOtherThanDebugLoc(MDs);
+
+      // If no metadata, ignore instruction.
+      if (MDs.empty())
+        continue;
+
+      Record.push_back(VE.getInstructionID(&I));
+
+      for (unsigned i = 0, e = MDs.size(); i != e; ++i) {
+        Record.push_back(MDs[i].first);
+        Record.push_back(VE.getMetadataID(MDs[i].second));
+      }
+      Stream.EmitRecord(bitc::METADATA_ATTACHMENT, Record, 0);
+      Record.clear();
+    }
+
+  Stream.ExitBlock();
+}
+
+void DXILBitcodeWriter::writeModuleMetadataKinds() {
+  SmallVector<uint64_t, 64> Record;
+
+  // Write metadata kinds
+  // METADATA_KIND - [n x [id, name]]
+  SmallVector<StringRef, 8> Names;
+  M.getMDKindNames(Names);
+
+  if (Names.empty())
+    return;
+
+  Stream.EnterSubblock(bitc::METADATA_BLOCK_ID, 3);
+
+  for (unsigned MDKindID = 0, e = Names.size(); MDKindID != e; ++MDKindID) {
+    Record.push_back(MDKindID);
+    StringRef KName = Names[MDKindID];
+    Record.append(KName.begin(), KName.end());
+
+    Stream.EmitRecord(bitc::METADATA_KIND, Record, 0);
+    Record.clear();
+  }
+
+  Stream.ExitBlock();
+}
+
+void DXILBitcodeWriter::writeConstants(unsigned FirstVal, unsigned LastVal,
+                                       bool isGlobal) {
+  if (FirstVal == LastVal)
+    return;
+
+  Stream.EnterSubblock(bitc::CONSTANTS_BLOCK_ID, 4);
+
+  unsigned AggregateAbbrev = 0;
+  unsigned String8Abbrev = 0;
+  unsigned CString7Abbrev = 0;
+  unsigned CString6Abbrev = 0;
+  // If this is a constant pool for the module, emit module-specific abbrevs.
+  if (isGlobal) {
+    // Abbrev for CST_CODE_AGGREGATE.
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::CST_CODE_AGGREGATE));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+    Abbv->Add(
+        BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, Log2_32_Ceil(LastVal + 1)));
+    AggregateAbbrev = Stream.EmitAbbrev(std::move(Abbv));
+
+    // Abbrev for CST_CODE_STRING.
+    Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::CST_CODE_STRING));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 8));
+    String8Abbrev = Stream.EmitAbbrev(std::move(Abbv));
+    // Abbrev for CST_CODE_CSTRING.
+    Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::CST_CODE_CSTRING));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 7));
+    CString7Abbrev = Stream.EmitAbbrev(std::move(Abbv));
+    // Abbrev for CST_CODE_CSTRING.
+    Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::CST_CODE_CSTRING));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Char6));
+    CString6Abbrev = Stream.EmitAbbrev(std::move(Abbv));
+  }
+
+  SmallVector<uint64_t, 64> Record;
+
+  const ValueEnumerator::ValueList &Vals = VE.getValues();
+  Type *LastTy = nullptr;
+  for (unsigned i = FirstVal; i != LastVal; ++i) {
+    const Value *V = Vals[i].first;
+    // If we need to switch types, do so now.
+    if (V->getType() != LastTy) {
+      LastTy = V->getType();
+      Record.push_back(VE.getTypeID(LastTy));
+      Stream.EmitRecord(bitc::CST_CODE_SETTYPE, Record,
+                        CONSTANTS_SETTYPE_ABBREV);
+      Record.clear();
+    }
+
+    if (const InlineAsm *IA = dyn_cast<InlineAsm>(V)) {
+      Record.push_back(unsigned(IA->hasSideEffects()) |
+                       unsigned(IA->isAlignStack()) << 1 |
+                       unsigned(IA->getDialect() & 1) << 2);
+
+      // Add the asm string.
+      const std::string &AsmStr = IA->getAsmString();
+      Record.push_back(AsmStr.size());
+      Record.append(AsmStr.begin(), AsmStr.end());
+
+      // Add the constraint string.
+      const std::string &ConstraintStr = IA->getConstraintString();
+      Record.push_back(ConstraintStr.size());
+      Record.append(ConstraintStr.begin(), ConstraintStr.end());
+      Stream.EmitRecord(bitc::CST_CODE_INLINEASM, Record);
+      Record.clear();
+      continue;
+    }
+    const Constant *C = cast<Constant>(V);
+    unsigned Code = -1U;
+    unsigned AbbrevToUse = 0;
+    if (C->isNullValue()) {
+      Code = bitc::CST_CODE_NULL;
+    } else if (isa<UndefValue>(C)) {
+      Code = bitc::CST_CODE_UNDEF;
+    } else if (const ConstantInt *IV = dyn_cast<ConstantInt>(C)) {
+      if (IV->getBitWidth() <= 64) {
+        uint64_t V = IV->getSExtValue();
+        emitSignedInt64(Record, V);
+        Code = bitc::CST_CODE_INTEGER;
+        AbbrevToUse = CONSTANTS_INTEGER_ABBREV;
+      } else { // Wide integers, > 64 bits in size.
+        // We have an arbitrary precision integer value to write whose
+        // bit width is > 64. However, in canonical unsigned integer
+        // format it is likely that the high bits are going to be zero.
+        // So, we only write the number of active words.
+        unsigned NWords = IV->getValue().getActiveWords();
+        const uint64_t *RawWords = IV->getValue().getRawData();
+        for (unsigned i = 0; i != NWords; ++i) {
+          emitSignedInt64(Record, RawWords[i]);
+        }
+        Code = bitc::CST_CODE_WIDE_INTEGER;
+      }
+    } else if (const ConstantFP *CFP = dyn_cast<ConstantFP>(C)) {
+      Code = bitc::CST_CODE_FLOAT;
+      Type *Ty = CFP->getType();
+      if (Ty->isHalfTy() || Ty->isFloatTy() || Ty->isDoubleTy()) {
+        Record.push_back(CFP->getValueAPF().bitcastToAPInt().getZExtValue());
+      } else if (Ty->isX86_FP80Ty()) {
+        // api needed to prevent premature destruction
+        // bits are not in the same order as a normal i80 APInt, compensate.
+        APInt api = CFP->getValueAPF().bitcastToAPInt();
+        const uint64_t *p = api.getRawData();
+        Record.push_back((p[1] << 48) | (p[0] >> 16));
+        Record.push_back(p[0] & 0xffffLL);
+      } else if (Ty->isFP128Ty() || Ty->isPPC_FP128Ty()) {
+        APInt api = CFP->getValueAPF().bitcastToAPInt();
+        const uint64_t *p = api.getRawData();
+        Record.push_back(p[0]);
+        Record.push_back(p[1]);
+      } else {
+        assert(0 && "Unknown FP type!");
+      }
+    } else if (isa<ConstantDataSequential>(C) &&
+               cast<ConstantDataSequential>(C)->isString()) {
+      const ConstantDataSequential *Str = cast<ConstantDataSequential>(C);
+      // Emit constant strings specially.
+      unsigned NumElts = Str->getNumElements();
+      // If this is a null-terminated string, use the denser CSTRING encoding.
+      if (Str->isCString()) {
+        Code = bitc::CST_CODE_CSTRING;
+        --NumElts; // Don't encode the null, which isn't allowed by char6.
+      } else {
+        Code = bitc::CST_CODE_STRING;
+        AbbrevToUse = String8Abbrev;
+      }
+      bool isCStr7 = Code == bitc::CST_CODE_CSTRING;
+      bool isCStrChar6 = Code == bitc::CST_CODE_CSTRING;
+      for (unsigned i = 0; i != NumElts; ++i) {
+        unsigned char V = Str->getElementAsInteger(i);
+        Record.push_back(V);
+        isCStr7 &= (V & 128) == 0;
+        if (isCStrChar6)
+          isCStrChar6 = BitCodeAbbrevOp::isChar6(V);
+      }
+
+      if (isCStrChar6)
+        AbbrevToUse = CString6Abbrev;
+      else if (isCStr7)
+        AbbrevToUse = CString7Abbrev;
+    } else if (const ConstantDataSequential *CDS =
+                   dyn_cast<ConstantDataSequential>(C)) {
+      Code = bitc::CST_CODE_DATA;
+      Type *EltTy = CDS->getType()->getArrayElementType();
+      if (isa<IntegerType>(EltTy)) {
+        for (unsigned i = 0, e = CDS->getNumElements(); i != e; ++i)
+          Record.push_back(CDS->getElementAsInteger(i));
+      } else if (EltTy->isFloatTy()) {
+        for (unsigned i = 0, e = CDS->getNumElements(); i != e; ++i) {
+          union {
+            float F;
+            uint32_t I;
+          };
+          F = CDS->getElementAsFloat(i);
+          Record.push_back(I);
+        }
+      } else {
+        assert(EltTy->isDoubleTy() && "Unknown ConstantData element type");
+        for (unsigned i = 0, e = CDS->getNumElements(); i != e; ++i) {
+          union {
+            double F;
+            uint64_t I;
+          };
+          F = CDS->getElementAsDouble(i);
+          Record.push_back(I);
+        }
+      }
+    } else if (isa<ConstantArray>(C) || isa<ConstantStruct>(C) ||
+               isa<ConstantVector>(C)) {
+      Code = bitc::CST_CODE_AGGREGATE;
+      for (const Value *Op : C->operands())
+        Record.push_back(VE.getValueID(Op));
+      AbbrevToUse = AggregateAbbrev;
+    } else if (const ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) {
+      switch (CE->getOpcode()) {
+      default:
+        if (Instruction::isCast(CE->getOpcode())) {
+          Code = bitc::CST_CODE_CE_CAST;
+          Record.push_back(getEncodedCastOpcode(CE->getOpcode()));
+          Record.push_back(VE.getTypeID(C->getOperand(0)->getType()));
+          Record.push_back(VE.getValueID(C->getOperand(0)));
+          AbbrevToUse = CONSTANTS_CE_CAST_Abbrev;
+        } else {
+          assert(CE->getNumOperands() == 2 && "Unknown constant expr!");
+          Code = bitc::CST_CODE_CE_BINOP;
+          Record.push_back(getEncodedBinaryOpcode(CE->getOpcode()));
+          Record.push_back(VE.getValueID(C->getOperand(0)));
+          Record.push_back(VE.getValueID(C->getOperand(1)));
+          uint64_t Flags = getOptimizationFlags(CE);
+          if (Flags != 0)
+            Record.push_back(Flags);
+        }
+        break;
+      case Instruction::GetElementPtr: {
+        Code = bitc::CST_CODE_CE_GEP;
+        const auto *GO = cast<GEPOperator>(C);
+        if (GO->isInBounds())
+          Code = bitc::CST_CODE_CE_INBOUNDS_GEP;
+        Record.push_back(VE.getTypeID(GO->getSourceElementType()));
+        for (unsigned i = 0, e = CE->getNumOperands(); i != e; ++i) {
+          Record.push_back(VE.getTypeID(C->getOperand(i)->getType()));
+          Record.push_back(VE.getValueID(C->getOperand(i)));
+        }
+        break;
+      }
+      case Instruction::Select:
+        Code = bitc::CST_CODE_CE_SELECT;
+        Record.push_back(VE.getValueID(C->getOperand(0)));
+        Record.push_back(VE.getValueID(C->getOperand(1)));
+        Record.push_back(VE.getValueID(C->getOperand(2)));
+        break;
+      case Instruction::ExtractElement:
+        Code = bitc::CST_CODE_CE_EXTRACTELT;
+        Record.push_back(VE.getTypeID(C->getOperand(0)->getType()));
+        Record.push_back(VE.getValueID(C->getOperand(0)));
+        Record.push_back(VE.getTypeID(C->getOperand(1)->getType()));
+        Record.push_back(VE.getValueID(C->getOperand(1)));
+        break;
+      case Instruction::InsertElement:
+        Code = bitc::CST_CODE_CE_INSERTELT;
+        Record.push_back(VE.getValueID(C->getOperand(0)));
+        Record.push_back(VE.getValueID(C->getOperand(1)));
+        Record.push_back(VE.getTypeID(C->getOperand(2)->getType()));
+        Record.push_back(VE.getValueID(C->getOperand(2)));
+        break;
+      case Instruction::ShuffleVector:
+        // If the return type and argument types are the same, this is a
+        // standard shufflevector instruction.  If the types are different,
+        // then the shuffle is widening or truncating the input vectors, and
+        // the argument type must also be encoded.
+        if (C->getType() == C->getOperand(0)->getType()) {
+          Code = bitc::CST_CODE_CE_SHUFFLEVEC;
+        } else {
+          Code = bitc::CST_CODE_CE_SHUFVEC_EX;
+          Record.push_back(VE.getTypeID(C->getOperand(0)->getType()));
+        }
+        Record.push_back(VE.getValueID(C->getOperand(0)));
+        Record.push_back(VE.getValueID(C->getOperand(1)));
+        Record.push_back(VE.getValueID(C->getOperand(2)));
+        break;
+      case Instruction::ICmp:
+      case Instruction::FCmp:
+        Code = bitc::CST_CODE_CE_CMP;
+        Record.push_back(VE.getTypeID(C->getOperand(0)->getType()));
+        Record.push_back(VE.getValueID(C->getOperand(0)));
+        Record.push_back(VE.getValueID(C->getOperand(1)));
+        Record.push_back(CE->getPredicate());
+        break;
+      }
+    } else if (const BlockAddress *BA = dyn_cast<BlockAddress>(C)) {
+      Code = bitc::CST_CODE_BLOCKADDRESS;
+      Record.push_back(VE.getTypeID(BA->getFunction()->getType()));
+      Record.push_back(VE.getValueID(BA->getFunction()));
+      Record.push_back(VE.getGlobalBasicBlockID(BA->getBasicBlock()));
+    } else {
+#ifndef NDEBUG
+      C->dump();
+#endif
+      llvm_unreachable("Unknown constant!");
+    }
+    Stream.EmitRecord(Code, Record, AbbrevToUse);
+    Record.clear();
+  }
+
+  Stream.ExitBlock();
+}
+
+void DXILBitcodeWriter::writeModuleConstants() {
+  const ValueEnumerator::ValueList &Vals = VE.getValues();
+
+  // Find the first constant to emit, which is the first non-globalvalue value.
+  // We know globalvalues have been emitted by WriteModuleInfo.
+  for (unsigned i = 0, e = Vals.size(); i != e; ++i) {
+    if (!isa<GlobalValue>(Vals[i].first)) {
+      writeConstants(i, Vals.size(), true);
+      return;
+    }
+  }
+}
+
+/// pushValueAndType - The file has to encode both the value and type id for
+/// many values, because we need to know what type to create for forward
+/// references.  However, most operands are not forward references, so this type
+/// field is not needed.
+///
+/// This function adds V's value ID to Vals.  If the value ID is higher than the
+/// instruction ID, then it is a forward reference, and it also includes the
+/// type ID.  The value ID that is written is encoded relative to the InstID.
+bool DXILBitcodeWriter::pushValueAndType(const Value *V, unsigned InstID,
+                                         SmallVectorImpl<unsigned> &Vals) {
+  unsigned ValID = VE.getValueID(V);
+  // Make encoding relative to the InstID.
+  Vals.push_back(InstID - ValID);
+  if (ValID >= InstID) {
+    Vals.push_back(VE.getTypeID(V->getType()));
+    return true;
+  }
+  return false;
+}
+
+/// pushValue - Like pushValueAndType, but where the type of the value is
+/// omitted (perhaps it was already encoded in an earlier operand).
+void DXILBitcodeWriter::pushValue(const Value *V, unsigned InstID,
+                                  SmallVectorImpl<unsigned> &Vals) {
+  unsigned ValID = VE.getValueID(V);
+  Vals.push_back(InstID - ValID);
+}
+
+void DXILBitcodeWriter::pushValueSigned(const Value *V, unsigned InstID,
+                                        SmallVectorImpl<uint64_t> &Vals) {
+  unsigned ValID = VE.getValueID(V);
+  int64_t diff = ((int32_t)InstID - (int32_t)ValID);
+  emitSignedInt64(Vals, diff);
+}
+
+/// WriteInstruction - Emit an instruction
+void DXILBitcodeWriter::writeInstruction(const Instruction &I, unsigned InstID,
+                                         SmallVectorImpl<unsigned> &Vals) {
+  unsigned Code = 0;
+  unsigned AbbrevToUse = 0;
+  VE.setInstructionID(&I);
+  switch (I.getOpcode()) {
+  default:
+    if (Instruction::isCast(I.getOpcode())) {
+      Code = bitc::FUNC_CODE_INST_CAST;
+      if (!pushValueAndType(I.getOperand(0), InstID, Vals))
+        AbbrevToUse = (unsigned)FUNCTION_INST_CAST_ABBREV;
+      Vals.push_back(VE.getTypeID(I.getType()));
+      Vals.push_back(getEncodedCastOpcode(I.getOpcode()));
+    } else {
+      assert(isa<BinaryOperator>(I) && "Unknown instruction!");
+      Code = bitc::FUNC_CODE_INST_BINOP;
+      if (!pushValueAndType(I.getOperand(0), InstID, Vals))
+        AbbrevToUse = (unsigned)FUNCTION_INST_BINOP_ABBREV;
+      pushValue(I.getOperand(1), InstID, Vals);
+      Vals.push_back(getEncodedBinaryOpcode(I.getOpcode()));
+      uint64_t Flags = getOptimizationFlags(&I);
+      if (Flags != 0) {
+        if (AbbrevToUse == (unsigned)FUNCTION_INST_BINOP_ABBREV)
+          AbbrevToUse = (unsigned)FUNCTION_INST_BINOP_FLAGS_ABBREV;
+        Vals.push_back(Flags);
+      }
+    }
+    break;
+
+  case Instruction::GetElementPtr: {
+    Code = bitc::FUNC_CODE_INST_GEP;
+    AbbrevToUse = (unsigned)FUNCTION_INST_GEP_ABBREV;
+    auto &GEPInst = cast<GetElementPtrInst>(I);
+    Vals.push_back(GEPInst.isInBounds());
+    Vals.push_back(VE.getTypeID(GEPInst.getSourceElementType()));
+    for (unsigned i = 0, e = I.getNumOperands(); i != e; ++i)
+      pushValueAndType(I.getOperand(i), InstID, Vals);
+    break;
+  }
+  case Instruction::ExtractValue: {
+    Code = bitc::FUNC_CODE_INST_EXTRACTVAL;
+    pushValueAndType(I.getOperand(0), InstID, Vals);
+    const ExtractValueInst *EVI = cast<ExtractValueInst>(&I);
+    Vals.append(EVI->idx_begin(), EVI->idx_end());
+    break;
+  }
+  case Instruction::InsertValue: {
+    Code = bitc::FUNC_CODE_INST_INSERTVAL;
+    pushValueAndType(I.getOperand(0), InstID, Vals);
+    pushValueAndType(I.getOperand(1), InstID, Vals);
+    const InsertValueInst *IVI = cast<InsertValueInst>(&I);
+    Vals.append(IVI->idx_begin(), IVI->idx_end());
+    break;
+  }
+  case Instruction::Select:
+    Code = bitc::FUNC_CODE_INST_VSELECT;
+    pushValueAndType(I.getOperand(1), InstID, Vals);
+    pushValue(I.getOperand(2), InstID, Vals);
+    pushValueAndType(I.getOperand(0), InstID, Vals);
+    break;
+  case Instruction::ExtractElement:
+    Code = bitc::FUNC_CODE_INST_EXTRACTELT;
+    pushValueAndType(I.getOperand(0), InstID, Vals);
+    pushValueAndType(I.getOperand(1), InstID, Vals);
+    break;
+  case Instruction::InsertElement:
+    Code = bitc::FUNC_CODE_INST_INSERTELT;
+    pushValueAndType(I.getOperand(0), InstID, Vals);
+    pushValue(I.getOperand(1), InstID, Vals);
+    pushValueAndType(I.getOperand(2), InstID, Vals);
+    break;
+  case Instruction::ShuffleVector:
+    Code = bitc::FUNC_CODE_INST_SHUFFLEVEC;
+    pushValueAndType(I.getOperand(0), InstID, Vals);
+    pushValue(I.getOperand(1), InstID, Vals);
+    pushValue(I.getOperand(2), InstID, Vals);
+    break;
+  case Instruction::ICmp:
+  case Instruction::FCmp: {
+    // compare returning Int1Ty or vector of Int1Ty
+    Code = bitc::FUNC_CODE_INST_CMP2;
+    pushValueAndType(I.getOperand(0), InstID, Vals);
+    pushValue(I.getOperand(1), InstID, Vals);
+    Vals.push_back(cast<CmpInst>(I).getPredicate());
+    uint64_t Flags = getOptimizationFlags(&I);
+    if (Flags != 0)
+      Vals.push_back(Flags);
+    break;
+  }
+
+  case Instruction::Ret: {
+    Code = bitc::FUNC_CODE_INST_RET;
+    unsigned NumOperands = I.getNumOperands();
+    if (NumOperands == 0)
+      AbbrevToUse = (unsigned)FUNCTION_INST_RET_VOID_ABBREV;
+    else if (NumOperands == 1) {
+      if (!pushValueAndType(I.getOperand(0), InstID, Vals))
+        AbbrevToUse = (unsigned)FUNCTION_INST_RET_VAL_ABBREV;
+    } else {
+      for (unsigned i = 0, e = NumOperands; i != e; ++i)
+        pushValueAndType(I.getOperand(i), InstID, Vals);
+    }
+  } break;
+  case Instruction::Br: {
+    Code = bitc::FUNC_CODE_INST_BR;
+    const BranchInst &II = cast<BranchInst>(I);
+    Vals.push_back(VE.getValueID(II.getSuccessor(0)));
+    if (II.isConditional()) {
+      Vals.push_back(VE.getValueID(II.getSuccessor(1)));
+      pushValue(II.getCondition(), InstID, Vals);
+    }
+  } break;
+  case Instruction::Switch: {
+    Code = bitc::FUNC_CODE_INST_SWITCH;
+    const SwitchInst &SI = cast<SwitchInst>(I);
+    Vals.push_back(VE.getTypeID(SI.getCondition()->getType()));
+    pushValue(SI.getCondition(), InstID, Vals);
+    Vals.push_back(VE.getValueID(SI.getDefaultDest()));
+    for (auto Case : SI.cases()) {
+      Vals.push_back(VE.getValueID(Case.getCaseValue()));
+      Vals.push_back(VE.getValueID(Case.getCaseSuccessor()));
+    }
+  } break;
+  case Instruction::IndirectBr:
+    Code = bitc::FUNC_CODE_INST_INDIRECTBR;
+    Vals.push_back(VE.getTypeID(I.getOperand(0)->getType()));
+    // Encode the address operand as relative, but not the basic blocks.
+    pushValue(I.getOperand(0), InstID, Vals);
+    for (unsigned i = 1, e = I.getNumOperands(); i != e; ++i)
+      Vals.push_back(VE.getValueID(I.getOperand(i)));
+    break;
+
+  case Instruction::Invoke: {
+    const InvokeInst *II = cast<InvokeInst>(&I);
+    const Value *Callee = II->getCalledOperand();
+    FunctionType *FTy = II->getFunctionType();
+    Code = bitc::FUNC_CODE_INST_INVOKE;
+
+    Vals.push_back(VE.getAttributeListID(II->getAttributes()));
+    Vals.push_back(II->getCallingConv() | 1 << 13);
+    Vals.push_back(VE.getValueID(II->getNormalDest()));
+    Vals.push_back(VE.getValueID(II->getUnwindDest()));
+    Vals.push_back(VE.getTypeID(FTy));
+    pushValueAndType(Callee, InstID, Vals);
+
+    // Emit value #'s for the fixed parameters.
+    for (unsigned i = 0, e = FTy->getNumParams(); i != e; ++i)
+      pushValue(I.getOperand(i), InstID, Vals); // fixed param.
+
+    // Emit type/value pairs for varargs params.
+    if (FTy->isVarArg()) {
+      for (unsigned i = FTy->getNumParams(), e = I.getNumOperands() - 3; i != e;
+           ++i)
+        pushValueAndType(I.getOperand(i), InstID, Vals); // vararg
+    }
+    break;
+  }
+  case Instruction::Resume:
+    Code = bitc::FUNC_CODE_INST_RESUME;
+    pushValueAndType(I.getOperand(0), InstID, Vals);
+    break;
+  case Instruction::Unreachable:
+    Code = bitc::FUNC_CODE_INST_UNREACHABLE;
+    AbbrevToUse = (unsigned)FUNCTION_INST_UNREACHABLE_ABBREV;
+    break;
+
+  case Instruction::PHI: {
+    const PHINode &PN = cast<PHINode>(I);
+    Code = bitc::FUNC_CODE_INST_PHI;
+    // With the newer instruction encoding, forward references could give
+    // negative valued IDs.  This is most common for PHIs, so we use
+    // signed VBRs.
+    SmallVector<uint64_t, 128> Vals64;
+    Vals64.push_back(VE.getTypeID(PN.getType()));
+    for (unsigned i = 0, e = PN.getNumIncomingValues(); i != e; ++i) {
+      pushValueSigned(PN.getIncomingValue(i), InstID, Vals64);
+      Vals64.push_back(VE.getValueID(PN.getIncomingBlock(i)));
+    }
+    // Emit a Vals64 vector and exit.
+    Stream.EmitRecord(Code, Vals64, AbbrevToUse);
+    Vals64.clear();
+    return;
+  }
+
+  case Instruction::LandingPad: {
+    const LandingPadInst &LP = cast<LandingPadInst>(I);
+    Code = bitc::FUNC_CODE_INST_LANDINGPAD;
+    Vals.push_back(VE.getTypeID(LP.getType()));
+    Vals.push_back(LP.isCleanup());
+    Vals.push_back(LP.getNumClauses());
+    for (unsigned I = 0, E = LP.getNumClauses(); I != E; ++I) {
+      if (LP.isCatch(I))
+        Vals.push_back(LandingPadInst::Catch);
+      else
+        Vals.push_back(LandingPadInst::Filter);
+      pushValueAndType(LP.getClause(I), InstID, Vals);
+    }
+    break;
+  }
+
+  case Instruction::Alloca: {
+    Code = bitc::FUNC_CODE_INST_ALLOCA;
+    const AllocaInst &AI = cast<AllocaInst>(I);
+    Vals.push_back(VE.getTypeID(AI.getAllocatedType()));
+    Vals.push_back(VE.getTypeID(I.getOperand(0)->getType()));
+    Vals.push_back(VE.getValueID(I.getOperand(0))); // size.
+    using APV = AllocaPackedValues;
+    unsigned Record = 0;
+    unsigned EncodedAlign = getEncodedAlign(AI.getAlign());
+    Bitfield::set<APV::AlignLower>(
+        Record, EncodedAlign & ((1 << APV::AlignLower::Bits) - 1));
+    Bitfield::set<APV::AlignUpper>(Record,
+                                   EncodedAlign >> APV::AlignLower::Bits);
+    Bitfield::set<APV::UsedWithInAlloca>(Record, AI.isUsedWithInAlloca());
+    Vals.push_back(Record);
+    break;
+  }
+
+  case Instruction::Load:
+    if (cast<LoadInst>(I).isAtomic()) {
+      Code = bitc::FUNC_CODE_INST_LOADATOMIC;
+      pushValueAndType(I.getOperand(0), InstID, Vals);
+    } else {
+      Code = bitc::FUNC_CODE_INST_LOAD;
+      if (!pushValueAndType(I.getOperand(0), InstID, Vals)) // ptr
+        AbbrevToUse = (unsigned)FUNCTION_INST_LOAD_ABBREV;
+    }
+    Vals.push_back(VE.getTypeID(I.getType()));
+    Vals.push_back(Log2_32(cast<LoadInst>(I).getAlignment()) + 1);
+    Vals.push_back(cast<LoadInst>(I).isVolatile());
+    if (cast<LoadInst>(I).isAtomic()) {
+      Vals.push_back(getEncodedOrdering(cast<LoadInst>(I).getOrdering()));
+      Vals.push_back(getEncodedSyncScopeID(cast<LoadInst>(I).getSyncScopeID()));
+    }
+    break;
+  case Instruction::Store:
+    if (cast<StoreInst>(I).isAtomic())
+      Code = bitc::FUNC_CODE_INST_STOREATOMIC;
+    else
+      Code = bitc::FUNC_CODE_INST_STORE;
+    pushValueAndType(I.getOperand(1), InstID, Vals); // ptrty + ptr
+    pushValueAndType(I.getOperand(0), InstID, Vals); // valty + val
+    Vals.push_back(Log2_32(cast<StoreInst>(I).getAlignment()) + 1);
+    Vals.push_back(cast<StoreInst>(I).isVolatile());
+    if (cast<StoreInst>(I).isAtomic()) {
+      Vals.push_back(getEncodedOrdering(cast<StoreInst>(I).getOrdering()));
+      Vals.push_back(
+          getEncodedSyncScopeID(cast<StoreInst>(I).getSyncScopeID()));
+    }
+    break;
+  case Instruction::AtomicCmpXchg:
+    Code = bitc::FUNC_CODE_INST_CMPXCHG;
+    pushValueAndType(I.getOperand(0), InstID, Vals); // ptrty + ptr
+    pushValueAndType(I.getOperand(1), InstID, Vals); // cmp.
+    pushValue(I.getOperand(2), InstID, Vals);        // newval.
+    Vals.push_back(cast<AtomicCmpXchgInst>(I).isVolatile());
+    Vals.push_back(
+        getEncodedOrdering(cast<AtomicCmpXchgInst>(I).getSuccessOrdering()));
+    Vals.push_back(
+        getEncodedSyncScopeID(cast<AtomicCmpXchgInst>(I).getSyncScopeID()));
+    Vals.push_back(
+        getEncodedOrdering(cast<AtomicCmpXchgInst>(I).getFailureOrdering()));
+    Vals.push_back(cast<AtomicCmpXchgInst>(I).isWeak());
+    break;
+  case Instruction::AtomicRMW:
+    Code = bitc::FUNC_CODE_INST_ATOMICRMW;
+    pushValueAndType(I.getOperand(0), InstID, Vals); // ptrty + ptr
+    pushValue(I.getOperand(1), InstID, Vals);        // val.
+    Vals.push_back(
+        getEncodedRMWOperation(cast<AtomicRMWInst>(I).getOperation()));
+    Vals.push_back(cast<AtomicRMWInst>(I).isVolatile());
+    Vals.push_back(getEncodedOrdering(cast<AtomicRMWInst>(I).getOrdering()));
+    Vals.push_back(
+        getEncodedSyncScopeID(cast<AtomicRMWInst>(I).getSyncScopeID()));
+    break;
+  case Instruction::Fence:
+    Code = bitc::FUNC_CODE_INST_FENCE;
+    Vals.push_back(getEncodedOrdering(cast<FenceInst>(I).getOrdering()));
+    Vals.push_back(getEncodedSyncScopeID(cast<FenceInst>(I).getSyncScopeID()));
+    break;
+  case Instruction::Call: {
+    const CallInst &CI = cast<CallInst>(I);
+    FunctionType *FTy = CI.getFunctionType();
+
+    Code = bitc::FUNC_CODE_INST_CALL;
+
+    Vals.push_back(VE.getAttributeListID(CI.getAttributes()));
+    Vals.push_back((CI.getCallingConv() << 1) | unsigned(CI.isTailCall()) |
+                   unsigned(CI.isMustTailCall()) << 14 | 1 << 15);
+    Vals.push_back(VE.getTypeID(FTy));
+    pushValueAndType(CI.getCalledOperand(), InstID, Vals); // Callee
+
+    // Emit value #'s for the fixed parameters.
+    for (unsigned i = 0, e = FTy->getNumParams(); i != e; ++i) {
+      // Check for labels (can happen with asm labels).
+      if (FTy->getParamType(i)->isLabelTy())
+        Vals.push_back(VE.getValueID(CI.getArgOperand(i)));
+      else
+        pushValue(CI.getArgOperand(i), InstID, Vals); // fixed param.
+    }
+
+    // Emit type/value pairs for varargs params.
+    if (FTy->isVarArg()) {
+      for (unsigned i = FTy->getNumParams(), e = CI.arg_size(); i != e; ++i)
+        pushValueAndType(CI.getArgOperand(i), InstID, Vals); // varargs
+    }
+    break;
+  }
+  case Instruction::VAArg:
+    Code = bitc::FUNC_CODE_INST_VAARG;
+    Vals.push_back(VE.getTypeID(I.getOperand(0)->getType())); // valistty
+    pushValue(I.getOperand(0), InstID, Vals);                 // valist.
+    Vals.push_back(VE.getTypeID(I.getType()));                // restype.
+    break;
+  }
+
+  Stream.EmitRecord(Code, Vals, AbbrevToUse);
+  Vals.clear();
+}
+
+// Emit names for globals/functions etc.
+void DXILBitcodeWriter::writeFunctionLevelValueSymbolTable(
+    const ValueSymbolTable &VST) {
+  if (VST.empty())
+    return;
+  Stream.EnterSubblock(bitc::VALUE_SYMTAB_BLOCK_ID, 4);
+
+  SmallVector<unsigned, 64> NameVals;
+
+  // HLSL Change
+  // Read the named values from a sorted list instead of the original list
+  // to ensure the binary is the same no matter what values ever existed.
+  SmallVector<const ValueName *, 16> SortedTable;
+
+  for (auto &VI : VST) {
+    SortedTable.push_back(VI.second->getValueName());
+  }
+  // The keys are unique, so there shouldn't be stability issues.
+  std::sort(SortedTable.begin(), SortedTable.end(),
+            [](const ValueName *A, const ValueName *B) {
+              return A->first() < B->first();
+            });
+
+  for (const ValueName *SI : SortedTable) {
+    auto &Name = *SI;
+
+    // Figure out the encoding to use for the name.
+    bool is7Bit = true;
+    bool isChar6 = true;
+    for (const char *C = Name.getKeyData(), *E = C + Name.getKeyLength();
+         C != E; ++C) {
+      if (isChar6)
+        isChar6 = BitCodeAbbrevOp::isChar6(*C);
+      if ((unsigned char)*C & 128) {
+        is7Bit = false;
+        break; // don't bother scanning the rest.
+      }
+    }
+
+    unsigned AbbrevToUse = VST_ENTRY_8_ABBREV;
+
+    // VST_ENTRY:   [valueid, namechar x N]
+    // VST_BBENTRY: [bbid, namechar x N]
+    unsigned Code;
+    if (isa<BasicBlock>(SI->getValue())) {
+      Code = bitc::VST_CODE_BBENTRY;
+      if (isChar6)
+        AbbrevToUse = VST_BBENTRY_6_ABBREV;
+    } else {
+      Code = bitc::VST_CODE_ENTRY;
+      if (isChar6)
+        AbbrevToUse = VST_ENTRY_6_ABBREV;
+      else if (is7Bit)
+        AbbrevToUse = VST_ENTRY_7_ABBREV;
+    }
+
+    NameVals.push_back(VE.getValueID(SI->getValue()));
+    for (const char *P = Name.getKeyData(),
+                    *E = Name.getKeyData() + Name.getKeyLength();
+         P != E; ++P)
+      NameVals.push_back((unsigned char)*P);
+
+    // Emit the finished record.
+    Stream.EmitRecord(Code, NameVals, AbbrevToUse);
+    NameVals.clear();
+  }
+  Stream.ExitBlock();
+}
+
+void DXILBitcodeWriter::writeUseList(UseListOrder &&Order) {
+  assert(Order.Shuffle.size() >= 2 && "Shuffle too small");
+  unsigned Code;
+  if (isa<BasicBlock>(Order.V))
+    Code = bitc::USELIST_CODE_BB;
+  else
+    Code = bitc::USELIST_CODE_DEFAULT;
+
+  SmallVector<uint64_t, 64> Record(Order.Shuffle.begin(), Order.Shuffle.end());
+  Record.push_back(VE.getValueID(Order.V));
+  Stream.EmitRecord(Code, Record);
+}
+
+void DXILBitcodeWriter::writeUseListBlock(const Function *F) {
+  assert(VE.shouldPreserveUseListOrder() &&
+         "Expected to be preserving use-list order");
+
+  auto hasMore = [&]() {
+    return !VE.UseListOrders.empty() && VE.UseListOrders.back().F == F;
+  };
+  if (!hasMore())
+    // Nothing to do.
+    return;
+
+  Stream.EnterSubblock(bitc::USELIST_BLOCK_ID, 3);
+  while (hasMore()) {
+    writeUseList(std::move(VE.UseListOrders.back()));
+    VE.UseListOrders.pop_back();
+  }
+  Stream.ExitBlock();
+}
+
+/// Emit a function body to the module stream.
+void DXILBitcodeWriter::writeFunction(const Function &F) {
+  Stream.EnterSubblock(bitc::FUNCTION_BLOCK_ID, 4);
+  VE.incorporateFunction(F);
+
+  SmallVector<unsigned, 64> Vals;
+
+  // Emit the number of basic blocks, so the reader can create them ahead of
+  // time.
+  Vals.push_back(VE.getBasicBlocks().size());
+  Stream.EmitRecord(bitc::FUNC_CODE_DECLAREBLOCKS, Vals);
+  Vals.clear();
+
+  // If there are function-local constants, emit them now.
+  unsigned CstStart, CstEnd;
+  VE.getFunctionConstantRange(CstStart, CstEnd);
+  writeConstants(CstStart, CstEnd, false);
+
+  // If there is function-local metadata, emit it now.
+  writeFunctionMetadata(F);
+
+  // Keep a running idea of what the instruction ID is.
+  unsigned InstID = CstEnd;
+
+  bool NeedsMetadataAttachment = F.hasMetadata();
+
+  DILocation *LastDL = nullptr;
+
+  // Finally, emit all the instructions, in order.
+  for (Function::const_iterator BB = F.begin(), E = F.end(); BB != E; ++BB)
+    for (BasicBlock::const_iterator I = BB->begin(), E = BB->end(); I != E;
+         ++I) {
+      writeInstruction(*I, InstID, Vals);
+
+      if (!I->getType()->isVoidTy())
+        ++InstID;
+
+      // If the instruction has metadata, write a metadata attachment later.
+      NeedsMetadataAttachment |= I->hasMetadataOtherThanDebugLoc();
+
+      // If the instruction has a debug location, emit it.
+      DILocation *DL = I->getDebugLoc();
+      if (!DL)
+        continue;
+
+      if (DL == LastDL) {
+        // Just repeat the same debug loc as last time.
+        Stream.EmitRecord(bitc::FUNC_CODE_DEBUG_LOC_AGAIN, Vals);
+        continue;
+      }
+
+      Vals.push_back(DL->getLine());
+      Vals.push_back(DL->getColumn());
+      Vals.push_back(VE.getMetadataOrNullID(DL->getScope()));
+      Vals.push_back(VE.getMetadataOrNullID(DL->getInlinedAt()));
+      Stream.EmitRecord(bitc::FUNC_CODE_DEBUG_LOC, Vals);
+      Vals.clear();
+
+      LastDL = DL;
+    }
+
+  // Emit names for all the instructions etc.
+  if (auto *Symtab = F.getValueSymbolTable())
+    writeFunctionLevelValueSymbolTable(*Symtab);
+
+  if (NeedsMetadataAttachment)
+    writeFunctionMetadataAttachment(F);
+  if (VE.shouldPreserveUseListOrder())
+    writeUseListBlock(&F);
+  VE.purgeFunction();
+  Stream.ExitBlock();
+}
+
+// Emit blockinfo, which defines the standard abbreviations etc.
+void DXILBitcodeWriter::writeBlockInfo() {
+  // We only want to emit block info records for blocks that have multiple
+  // instances: CONSTANTS_BLOCK, FUNCTION_BLOCK and VALUE_SYMTAB_BLOCK.
+  // Other blocks can define their abbrevs inline.
+  Stream.EnterBlockInfoBlock();
+
+  { // 8-bit fixed-width VST_ENTRY/VST_BBENTRY strings.
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 3));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 8));
+    if (Stream.EmitBlockInfoAbbrev(bitc::VALUE_SYMTAB_BLOCK_ID,
+                                   std::move(Abbv)) != VST_ENTRY_8_ABBREV)
+      assert(false && "Unexpected abbrev ordering!");
+  }
+
+  { // 7-bit fixed width VST_ENTRY strings.
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::VST_CODE_ENTRY));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 7));
+    if (Stream.EmitBlockInfoAbbrev(bitc::VALUE_SYMTAB_BLOCK_ID,
+                                   std::move(Abbv)) != VST_ENTRY_7_ABBREV)
+      assert(false && "Unexpected abbrev ordering!");
+  }
+  { // 6-bit char6 VST_ENTRY strings.
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::VST_CODE_ENTRY));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Char6));
+    if (Stream.EmitBlockInfoAbbrev(bitc::VALUE_SYMTAB_BLOCK_ID,
+                                   std::move(Abbv)) != VST_ENTRY_6_ABBREV)
+      assert(false && "Unexpected abbrev ordering!");
+  }
+  { // 6-bit char6 VST_BBENTRY strings.
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::VST_CODE_BBENTRY));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Char6));
+    if (Stream.EmitBlockInfoAbbrev(bitc::VALUE_SYMTAB_BLOCK_ID,
+                                   std::move(Abbv)) != VST_BBENTRY_6_ABBREV)
+      assert(false && "Unexpected abbrev ordering!");
+  }
+
+  { // SETTYPE abbrev for CONSTANTS_BLOCK.
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::CST_CODE_SETTYPE));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed,
+                              VE.computeBitsRequiredForTypeIndicies()));
+    if (Stream.EmitBlockInfoAbbrev(bitc::CONSTANTS_BLOCK_ID, std::move(Abbv)) !=
+        CONSTANTS_SETTYPE_ABBREV)
+      assert(false && "Unexpected abbrev ordering!");
+  }
+
+  { // INTEGER abbrev for CONSTANTS_BLOCK.
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::CST_CODE_INTEGER));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8));
+    if (Stream.EmitBlockInfoAbbrev(bitc::CONSTANTS_BLOCK_ID, std::move(Abbv)) !=
+        CONSTANTS_INTEGER_ABBREV)
+      assert(false && "Unexpected abbrev ordering!");
+  }
+
+  { // CE_CAST abbrev for CONSTANTS_BLOCK.
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::CST_CODE_CE_CAST));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 4)); // cast opc
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed,      // typeid
+                              VE.computeBitsRequiredForTypeIndicies()));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8)); // value id
+
+    if (Stream.EmitBlockInfoAbbrev(bitc::CONSTANTS_BLOCK_ID, std::move(Abbv)) !=
+        CONSTANTS_CE_CAST_Abbrev)
+      assert(false && "Unexpected abbrev ordering!");
+  }
+  { // NULL abbrev for CONSTANTS_BLOCK.
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::CST_CODE_NULL));
+    if (Stream.EmitBlockInfoAbbrev(bitc::CONSTANTS_BLOCK_ID, std::move(Abbv)) !=
+        CONSTANTS_NULL_Abbrev)
+      assert(false && "Unexpected abbrev ordering!");
+  }
+
+  // FIXME: This should only use space for first class types!
+
+  { // INST_LOAD abbrev for FUNCTION_BLOCK.
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::FUNC_CODE_INST_LOAD));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 6)); // Ptr
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed,    // dest ty
+                              VE.computeBitsRequiredForTypeIndicies()));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 4));   // Align
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 1)); // volatile
+    if (Stream.EmitBlockInfoAbbrev(bitc::FUNCTION_BLOCK_ID, std::move(Abbv)) !=
+        (unsigned)FUNCTION_INST_LOAD_ABBREV)
+      assert(false && "Unexpected abbrev ordering!");
+  }
+  { // INST_BINOP abbrev for FUNCTION_BLOCK.
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::FUNC_CODE_INST_BINOP));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 6));   // LHS
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 6));   // RHS
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 4)); // opc
+    if (Stream.EmitBlockInfoAbbrev(bitc::FUNCTION_BLOCK_ID, std::move(Abbv)) !=
+        (unsigned)FUNCTION_INST_BINOP_ABBREV)
+      assert(false && "Unexpected abbrev ordering!");
+  }
+  { // INST_BINOP_FLAGS abbrev for FUNCTION_BLOCK.
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::FUNC_CODE_INST_BINOP));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 6));   // LHS
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 6));   // RHS
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 4)); // opc
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 7)); // flags
+    if (Stream.EmitBlockInfoAbbrev(bitc::FUNCTION_BLOCK_ID, std::move(Abbv)) !=
+        (unsigned)FUNCTION_INST_BINOP_FLAGS_ABBREV)
+      assert(false && "Unexpected abbrev ordering!");
+  }
+  { // INST_CAST abbrev for FUNCTION_BLOCK.
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::FUNC_CODE_INST_CAST));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 6)); // OpVal
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed,    // dest ty
+                              VE.computeBitsRequiredForTypeIndicies()));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 4)); // opc
+    if (Stream.EmitBlockInfoAbbrev(bitc::FUNCTION_BLOCK_ID, std::move(Abbv)) !=
+        (unsigned)FUNCTION_INST_CAST_ABBREV)
+      assert(false && "Unexpected abbrev ordering!");
+  }
+
+  { // INST_RET abbrev for FUNCTION_BLOCK.
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::FUNC_CODE_INST_RET));
+    if (Stream.EmitBlockInfoAbbrev(bitc::FUNCTION_BLOCK_ID, std::move(Abbv)) !=
+        (unsigned)FUNCTION_INST_RET_VOID_ABBREV)
+      assert(false && "Unexpected abbrev ordering!");
+  }
+  { // INST_RET abbrev for FUNCTION_BLOCK.
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::FUNC_CODE_INST_RET));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 6)); // ValID
+    if (Stream.EmitBlockInfoAbbrev(bitc::FUNCTION_BLOCK_ID, std::move(Abbv)) !=
+        (unsigned)FUNCTION_INST_RET_VAL_ABBREV)
+      assert(false && "Unexpected abbrev ordering!");
+  }
+  { // INST_UNREACHABLE abbrev for FUNCTION_BLOCK.
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::FUNC_CODE_INST_UNREACHABLE));
+    if (Stream.EmitBlockInfoAbbrev(bitc::FUNCTION_BLOCK_ID, std::move(Abbv)) !=
+        (unsigned)FUNCTION_INST_UNREACHABLE_ABBREV)
+      assert(false && "Unexpected abbrev ordering!");
+  }
+  {
+    auto Abbv = std::make_shared<BitCodeAbbrev>();
+    Abbv->Add(BitCodeAbbrevOp(bitc::FUNC_CODE_INST_GEP));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 1));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, // dest ty
+                              Log2_32_Ceil(VE.getTypes().size() + 1)));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+    Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 6));
+    if (Stream.EmitBlockInfoAbbrev(bitc::FUNCTION_BLOCK_ID, std::move(Abbv)) !=
+        (unsigned)FUNCTION_INST_GEP_ABBREV)
+      assert(false && "Unexpected abbrev ordering!");
+  }
+
+  Stream.ExitBlock();
+}
+
+void DXILBitcodeWriter::writeModuleVersion() {
+  // VERSION: [version#]
+  Stream.EmitRecord(bitc::MODULE_CODE_VERSION, ArrayRef<unsigned>{1});
+}
+
+/// WriteModule - Emit the specified module to the bitstream.
+void DXILBitcodeWriter::write() {
+  // The identification block is new since llvm-3.7, but the old bitcode reader
+  // will skip it.
+  // writeIdentificationBlock(Stream);
+
+  Stream.EnterSubblock(bitc::MODULE_BLOCK_ID, 3);
+
+  // It is redundant to fully-specify this here, but nice to make it explicit
+  // so that it is clear the DXIL module version is different.
+  DXILBitcodeWriter::writeModuleVersion();
+
+  // Emit blockinfo, which defines the standard abbreviations etc.
+  writeBlockInfo();
+
+  // Emit information about attribute groups.
+  writeAttributeGroupTable();
+
+  // Emit information about parameter attributes.
+  writeAttributeTable();
+
+  // Emit information describing all of the types in the module.
+  writeTypeTable();
+
+  writeComdats();
+
+  // Emit top-level description of module, including target triple, inline asm,
+  // descriptors for global variables, and function prototype info.
+  writeModuleInfo();
+
+  // Emit constants.
+  writeModuleConstants();
+
+  // Emit metadata.
+  writeModuleMetadataKinds();
+
+  // Emit metadata.
+  writeModuleMetadata();
+
+  // Emit names for globals/functions etc.
+  // DXIL uses the same format for module-level value symbol table as for the
+  // function level table.
+  writeFunctionLevelValueSymbolTable(M.getValueSymbolTable());
+
+  // Emit module-level use-lists.
+  if (VE.shouldPreserveUseListOrder())
+    writeUseListBlock(nullptr);
+
+  // Emit function bodies.
+  for (const Function &F : M)
+    if (!F.isDeclaration())
+      writeFunction(F);
+
+  Stream.ExitBlock();
+}

llvm/lib/Target/DirectX/DXILWriter/DXILValueEnumerator.h

+//===- DirectX/DXILWriter/ValueEnumerator.h - Number values -----*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This class gives values and types Unique ID's.
+// Forked from lib/Bitcode/Writer
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_DXILWRITER_VALUEENUMERATOR_H
+#define LLVM_DXILWRITER_VALUEENUMERATOR_H
+
+#include "llvm/ADT/ArrayRef.h"
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/UniqueVector.h"
+#include "llvm/IR/Attributes.h"
+#include "llvm/IR/UseListOrder.h"
+#include <cassert>
+#include <cstdint>
+#include <utility>
+#include <vector>
+
+namespace llvm {
+
+class BasicBlock;
+class Comdat;
+class DIArgList;
+class Function;
+class Instruction;
+class LocalAsMetadata;
+class MDNode;
+class Metadata;
+class Module;
+class NamedMDNode;
+class raw_ostream;
+class Type;
+class Value;
+class ValueSymbolTable;
+
+namespace dxil {
+
+class ValueEnumerator {
+public:
+  using TypeList = std::vector<Type *>;
+
+  // For each value, we remember its Value* and occurrence frequency.
+  using ValueList = std::vector<std::pair<const Value *, unsigned>>;
+
+  /// Attribute groups as encoded in bitcode are almost AttributeSets, but they
+  /// include the AttributeList index, so we have to track that in our map.
+  using IndexAndAttrSet = std::pair<unsigned, AttributeSet>;
+
+  UseListOrderStack UseListOrders;
+
+private:
+  using TypeMapType = DenseMap<Type *, unsigned>;
+  TypeMapType TypeMap;
+  TypeList Types;
+
+  using ValueMapType = DenseMap<const Value *, unsigned>;
+  ValueMapType ValueMap;
+  ValueList Values;
+
+  using ComdatSetType = UniqueVector<const Comdat *>;
+  ComdatSetType Comdats;
+
+  std::vector<const Metadata *> MDs;
+  std::vector<const Metadata *> FunctionMDs;
+
+  /// Index of information about a piece of metadata.
+  struct MDIndex {
+    unsigned F = 0;  ///< The ID of the function for this metadata, if any.
+    unsigned ID = 0; ///< The implicit ID of this metadata in bitcode.
+
+    MDIndex() = default;
+    explicit MDIndex(unsigned F) : F(F) {}
+
+    /// Check if this has a function tag, and it's different from NewF.
+    bool hasDifferentFunction(unsigned NewF) const { return F && F != NewF; }
+
+    /// Fetch the MD this references out of the given metadata array.
+    const Metadata *get(ArrayRef<const Metadata *> MDs) const {
+      assert(ID && "Expected non-zero ID");
+      assert(ID <= MDs.size() && "Expected valid ID");
+      return MDs[ID - 1];
+    }
+  };
+
+  using MetadataMapType = DenseMap<const Metadata *, MDIndex>;
+  MetadataMapType MetadataMap;
+
+  /// Range of metadata IDs, as a half-open range.
+  struct MDRange {
+    unsigned First = 0;
+    unsigned Last = 0;
+
+    /// Number of strings in the prefix of the metadata range.
+    unsigned NumStrings = 0;
+
+    MDRange() = default;
+    explicit MDRange(unsigned First) : First(First) {}
+  };
+  SmallDenseMap<unsigned, MDRange, 1> FunctionMDInfo;
+
+  bool ShouldPreserveUseListOrder;
+
+  using AttributeGroupMapType = DenseMap<IndexAndAttrSet, unsigned>;
+  AttributeGroupMapType AttributeGroupMap;
+  std::vector<IndexAndAttrSet> AttributeGroups;
+
+  using AttributeListMapType = DenseMap<AttributeList, unsigned>;
+  AttributeListMapType AttributeListMap;
+  std::vector<AttributeList> AttributeLists;
+
+  /// GlobalBasicBlockIDs - This map memoizes the basic block ID's referenced by
+  /// the "getGlobalBasicBlockID" method.
+  mutable DenseMap<const BasicBlock *, unsigned> GlobalBasicBlockIDs;
+
+  using InstructionMapType = DenseMap<const Instruction *, unsigned>;
+  InstructionMapType InstructionMap;
+  unsigned InstructionCount;
+
+  /// BasicBlocks - This contains all the basic blocks for the currently
+  /// incorporated function.  Their reverse mapping is stored in ValueMap.
+  std::vector<const BasicBlock *> BasicBlocks;
+
+  /// When a function is incorporated, this is the size of the Values list
+  /// before incorporation.
+  unsigned NumModuleValues;
+
+  /// When a function is incorporated, this is the size of the Metadatas list
+  /// before incorporation.
+  unsigned NumModuleMDs = 0;
+  unsigned NumMDStrings = 0;
+
+  unsigned FirstFuncConstantID;
+  unsigned FirstInstID;
+
+public:
+  ValueEnumerator(const Module &M, bool ShouldPreserveUseListOrder);
+  ValueEnumerator(const ValueEnumerator &) = delete;
+  ValueEnumerator &operator=(const ValueEnumerator &) = delete;
+
+  void dump() const;
+  void print(raw_ostream &OS, const ValueMapType &Map, const char *Name) const;
+  void print(raw_ostream &OS, const MetadataMapType &Map,
+             const char *Name) const;
+
+  unsigned getValueID(const Value *V) const;
+
+  unsigned getMetadataID(const Metadata *MD) const {
+    auto ID = getMetadataOrNullID(MD);
+    assert(ID != 0 && "Metadata not in slotcalculator!");
+    return ID - 1;
+  }
+
+  unsigned getMetadataOrNullID(const Metadata *MD) const {
+    return MetadataMap.lookup(MD).ID;
+  }
+
+  unsigned numMDs() const { return MDs.size(); }
+
+  bool shouldPreserveUseListOrder() const { return ShouldPreserveUseListOrder; }
+
+  unsigned getTypeID(Type *T) const {
+    TypeMapType::const_iterator I = TypeMap.find(T);
+    assert(I != TypeMap.end() && "Type not in ValueEnumerator!");
+    return I->second - 1;
+  }
+
+  unsigned getInstructionID(const Instruction *I) const;
+  void setInstructionID(const Instruction *I);
+
+  unsigned getAttributeListID(AttributeList PAL) const {
+    if (PAL.isEmpty())
+      return 0; // Null maps to zero.
+    AttributeListMapType::const_iterator I = AttributeListMap.find(PAL);
+    assert(I != AttributeListMap.end() && "Attribute not in ValueEnumerator!");
+    return I->second;
+  }
+
+  unsigned getAttributeGroupID(IndexAndAttrSet Group) const {
+    if (!Group.second.hasAttributes())
+      return 0; // Null maps to zero.
+    AttributeGroupMapType::const_iterator I = AttributeGroupMap.find(Group);
+    assert(I != AttributeGroupMap.end() && "Attribute not in ValueEnumerator!");
+    return I->second;
+  }
+
+  /// getFunctionConstantRange - Return the range of values that corresponds to
+  /// function-local constants.
+  void getFunctionConstantRange(unsigned &Start, unsigned &End) const {
+    Start = FirstFuncConstantID;
+    End = FirstInstID;
+  }
+
+  const ValueList &getValues() const { return Values; }
+
+  /// Check whether the current block has any metadata to emit.
+  bool hasMDs() const { return NumModuleMDs < MDs.size(); }
+
+  /// Get the MDString metadata for this block.
+  ArrayRef<const Metadata *> getMDStrings() const {
+    return makeArrayRef(MDs).slice(NumModuleMDs, NumMDStrings);
+  }
+
+  /// Get the non-MDString metadata for this block.
+  ArrayRef<const Metadata *> getNonMDStrings() const {
+    return makeArrayRef(MDs).slice(NumModuleMDs).slice(NumMDStrings);
+  }
+
+  const TypeList &getTypes() const { return Types; }
+
+  const std::vector<const BasicBlock *> &getBasicBlocks() const {
+    return BasicBlocks;
+  }
+
+  const std::vector<AttributeList> &getAttributeLists() const {
+    return AttributeLists;
+  }
+
+  const std::vector<IndexAndAttrSet> &getAttributeGroups() const {
+    return AttributeGroups;
+  }
+
+  const ComdatSetType &getComdats() const { return Comdats; }
+  unsigned getComdatID(const Comdat *C) const;
+
+  /// getGlobalBasicBlockID - This returns the function-specific ID for the
+  /// specified basic block.  This is relatively expensive information, so it
+  /// should only be used by rare constructs such as address-of-label.
+  unsigned getGlobalBasicBlockID(const BasicBlock *BB) const;
+
+  /// incorporateFunction/purgeFunction - If you'd like to deal with a function,
+  /// use these two methods to get its data into the ValueEnumerator!
+  void incorporateFunction(const Function &F);
+
+  void purgeFunction();
+  uint64_t computeBitsRequiredForTypeIndicies() const;
+
+private:
+  void OptimizeConstants(unsigned CstStart, unsigned CstEnd);
+
+  /// Reorder the reachable metadata.
+  ///
+  /// This is not just an optimization, but is mandatory for emitting MDString
+  /// correctly.
+  void organizeMetadata();
+
+  /// Drop the function tag from the transitive operands of the given node.
+  void dropFunctionFromMetadata(MetadataMapType::value_type &FirstMD);
+
+  /// Incorporate the function metadata.
+  ///
+  /// This should be called before enumerating LocalAsMetadata for the
+  /// function.
+  void incorporateFunctionMetadata(const Function &F);
+
+  /// Enumerate a single instance of metadata with the given function tag.
+  ///
+  /// If \c MD has already been enumerated, check that \c F matches its
+  /// function tag.  If not, call \a dropFunctionFromMetadata().
+  ///
+  /// Otherwise, mark \c MD as visited.  Assign it an ID, or just return it if
+  /// it's an \a MDNode.
+  const MDNode *enumerateMetadataImpl(unsigned F, const Metadata *MD);
+
+  unsigned getMetadataFunctionID(const Function *F) const;
+
+  /// Enumerate reachable metadata in (almost) post-order.
+  ///
+  /// Enumerate all the metadata reachable from MD.  We want to minimize the
+  /// cost of reading bitcode records, and so the primary consideration is that
+  /// operands of uniqued nodes are resolved before the nodes are read.  This
+  /// avoids re-uniquing them on the context and factors away RAUW support.
+  ///
+  /// This algorithm guarantees that subgraphs of uniqued nodes are in
+  /// post-order.  Distinct subgraphs reachable only from a single uniqued node
+  /// will be in post-order.
+  ///
+  /// \note The relative order of a distinct and uniqued node is irrelevant.
+  /// \a organizeMetadata() will later partition distinct nodes ahead of
+  /// uniqued ones.
+  ///{
+  void EnumerateMetadata(const Function *F, const Metadata *MD);
+  void EnumerateMetadata(unsigned F, const Metadata *MD);
+  ///}
+
+  void EnumerateFunctionLocalMetadata(const Function &F,
+                                      const LocalAsMetadata *Local);
+  void EnumerateFunctionLocalMetadata(unsigned F, const LocalAsMetadata *Local);
+  void EnumerateFunctionLocalListMetadata(const Function &F,
+                                          const DIArgList *ArgList);
+  void EnumerateFunctionLocalListMetadata(unsigned F, const DIArgList *Arglist);
+  void EnumerateNamedMDNode(const NamedMDNode *NMD);
+  void EnumerateValue(const Value *V);
+  void EnumerateType(Type *T);
+  void EnumerateOperandType(const Value *V);
+  void EnumerateAttributes(AttributeList PAL);
+
+  void EnumerateValueSymbolTable(const ValueSymbolTable &ST);
+  void EnumerateNamedMetadata(const Module &M);
+};
+
+} // end namespace dxil
+} // end namespace llvm
+
+#endif // LLVM_DXILWRITER_VALUEENUMERATOR_H

llvm/lib/Target/DirectX/DXILWriter/DXILValueEnumerator.cpp

+//===- ValueEnumerator.cpp - Number values and types for bitcode writer ---===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements the ValueEnumerator class.
+// Forked from lib/Bitcode/Writer
+//
+//===----------------------------------------------------------------------===//
+
+#include "DXILValueEnumerator.h"
+#include "llvm/ADT/SmallVector.h"
+#include "llvm/Config/llvm-config.h"
+#include "llvm/IR/Argument.h"
+#include "llvm/IR/BasicBlock.h"
+#include "llvm/IR/Constant.h"
+#include "llvm/IR/DebugInfoMetadata.h"
+#include "llvm/IR/DerivedTypes.h"
+#include "llvm/IR/Function.h"
+#include "llvm/IR/GlobalAlias.h"
+#include "llvm/IR/GlobalIFunc.h"
+#include "llvm/IR/GlobalObject.h"
+#include "llvm/IR/GlobalValue.h"
+#include "llvm/IR/GlobalVariable.h"
+#include "llvm/IR/Instruction.h"
+#include "llvm/IR/Instructions.h"
+#include "llvm/IR/Metadata.h"
+#include "llvm/IR/Module.h"
+#include "llvm/IR/Operator.h"
+#include "llvm/IR/Type.h"
+#include "llvm/IR/Use.h"
+#include "llvm/IR/User.h"
+#include "llvm/IR/Value.h"
+#include "llvm/IR/ValueSymbolTable.h"
+#include "llvm/Support/Casting.h"
+#include "llvm/Support/Compiler.h"
+#include "llvm/Support/Debug.h"
+#include "llvm/Support/MathExtras.h"
+#include "llvm/Support/raw_ostream.h"
+#include <algorithm>
+#include <cstddef>
+#include <iterator>
+#include <tuple>
+
+using namespace llvm;
+using namespace llvm::dxil;
+
+namespace {
+
+struct OrderMap {
+  DenseMap<const Value *, std::pair<unsigned, bool>> IDs;
+  unsigned LastGlobalConstantID = 0;
+  unsigned LastGlobalValueID = 0;
+
+  OrderMap() = default;
+
+  bool isGlobalConstant(unsigned ID) const {
+    return ID <= LastGlobalConstantID;
+  }
+
+  bool isGlobalValue(unsigned ID) const {
+    return ID <= LastGlobalValueID && !isGlobalConstant(ID);
+  }
+
+  unsigned size() const { return IDs.size(); }
+  std::pair<unsigned, bool> &operator[](const Value *V) { return IDs[V]; }
+
+  std::pair<unsigned, bool> lookup(const Value *V) const {
+    return IDs.lookup(V);
+  }
+
+  void index(const Value *V) {
+    // Explicitly sequence get-size and insert-value operations to avoid UB.
+    unsigned ID = IDs.size() + 1;
+    IDs[V].first = ID;
+  }
+};
+
+} // end anonymous namespace
+
+static void orderValue(const Value *V, OrderMap &OM) {
+  if (OM.lookup(V).first)
+    return;
+
+  if (const Constant *C = dyn_cast<Constant>(V)) {
+    if (C->getNumOperands() && !isa<GlobalValue>(C)) {
+      for (const Value *Op : C->operands())
+        if (!isa<BasicBlock>(Op) && !isa<GlobalValue>(Op))
+          orderValue(Op, OM);
+      if (auto *CE = dyn_cast<ConstantExpr>(C))
+        if (CE->getOpcode() == Instruction::ShuffleVector)
+          orderValue(CE->getShuffleMaskForBitcode(), OM);
+    }
+  }
+
+  // Note: we cannot cache this lookup above, since inserting into the map
+  // changes the map's size, and thus affects the other IDs.
+  OM.index(V);
+}
+
+static OrderMap orderModule(const Module &M) {
+  // This needs to match the order used by ValueEnumerator::ValueEnumerator()
+  // and ValueEnumerator::incorporateFunction().
+  OrderMap OM;
+
+  // In the reader, initializers of GlobalValues are set *after* all the
+  // globals have been read.  Rather than awkwardly modeling this behaviour
+  // directly in predictValueUseListOrderImpl(), just assign IDs to
+  // initializers of GlobalValues before GlobalValues themselves to model this
+  // implicitly.
+  for (const GlobalVariable &G : M.globals())
+    if (G.hasInitializer())
+      if (!isa<GlobalValue>(G.getInitializer()))
+        orderValue(G.getInitializer(), OM);
+  for (const GlobalAlias &A : M.aliases())
+    if (!isa<GlobalValue>(A.getAliasee()))
+      orderValue(A.getAliasee(), OM);
+  for (const GlobalIFunc &I : M.ifuncs())
+    if (!isa<GlobalValue>(I.getResolver()))
+      orderValue(I.getResolver(), OM);
+  for (const Function &F : M) {
+    for (const Use &U : F.operands())
+      if (!isa<GlobalValue>(U.get()))
+        orderValue(U.get(), OM);
+  }
+
+  // As constants used in metadata operands are emitted as module-level
+  // constants, we must order them before other operands. Also, we must order
+  // these before global values, as these will be read before setting the
+  // global values' initializers. The latter matters for constants which have
+  // uses towards other constants that are used as initializers.
+  auto orderConstantValue = [&OM](const Value *V) {
+    if ((isa<Constant>(V) && !isa<GlobalValue>(V)) || isa<InlineAsm>(V))
+      orderValue(V, OM);
+  };
+  for (const Function &F : M) {
+    if (F.isDeclaration())
+      continue;
+    for (const BasicBlock &BB : F)
+      for (const Instruction &I : BB)
+        for (const Value *V : I.operands()) {
+          if (const auto *MAV = dyn_cast<MetadataAsValue>(V)) {
+            if (const auto *VAM =
+                    dyn_cast<ValueAsMetadata>(MAV->getMetadata())) {
+              orderConstantValue(VAM->getValue());
+            } else if (const auto *AL =
+                           dyn_cast<DIArgList>(MAV->getMetadata())) {
+              for (const auto *VAM : AL->getArgs())
+                orderConstantValue(VAM->getValue());
+            }
+          }
+        }
+  }
+  OM.LastGlobalConstantID = OM.size();
+
+  // Initializers of GlobalValues are processed in
+  // BitcodeReader::ResolveGlobalAndAliasInits().  Match the order there rather
+  // than ValueEnumerator, and match the code in predictValueUseListOrderImpl()
+  // by giving IDs in reverse order.
+  //
+  // Since GlobalValues never reference each other directly (just through
+  // initializers), their relative IDs only matter for determining order of
+  // uses in their initializers.
+  for (const Function &F : M)
+    orderValue(&F, OM);
+  for (const GlobalAlias &A : M.aliases())
+    orderValue(&A, OM);
+  for (const GlobalIFunc &I : M.ifuncs())
+    orderValue(&I, OM);
+  for (const GlobalVariable &G : M.globals())
+    orderValue(&G, OM);
+  OM.LastGlobalValueID = OM.size();
+
+  for (const Function &F : M) {
+    if (F.isDeclaration())
+      continue;
+    // Here we need to match the union of ValueEnumerator::incorporateFunction()
+    // and WriteFunction().  Basic blocks are implicitly declared before
+    // anything else (by declaring their size).
+    for (const BasicBlock &BB : F)
+      orderValue(&BB, OM);
+    for (const Argument &A : F.args())
+      orderValue(&A, OM);
+    for (const BasicBlock &BB : F)
+      for (const Instruction &I : BB) {
+        for (const Value *Op : I.operands())
+          if ((isa<Constant>(*Op) && !isa<GlobalValue>(*Op)) ||
+              isa<InlineAsm>(*Op))
+            orderValue(Op, OM);
+        if (auto *SVI = dyn_cast<ShuffleVectorInst>(&I))
+          orderValue(SVI->getShuffleMaskForBitcode(), OM);
+      }
+    for (const BasicBlock &BB : F)
+      for (const Instruction &I : BB)
+        orderValue(&I, OM);
+  }
+  return OM;
+}
+
+static void predictValueUseListOrderImpl(const Value *V, const Function *F,
+                                         unsigned ID, const OrderMap &OM,
+                                         UseListOrderStack &Stack) {
+  // Predict use-list order for this one.
+  using Entry = std::pair<const Use *, unsigned>;
+  SmallVector<Entry, 64> List;
+  for (const Use &U : V->uses())
+    // Check if this user will be serialized.
+    if (OM.lookup(U.getUser()).first)
+      List.push_back(std::make_pair(&U, List.size()));
+
+  if (List.size() < 2)
+    // We may have lost some users.
+    return;
+
+  bool IsGlobalValue = OM.isGlobalValue(ID);
+  llvm::sort(List, [&](const Entry &L, const Entry &R) {
+    const Use *LU = L.first;
+    const Use *RU = R.first;
+    if (LU == RU)
+      return false;
+
+    auto LID = OM.lookup(LU->getUser()).first;
+    auto RID = OM.lookup(RU->getUser()).first;
+
+    // Global values are processed in reverse order.
+    //
+    // Moreover, initializers of GlobalValues are set *after* all the globals
+    // have been read (despite having earlier IDs).  Rather than awkwardly
+    // modeling this behaviour here, orderModule() has assigned IDs to
+    // initializers of GlobalValues before GlobalValues themselves.
+    if (OM.isGlobalValue(LID) && OM.isGlobalValue(RID)) {
+      if (LID == RID)
+        return LU->getOperandNo() > RU->getOperandNo();
+      return LID < RID;
+    }
+
+    // If ID is 4, then expect: 7 6 5 1 2 3.
+    if (LID < RID) {
+      if (RID <= ID)
+        if (!IsGlobalValue) // GlobalValue uses don't get reversed.
+          return true;
+      return false;
+    }
+    if (RID < LID) {
+      if (LID <= ID)
+        if (!IsGlobalValue) // GlobalValue uses don't get reversed.
+          return false;
+      return true;
+    }
+
+    // LID and RID are equal, so we have different operands of the same user.
+    // Assume operands are added in order for all instructions.
+    if (LID <= ID)
+      if (!IsGlobalValue) // GlobalValue uses don't get reversed.
+        return LU->getOperandNo() < RU->getOperandNo();
+    return LU->getOperandNo() > RU->getOperandNo();
+  });
+
+  if (llvm::is_sorted(List, [](const Entry &L, const Entry &R) {
+        return L.second < R.second;
+      }))
+    // Order is already correct.
+    return;
+
+  // Store the shuffle.
+  Stack.emplace_back(V, F, List.size());
+  assert(List.size() == Stack.back().Shuffle.size() && "Wrong size");
+  for (size_t I = 0, E = List.size(); I != E; ++I)
+    Stack.back().Shuffle[I] = List[I].second;
+}
+
+static void predictValueUseListOrder(const Value *V, const Function *F,
+                                     OrderMap &OM, UseListOrderStack &Stack) {
+  auto &IDPair = OM[V];
+  assert(IDPair.first && "Unmapped value");
+  if (IDPair.second)
+    // Already predicted.
+    return;
+
+  // Do the actual prediction.
+  IDPair.second = true;
+  if (!V->use_empty() && std::next(V->use_begin()) != V->use_end())
+    predictValueUseListOrderImpl(V, F, IDPair.first, OM, Stack);
+
+  // Recursive descent into constants.
+  if (const Constant *C = dyn_cast<Constant>(V)) {
+    if (C->getNumOperands()) { // Visit GlobalValues.
+      for (const Value *Op : C->operands())
+        if (isa<Constant>(Op)) // Visit GlobalValues.
+          predictValueUseListOrder(Op, F, OM, Stack);
+      if (auto *CE = dyn_cast<ConstantExpr>(C))
+        if (CE->getOpcode() == Instruction::ShuffleVector)
+          predictValueUseListOrder(CE->getShuffleMaskForBitcode(), F, OM,
+                                   Stack);
+    }
+  }
+}
+
+static UseListOrderStack predictUseListOrder(const Module &M) {
+  OrderMap OM = orderModule(M);
+
+  // Use-list orders need to be serialized after all the users have been added
+  // to a value, or else the shuffles will be incomplete.  Store them per
+  // function in a stack.
+  //
+  // Aside from function order, the order of values doesn't matter much here.
+  UseListOrderStack Stack;
+
+  // We want to visit the functions backward now so we can list function-local
+  // constants in the last Function they're used in.  Module-level constants
+  // have already been visited above.
+  for (const Function &F : llvm::reverse(M)) {
+    if (F.isDeclaration())
+      continue;
+    for (const BasicBlock &BB : F)
+      predictValueUseListOrder(&BB, &F, OM, Stack);
+    for (const Argument &A : F.args())
+      predictValueUseListOrder(&A, &F, OM, Stack);
+    for (const BasicBlock &BB : F)
+      for (const Instruction &I : BB) {
+        for (const Value *Op : I.operands())
+          if (isa<Constant>(*Op) || isa<InlineAsm>(*Op)) // Visit GlobalValues.
+            predictValueUseListOrder(Op, &F, OM, Stack);
+        if (auto *SVI = dyn_cast<ShuffleVectorInst>(&I))
+          predictValueUseListOrder(SVI->getShuffleMaskForBitcode(), &F, OM,
+                                   Stack);
+      }
+    for (const BasicBlock &BB : F)
+      for (const Instruction &I : BB)
+        predictValueUseListOrder(&I, &F, OM, Stack);
+  }
+
+  // Visit globals last, since the module-level use-list block will be seen
+  // before the function bodies are processed.
+  for (const GlobalVariable &G : M.globals())
+    predictValueUseListOrder(&G, nullptr, OM, Stack);
+  for (const Function &F : M)
+    predictValueUseListOrder(&F, nullptr, OM, Stack);
+  for (const GlobalAlias &A : M.aliases())
+    predictValueUseListOrder(&A, nullptr, OM, Stack);
+  for (const GlobalIFunc &I : M.ifuncs())
+    predictValueUseListOrder(&I, nullptr, OM, Stack);
+  for (const GlobalVariable &G : M.globals())
+    if (G.hasInitializer())
+      predictValueUseListOrder(G.getInitializer(), nullptr, OM, Stack);
+  for (const GlobalAlias &A : M.aliases())
+    predictValueUseListOrder(A.getAliasee(), nullptr, OM, Stack);
+  for (const GlobalIFunc &I : M.ifuncs())
+    predictValueUseListOrder(I.getResolver(), nullptr, OM, Stack);
+  for (const Function &F : M) {
+    for (const Use &U : F.operands())
+      predictValueUseListOrder(U.get(), nullptr, OM, Stack);
+  }
+
+  return Stack;
+}
+
+static bool isIntOrIntVectorValue(const std::pair<const Value *, unsigned> &V) {
+  return V.first->getType()->isIntOrIntVectorTy();
+}
+
+ValueEnumerator::ValueEnumerator(const Module &M,
+                                 bool ShouldPreserveUseListOrder)
+    : ShouldPreserveUseListOrder(ShouldPreserveUseListOrder) {
+  if (ShouldPreserveUseListOrder)
+    UseListOrders = predictUseListOrder(M);
+
+  // Enumerate the global variables.
+  for (const GlobalVariable &GV : M.globals()) {
+    EnumerateValue(&GV);
+    EnumerateType(GV.getValueType());
+  }
+
+  // Enumerate the functions.
+  for (const Function &F : M) {
+    EnumerateValue(&F);
+    EnumerateType(F.getValueType());
+    EnumerateAttributes(F.getAttributes());
+  }
+
+  // Enumerate the aliases.
+  for (const GlobalAlias &GA : M.aliases()) {
+    EnumerateValue(&GA);
+    EnumerateType(GA.getValueType());
+  }
+
+  // Enumerate the ifuncs.
+  for (const GlobalIFunc &GIF : M.ifuncs()) {
+    EnumerateValue(&GIF);
+    EnumerateType(GIF.getValueType());
+  }
+
+  // Remember what is the cutoff between globalvalue's and other constants.
+  unsigned FirstConstant = Values.size();
+
+  // Enumerate the global variable initializers and attributes.
+  for (const GlobalVariable &GV : M.globals()) {
+    if (GV.hasInitializer())
+      EnumerateValue(GV.getInitializer());
+    if (GV.hasAttributes())
+      EnumerateAttributes(GV.getAttributesAsList(AttributeList::FunctionIndex));
+  }
+
+  // Enumerate the aliasees.
+  for (const GlobalAlias &GA : M.aliases())
+    EnumerateValue(GA.getAliasee());
+
+  // Enumerate the ifunc resolvers.
+  for (const GlobalIFunc &GIF : M.ifuncs())
+    EnumerateValue(GIF.getResolver());
+
+  // Enumerate any optional Function data.
+  for (const Function &F : M)
+    for (const Use &U : F.operands())
+      EnumerateValue(U.get());
+
+  // Enumerate the metadata type.
+  //
+  // TODO: Move this to ValueEnumerator::EnumerateOperandType() once bitcode
+  // only encodes the metadata type when it's used as a value.
+  EnumerateType(Type::getMetadataTy(M.getContext()));
+
+  // Insert constants and metadata that are named at module level into the slot
+  // pool so that the module symbol table can refer to them...
+  EnumerateValueSymbolTable(M.getValueSymbolTable());
+  EnumerateNamedMetadata(M);
+
+  SmallVector<std::pair<unsigned, MDNode *>, 8> MDs;
+  for (const GlobalVariable &GV : M.globals()) {
+    MDs.clear();
+    GV.getAllMetadata(MDs);
+    for (const auto &I : MDs)
+      // FIXME: Pass GV to EnumerateMetadata and arrange for the bitcode writer
+      // to write metadata to the global variable's own metadata block
+      // (PR28134).
+      EnumerateMetadata(nullptr, I.second);
+  }
+
+  // Enumerate types used by function bodies and argument lists.
+  for (const Function &F : M) {
+    for (const Argument &A : F.args())
+      EnumerateType(A.getType());
+
+    // Enumerate metadata attached to this function.
+    MDs.clear();
+    F.getAllMetadata(MDs);
+    for (const auto &I : MDs)
+      EnumerateMetadata(F.isDeclaration() ? nullptr : &F, I.second);
+
+    for (const BasicBlock &BB : F)
+      for (const Instruction &I : BB) {
+        for (const Use &Op : I.operands()) {
+          auto *MD = dyn_cast<MetadataAsValue>(&Op);
+          if (!MD) {
+            EnumerateOperandType(Op);
+            continue;
+          }
+
+          // Local metadata is enumerated during function-incorporation, but
+          // any ConstantAsMetadata arguments in a DIArgList should be examined
+          // now.
+          if (isa<LocalAsMetadata>(MD->getMetadata()))
+            continue;
+          if (auto *AL = dyn_cast<DIArgList>(MD->getMetadata())) {
+            for (auto *VAM : AL->getArgs())
+              if (isa<ConstantAsMetadata>(VAM))
+                EnumerateMetadata(&F, VAM);
+            continue;
+          }
+
+          EnumerateMetadata(&F, MD->getMetadata());
+        }
+        if (auto *SVI = dyn_cast<ShuffleVectorInst>(&I))
+          EnumerateType(SVI->getShuffleMaskForBitcode()->getType());
+        if (auto *GEP = dyn_cast<GetElementPtrInst>(&I))
+          EnumerateType(GEP->getSourceElementType());
+        if (auto *AI = dyn_cast<AllocaInst>(&I))
+          EnumerateType(AI->getAllocatedType());
+        EnumerateType(I.getType());
+        if (const auto *Call = dyn_cast<CallBase>(&I)) {
+          EnumerateAttributes(Call->getAttributes());
+          EnumerateType(Call->getFunctionType());
+        }
+
+        // Enumerate metadata attached with this instruction.
+        MDs.clear();
+        I.getAllMetadataOtherThanDebugLoc(MDs);
+        for (unsigned i = 0, e = MDs.size(); i != e; ++i)
+          EnumerateMetadata(&F, MDs[i].second);
+
+        // Don't enumerate the location directly -- it has a special record
+        // type -- but enumerate its operands.
+        if (DILocation *L = I.getDebugLoc())
+          for (const Metadata *Op : L->operands())
+            EnumerateMetadata(&F, Op);
+      }
+  }
+
+  // Optimize constant ordering.
+  OptimizeConstants(FirstConstant, Values.size());
+
+  // Organize metadata ordering.
+  organizeMetadata();
+}
+
+unsigned ValueEnumerator::getInstructionID(const Instruction *Inst) const {
+  InstructionMapType::const_iterator I = InstructionMap.find(Inst);
+  assert(I != InstructionMap.end() && "Instruction is not mapped!");
+  return I->second;
+}
+
+unsigned ValueEnumerator::getComdatID(const Comdat *C) const {
+  unsigned ComdatID = Comdats.idFor(C);
+  assert(ComdatID && "Comdat not found!");
+  return ComdatID;
+}
+
+void ValueEnumerator::setInstructionID(const Instruction *I) {
+  InstructionMap[I] = InstructionCount++;
+}
+
+unsigned ValueEnumerator::getValueID(const Value *V) const {
+  if (auto *MD = dyn_cast<MetadataAsValue>(V))
+    return getMetadataID(MD->getMetadata());
+
+  ValueMapType::const_iterator I = ValueMap.find(V);
+  assert(I != ValueMap.end() && "Value not in slotcalculator!");
+  return I->second - 1;
+}
+
+#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
+LLVM_DUMP_METHOD void ValueEnumerator::dump() const {
+  print(dbgs(), ValueMap, "Default");
+  dbgs() << '\n';
+  print(dbgs(), MetadataMap, "MetaData");
+  dbgs() << '\n';
+}
+#endif
+
+void ValueEnumerator::print(raw_ostream &OS, const ValueMapType &Map,
+                            const char *Name) const {
+  OS << "Map Name: " << Name << "\n";
+  OS << "Size: " << Map.size() << "\n";
+  for (const auto &I : Map) {
+    const Value *V = I.first;
+    if (V->hasName())
+      OS << "Value: " << V->getName();
+    else
+      OS << "Value: [null]\n";
+    V->print(errs());
+    errs() << '\n';
+
+    OS << " Uses(" << V->getNumUses() << "):";
+    for (const Use &U : V->uses()) {
+      if (&U != &*V->use_begin())
+        OS << ",";
+      if (U->hasName())
+        OS << " " << U->getName();
+      else
+        OS << " [null]";
+    }
+    OS << "\n\n";
+  }
+}
+
+void ValueEnumerator::print(raw_ostream &OS, const MetadataMapType &Map,
+                            const char *Name) const {
+  OS << "Map Name: " << Name << "\n";
+  OS << "Size: " << Map.size() << "\n";
+  for (const auto &I : Map) {
+    const Metadata *MD = I.first;
+    OS << "Metadata: slot = " << I.second.ID << "\n";
+    OS << "Metadata: function = " << I.second.F << "\n";
+    MD->print(OS);
+    OS << "\n";
+  }
+}
+
+/// OptimizeConstants - Reorder constant pool for denser encoding.
+void ValueEnumerator::OptimizeConstants(unsigned CstStart, unsigned CstEnd) {
+  if (CstStart == CstEnd || CstStart + 1 == CstEnd)
+    return;
+
+  if (ShouldPreserveUseListOrder)
+    // Optimizing constants makes the use-list order difficult to predict.
+    // Disable it for now when trying to preserve the order.
+    return;
+
+  std::stable_sort(Values.begin() + CstStart, Values.begin() + CstEnd,
+                   [this](const std::pair<const Value *, unsigned> &LHS,
+                          const std::pair<const Value *, unsigned> &RHS) {
+                     // Sort by plane.
+                     if (LHS.first->getType() != RHS.first->getType())
+                       return getTypeID(LHS.first->getType()) <
+                              getTypeID(RHS.first->getType());
+                     // Then by frequency.
+                     return LHS.second > RHS.second;
+                   });
+
+  // Ensure that integer and vector of integer constants are at the start of the
+  // constant pool.  This is important so that GEP structure indices come before
+  // gep constant exprs.
+  std::stable_partition(Values.begin() + CstStart, Values.begin() + CstEnd,
+                        isIntOrIntVectorValue);
+
+  // Rebuild the modified portion of ValueMap.
+  for (; CstStart != CstEnd; ++CstStart)
+    ValueMap[Values[CstStart].first] = CstStart + 1;
+}
+
+/// EnumerateValueSymbolTable - Insert all of the values in the specified symbol
+/// table into the values table.
+void ValueEnumerator::EnumerateValueSymbolTable(const ValueSymbolTable &VST) {
+  for (ValueSymbolTable::const_iterator VI = VST.begin(), VE = VST.end();
+       VI != VE; ++VI)
+    EnumerateValue(VI->getValue());
+}
+
+/// Insert all of the values referenced by named metadata in the specified
+/// module.
+void ValueEnumerator::EnumerateNamedMetadata(const Module &M) {
+  for (const auto &I : M.named_metadata())
+    EnumerateNamedMDNode(&I);
+}
+
+void ValueEnumerator::EnumerateNamedMDNode(const NamedMDNode *MD) {
+  for (unsigned i = 0, e = MD->getNumOperands(); i != e; ++i)
+    EnumerateMetadata(nullptr, MD->getOperand(i));
+}
+
+unsigned ValueEnumerator::getMetadataFunctionID(const Function *F) const {
+  return F ? getValueID(F) + 1 : 0;
+}
+
+void ValueEnumerator::EnumerateMetadata(const Function *F, const Metadata *MD) {
+  EnumerateMetadata(getMetadataFunctionID(F), MD);
+}
+
+void ValueEnumerator::EnumerateFunctionLocalMetadata(
+    const Function &F, const LocalAsMetadata *Local) {
+  EnumerateFunctionLocalMetadata(getMetadataFunctionID(&F), Local);
+}
+
+void ValueEnumerator::EnumerateFunctionLocalListMetadata(
+    const Function &F, const DIArgList *ArgList) {
+  EnumerateFunctionLocalListMetadata(getMetadataFunctionID(&F), ArgList);
+}
+
+void ValueEnumerator::dropFunctionFromMetadata(
+    MetadataMapType::value_type &FirstMD) {
+  SmallVector<const MDNode *, 64> Worklist;
+  auto push = [&Worklist](MetadataMapType::value_type &MD) {
+    auto &Entry = MD.second;
+
+    // Nothing to do if this metadata isn't tagged.
+    if (!Entry.F)
+      return;
+
+    // Drop the function tag.
+    Entry.F = 0;
+
+    // If this is has an ID and is an MDNode, then its operands have entries as
+    // well.  We need to drop the function from them too.
+    if (Entry.ID)
+      if (auto *N = dyn_cast<MDNode>(MD.first))
+        Worklist.push_back(N);
+  };
+  push(FirstMD);
+  while (!Worklist.empty())
+    for (const Metadata *Op : Worklist.pop_back_val()->operands()) {
+      if (!Op)
+        continue;
+      auto MD = MetadataMap.find(Op);
+      if (MD != MetadataMap.end())
+        push(*MD);
+    }
+}
+
+void ValueEnumerator::EnumerateMetadata(unsigned F, const Metadata *MD) {
+  // It's vital for reader efficiency that uniqued subgraphs are done in
+  // post-order; it's expensive when their operands have forward references.
+  // If a distinct node is referenced from a uniqued node, it'll be delayed
+  // until the uniqued subgraph has been completely traversed.
+  SmallVector<const MDNode *, 32> DelayedDistinctNodes;
+
+  // Start by enumerating MD, and then work through its transitive operands in
+  // post-order.  This requires a depth-first search.
+  SmallVector<std::pair<const MDNode *, MDNode::op_iterator>, 32> Worklist;
+  if (const MDNode *N = enumerateMetadataImpl(F, MD))
+    Worklist.push_back(std::make_pair(N, N->op_begin()));
+
+  while (!Worklist.empty()) {
+    const MDNode *N = Worklist.back().first;
+
+    // Enumerate operands until we hit a new node.  We need to traverse these
+    // nodes' operands before visiting the rest of N's operands.
+    MDNode::op_iterator I = std::find_if(
+        Worklist.back().second, N->op_end(),
+        [&](const Metadata *MD) { return enumerateMetadataImpl(F, MD); });
+    if (I != N->op_end()) {
+      auto *Op = cast<MDNode>(*I);
+      Worklist.back().second = ++I;
+
+      // Delay traversing Op if it's a distinct node and N is uniqued.
+      if (Op->isDistinct() && !N->isDistinct())
+        DelayedDistinctNodes.push_back(Op);
+      else
+        Worklist.push_back(std::make_pair(Op, Op->op_begin()));
+      continue;
+    }
+
+    // All the operands have been visited.  Now assign an ID.
+    Worklist.pop_back();
+    MDs.push_back(N);
+    MetadataMap[N].ID = MDs.size();
+
+    // Flush out any delayed distinct nodes; these are all the distinct nodes
+    // that are leaves in last uniqued subgraph.
+    if (Worklist.empty() || Worklist.back().first->isDistinct()) {
+      for (const MDNode *N : DelayedDistinctNodes)
+        Worklist.push_back(std::make_pair(N, N->op_begin()));
+      DelayedDistinctNodes.clear();
+    }
+  }
+}
+
+const MDNode *ValueEnumerator::enumerateMetadataImpl(unsigned F,
+                                                     const Metadata *MD) {
+  if (!MD)
+    return nullptr;
+
+  assert(
+      (isa<MDNode>(MD) || isa<MDString>(MD) || isa<ConstantAsMetadata>(MD)) &&
+      "Invalid metadata kind");
+
+  auto Insertion = MetadataMap.insert(std::make_pair(MD, MDIndex(F)));
+  MDIndex &Entry = Insertion.first->second;
+  if (!Insertion.second) {
+    // Already mapped.  If F doesn't match the function tag, drop it.
+    if (Entry.hasDifferentFunction(F))
+      dropFunctionFromMetadata(*Insertion.first);
+    return nullptr;
+  }
+
+  // Don't assign IDs to metadata nodes.
+  if (auto *N = dyn_cast<MDNode>(MD))
+    return N;
+
+  // Save the metadata.
+  MDs.push_back(MD);
+  Entry.ID = MDs.size();
+
+  // Enumerate the constant, if any.
+  if (auto *C = dyn_cast<ConstantAsMetadata>(MD))
+    EnumerateValue(C->getValue());
+
+  return nullptr;
+}
+
+/// EnumerateFunctionLocalMetadata - Incorporate function-local metadata
+/// information reachable from the metadata.
+void ValueEnumerator::EnumerateFunctionLocalMetadata(
+    unsigned F, const LocalAsMetadata *Local) {
+  assert(F && "Expected a function");
+
+  // Check to see if it's already in!
+  MDIndex &Index = MetadataMap[Local];
+  if (Index.ID) {
+    assert(Index.F == F && "Expected the same function");
+    return;
+  }
+
+  MDs.push_back(Local);
+  Index.F = F;
+  Index.ID = MDs.size();
+
+  EnumerateValue(Local->getValue());
+}
+
+/// EnumerateFunctionLocalListMetadata - Incorporate function-local metadata
+/// information reachable from the metadata.
+void ValueEnumerator::EnumerateFunctionLocalListMetadata(
+    unsigned F, const DIArgList *ArgList) {
+  assert(F && "Expected a function");
+
+  // Check to see if it's already in!
+  MDIndex &Index = MetadataMap[ArgList];
+  if (Index.ID) {
+    assert(Index.F == F && "Expected the same function");
+    return;
+  }
+
+  for (ValueAsMetadata *VAM : ArgList->getArgs()) {
+    if (isa<LocalAsMetadata>(VAM)) {
+      assert(MetadataMap.count(VAM) &&
+             "LocalAsMetadata should be enumerated before DIArgList");
+      assert(MetadataMap[VAM].F == F &&
+             "Expected LocalAsMetadata in the same function");
+    } else {
+      assert(isa<ConstantAsMetadata>(VAM) &&
+             "Expected LocalAsMetadata or ConstantAsMetadata");
+      assert(ValueMap.count(VAM->getValue()) &&
+             "Constant should be enumerated beforeDIArgList");
+      EnumerateMetadata(F, VAM);
+    }
+  }
+
+  MDs.push_back(ArgList);
+  Index.F = F;
+  Index.ID = MDs.size();
+}
+
+static unsigned getMetadataTypeOrder(const Metadata *MD) {
+  // Strings are emitted in bulk and must come first.
+  if (isa<MDString>(MD))
+    return 0;
+
+  // ConstantAsMetadata doesn't reference anything.  We may as well shuffle it
+  // to the front since we can detect it.
+  auto *N = dyn_cast<MDNode>(MD);
+  if (!N)
+    return 1;
+
+  // The reader is fast forward references for distinct node operands, but slow
+  // when uniqued operands are unresolved.
+  return N->isDistinct() ? 2 : 3;
+}
+
+void ValueEnumerator::organizeMetadata() {
+  assert(MetadataMap.size() == MDs.size() &&
+         "Metadata map and vector out of sync");
+
+  if (MDs.empty())
+    return;
+
+  // Copy out the index information from MetadataMap in order to choose a new
+  // order.
+  SmallVector<MDIndex, 64> Order;
+  Order.reserve(MetadataMap.size());
+  for (const Metadata *MD : MDs)
+    Order.push_back(MetadataMap.lookup(MD));
+
+  // Partition:
+  //   - by function, then
+  //   - by isa<MDString>
+  // and then sort by the original/current ID.  Since the IDs are guaranteed to
+  // be unique, the result of std::sort will be deterministic.  There's no need
+  // for std::stable_sort.
+  llvm::sort(Order, [this](MDIndex LHS, MDIndex RHS) {
+    return std::make_tuple(LHS.F, getMetadataTypeOrder(LHS.get(MDs)), LHS.ID) <
+           std::make_tuple(RHS.F, getMetadataTypeOrder(RHS.get(MDs)), RHS.ID);
+  });
+
+  // Rebuild MDs, index the metadata ranges for each function in FunctionMDs,
+  // and fix up MetadataMap.
+  std::vector<const Metadata *> OldMDs;
+  MDs.swap(OldMDs);
+  MDs.reserve(OldMDs.size());
+  for (unsigned I = 0, E = Order.size(); I != E && !Order[I].F; ++I) {
+    auto *MD = Order[I].get(OldMDs);
+    MDs.push_back(MD);
+    MetadataMap[MD].ID = I + 1;
+    if (isa<MDString>(MD))
+      ++NumMDStrings;
+  }
+
+  // Return early if there's nothing for the functions.
+  if (MDs.size() == Order.size())
+    return;
+
+  // Build the function metadata ranges.
+  MDRange R;
+  FunctionMDs.reserve(OldMDs.size());
+  unsigned PrevF = 0;
+  for (unsigned I = MDs.size(), E = Order.size(), ID = MDs.size(); I != E;
+       ++I) {
+    unsigned F = Order[I].F;
+    if (!PrevF) {
+      PrevF = F;
+    } else if (PrevF != F) {
+      R.Last = FunctionMDs.size();
+      std::swap(R, FunctionMDInfo[PrevF]);
+      R.First = FunctionMDs.size();
+
+      ID = MDs.size();
+      PrevF = F;
+    }
+
+    auto *MD = Order[I].get(OldMDs);
+    FunctionMDs.push_back(MD);
+    MetadataMap[MD].ID = ++ID;
+    if (isa<MDString>(MD))
+      ++R.NumStrings;
+  }
+  R.Last = FunctionMDs.size();
+  FunctionMDInfo[PrevF] = R;
+}
+
+void ValueEnumerator::incorporateFunctionMetadata(const Function &F) {
+  NumModuleMDs = MDs.size();
+
+  auto R = FunctionMDInfo.lookup(getValueID(&F) + 1);
+  NumMDStrings = R.NumStrings;
+  MDs.insert(MDs.end(), FunctionMDs.begin() + R.First,
+             FunctionMDs.begin() + R.Last);
+}
+
+void ValueEnumerator::EnumerateValue(const Value *V) {
+  assert(!V->getType()->isVoidTy() && "Can't insert void values!");
+  assert(!isa<MetadataAsValue>(V) && "EnumerateValue doesn't handle Metadata!");
+
+  // Check to see if it's already in!
+  unsigned &ValueID = ValueMap[V];
+  if (ValueID) {
+    // Increment use count.
+    Values[ValueID - 1].second++;
+    return;
+  }
+
+  if (auto *GO = dyn_cast<GlobalObject>(V))
+    if (const Comdat *C = GO->getComdat())
+      Comdats.insert(C);
+
+  // Enumerate the type of this value.
+  EnumerateType(V->getType());
+
+  if (const Constant *C = dyn_cast<Constant>(V)) {
+    if (isa<GlobalValue>(C)) {
+      // Initializers for globals are handled explicitly elsewhere.
+    } else if (C->getNumOperands()) {
+      // If a constant has operands, enumerate them.  This makes sure that if a
+      // constant has uses (for example an array of const ints), that they are
+      // inserted also.
+
+      // We prefer to enumerate them with values before we enumerate the user
+      // itself.  This makes it more likely that we can avoid forward references
+      // in the reader.  We know that there can be no cycles in the constants
+      // graph that don't go through a global variable.
+      for (User::const_op_iterator I = C->op_begin(), E = C->op_end(); I != E;
+           ++I)
+        if (!isa<BasicBlock>(*I)) // Don't enumerate BB operand to BlockAddress.
+          EnumerateValue(*I);
+      if (auto *CE = dyn_cast<ConstantExpr>(C)) {
+        if (CE->getOpcode() == Instruction::ShuffleVector)
+          EnumerateValue(CE->getShuffleMaskForBitcode());
+        if (auto *GEP = dyn_cast<GEPOperator>(CE))
+          EnumerateType(GEP->getSourceElementType());
+      }
+
+      // Finally, add the value.  Doing this could make the ValueID reference be
+      // dangling, don't reuse it.
+      Values.push_back(std::make_pair(V, 1U));
+      ValueMap[V] = Values.size();
+      return;
+    }
+  }
+
+  // Add the value.
+  Values.push_back(std::make_pair(V, 1U));
+  ValueID = Values.size();
+}
+
+void ValueEnumerator::EnumerateType(Type *Ty) {
+  unsigned *TypeID = &TypeMap[Ty];
+
+  // We've already seen this type.
+  if (*TypeID)
+    return;
+
+  // If it is a non-anonymous struct, mark the type as being visited so that we
+  // don't recursively visit it.  This is safe because we allow forward
+  // references of these in the bitcode reader.
+  if (StructType *STy = dyn_cast<StructType>(Ty))
+    if (!STy->isLiteral())
+      *TypeID = ~0U;
+
+  // Enumerate all of the subtypes before we enumerate this type.  This ensures
+  // that the type will be enumerated in an order that can be directly built.
+  for (Type *SubTy : Ty->subtypes())
+    EnumerateType(SubTy);
+
+  // Refresh the TypeID pointer in case the table rehashed.
+  TypeID = &TypeMap[Ty];
+
+  // Check to see if we got the pointer another way.  This can happen when
+  // enumerating recursive types that hit the base case deeper than they start.
+  //
+  // If this is actually a struct that we are treating as forward ref'able,
+  // then emit the definition now that all of its contents are available.
+  if (*TypeID && *TypeID != ~0U)
+    return;
+
+  // Add this type now that its contents are all happily enumerated.
+  Types.push_back(Ty);
+
+  *TypeID = Types.size();
+}
+
+// Enumerate the types for the specified value.  If the value is a constant,
+// walk through it, enumerating the types of the constant.
+void ValueEnumerator::EnumerateOperandType(const Value *V) {
+  EnumerateType(V->getType());
+
+  assert(!isa<MetadataAsValue>(V) && "Unexpected metadata operand");
+
+  const Constant *C = dyn_cast<Constant>(V);
+  if (!C)
+    return;
+
+  // If this constant is already enumerated, ignore it, we know its type must
+  // be enumerated.
+  if (ValueMap.count(C))
+    return;
+
+  // This constant may have operands, make sure to enumerate the types in
+  // them.
+  for (const Value *Op : C->operands()) {
+    // Don't enumerate basic blocks here, this happens as operands to
+    // blockaddress.
+    if (isa<BasicBlock>(Op))
+      continue;
+
+    EnumerateOperandType(Op);
+  }
+  if (auto *CE = dyn_cast<ConstantExpr>(C)) {
+    if (CE->getOpcode() == Instruction::ShuffleVector)
+      EnumerateOperandType(CE->getShuffleMaskForBitcode());
+    if (CE->getOpcode() == Instruction::GetElementPtr)
+      EnumerateType(cast<GEPOperator>(CE)->getSourceElementType());
+  }
+}
+
+void ValueEnumerator::EnumerateAttributes(AttributeList PAL) {
+  if (PAL.isEmpty())
+    return; // null is always 0.
+
+  // Do a lookup.
+  unsigned &Entry = AttributeListMap[PAL];
+  if (Entry == 0) {
+    // Never saw this before, add it.
+    AttributeLists.push_back(PAL);
+    Entry = AttributeLists.size();
+  }
+
+  // Do lookups for all attribute groups.
+  for (unsigned i : PAL.indexes()) {
+    AttributeSet AS = PAL.getAttributes(i);
+    if (!AS.hasAttributes())
+      continue;
+    IndexAndAttrSet Pair = {i, AS};
+    unsigned &Entry = AttributeGroupMap[Pair];
+    if (Entry == 0) {
+      AttributeGroups.push_back(Pair);
+      Entry = AttributeGroups.size();
+
+      for (Attribute Attr : AS) {
+        if (Attr.isTypeAttribute())
+          EnumerateType(Attr.getValueAsType());
+      }
+    }
+  }
+}
+
+void ValueEnumerator::incorporateFunction(const Function &F) {
+  InstructionCount = 0;
+  NumModuleValues = Values.size();
+
+  // Add global metadata to the function block.  This doesn't include
+  // LocalAsMetadata.
+  incorporateFunctionMetadata(F);
+
+  // Adding function arguments to the value table.
+  for (const auto &I : F.args()) {
+    EnumerateValue(&I);
+    if (I.hasAttribute(Attribute::ByVal))
+      EnumerateType(I.getParamByValType());
+    else if (I.hasAttribute(Attribute::StructRet))
+      EnumerateType(I.getParamStructRetType());
+    else if (I.hasAttribute(Attribute::ByRef))
+      EnumerateType(I.getParamByRefType());
+  }
+  FirstFuncConstantID = Values.size();
+
+  // Add all function-level constants to the value table.
+  for (const BasicBlock &BB : F) {
+    for (const Instruction &I : BB) {
+      for (const Use &OI : I.operands()) {
+        if ((isa<Constant>(OI) && !isa<GlobalValue>(OI)) || isa<InlineAsm>(OI))
+          EnumerateValue(OI);
+      }
+      if (auto *SVI = dyn_cast<ShuffleVectorInst>(&I))
+        EnumerateValue(SVI->getShuffleMaskForBitcode());
+    }
+    BasicBlocks.push_back(&BB);
+    ValueMap[&BB] = BasicBlocks.size();
+  }
+
+  // Optimize the constant layout.
+  OptimizeConstants(FirstFuncConstantID, Values.size());
+
+  // Add the function's parameter attributes so they are available for use in
+  // the function's instruction.
+  EnumerateAttributes(F.getAttributes());
+
+  FirstInstID = Values.size();
+
+  SmallVector<LocalAsMetadata *, 8> FnLocalMDVector;
+  SmallVector<DIArgList *, 8> ArgListMDVector;
+  // Add all of the instructions.
+  for (const BasicBlock &BB : F) {
+    for (const Instruction &I : BB) {
+      for (const Use &OI : I.operands()) {
+        if (auto *MD = dyn_cast<MetadataAsValue>(&OI)) {
+          if (auto *Local = dyn_cast<LocalAsMetadata>(MD->getMetadata())) {
+            // Enumerate metadata after the instructions they might refer to.
+            FnLocalMDVector.push_back(Local);
+          } else if (auto *ArgList = dyn_cast<DIArgList>(MD->getMetadata())) {
+            ArgListMDVector.push_back(ArgList);
+            for (ValueAsMetadata *VMD : ArgList->getArgs()) {
+              if (auto *Local = dyn_cast<LocalAsMetadata>(VMD)) {
+                // Enumerate metadata after the instructions they might refer
+                // to.
+                FnLocalMDVector.push_back(Local);
+              }
+            }
+          }
+        }
+      }
+
+      if (!I.getType()->isVoidTy())
+        EnumerateValue(&I);
+    }
+  }
+
+  // Add all of the function-local metadata.
+  for (unsigned i = 0, e = FnLocalMDVector.size(); i != e; ++i) {
+    // At this point, every local values have been incorporated, we shouldn't
+    // have a metadata operand that references a value that hasn't been seen.
+    assert(ValueMap.count(FnLocalMDVector[i]->getValue()) &&
+           "Missing value for metadata operand");
+    EnumerateFunctionLocalMetadata(F, FnLocalMDVector[i]);
+  }
+  // DIArgList entries must come after function-local metadata, as it is not
+  // possible to forward-reference them.
+  for (const DIArgList *ArgList : ArgListMDVector)
+    EnumerateFunctionLocalListMetadata(F, ArgList);
+}
+
+void ValueEnumerator::purgeFunction() {
+  /// Remove purged values from the ValueMap.
+  for (unsigned i = NumModuleValues, e = Values.size(); i != e; ++i)
+    ValueMap.erase(Values[i].first);
+  for (unsigned i = NumModuleMDs, e = MDs.size(); i != e; ++i)
+    MetadataMap.erase(MDs[i]);
+  for (const BasicBlock *BB : BasicBlocks)
+    ValueMap.erase(BB);
+
+  Values.resize(NumModuleValues);
+  MDs.resize(NumModuleMDs);
+  BasicBlocks.clear();
+  NumMDStrings = 0;
+}
+
+static void IncorporateFunctionInfoGlobalBBIDs(
+    const Function *F, DenseMap<const BasicBlock *, unsigned> &IDMap) {
+  unsigned Counter = 0;
+  for (const BasicBlock &BB : *F)
+    IDMap[&BB] = ++Counter;
+}
+
+/// getGlobalBasicBlockID - This returns the function-specific ID for the
+/// specified basic block.  This is relatively expensive information, so it
+/// should only be used by rare constructs such as address-of-label.
+unsigned ValueEnumerator::getGlobalBasicBlockID(const BasicBlock *BB) const {
+  unsigned &Idx = GlobalBasicBlockIDs[BB];
+  if (Idx != 0)
+    return Idx - 1;
+
+  IncorporateFunctionInfoGlobalBBIDs(BB->getParent(), GlobalBasicBlockIDs);
+  return getGlobalBasicBlockID(BB);
+}
+
+uint64_t ValueEnumerator::computeBitsRequiredForTypeIndicies() const {
+  return Log2_32_Ceil(getTypes().size() + 1);
+}

llvm/lib/Target/DirectX/DXILWriter/DXILWriterPass.h

+//===-- DXILWriterPass.h - Bitcode writing pass --------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+/// \file
+///
+/// This file provides a bitcode writing pass.
+///
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_BITCODE_DXILWriterPass_H
+#define LLVM_BITCODE_DXILWriterPass_H
+
+#include "DirectX.h"
+#include "llvm/Bitcode/BitcodeWriter.h"
+#include "llvm/IR/PassManager.h"
+
+namespace llvm {
+class Module;
+class raw_ostream;
+
+/// Create and return a pass that writes the module to the specified
+/// ostream. Note that this pass is designed for use with the legacy pass
+/// manager.
+ModulePass *createDXILWriterPass(raw_ostream &Str);
+
+} // namespace llvm
+
+#endif

llvm/lib/Target/DirectX/DXILWriter/DXILWriterPass.cpp

+//===- DXILWriterPass.cpp - Bitcode writing pass --------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// DXILWriterPass implementation.
+//
+//===----------------------------------------------------------------------===//
+
+#include "DXILWriterPass.h"
+#include "DXILBitcodeWriter.h"
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/Analysis/ModuleSummaryAnalysis.h"
+#include "llvm/IR/Module.h"
+#include "llvm/IR/PassManager.h"
+#include "llvm/InitializePasses.h"
+#include "llvm/Pass.h"
+
+using namespace llvm;
+using namespace llvm::dxil;
+
+namespace {
+class WriteDXILPass : public llvm::ModulePass {
+  raw_ostream &OS; // raw_ostream to print on
+
+public:
+  static char ID; // Pass identification, replacement for typeid
+  WriteDXILPass() : ModulePass(ID), OS(dbgs()) {
+    initializeWriteDXILPassPass(*PassRegistry::getPassRegistry());
+  }
+
+  explicit WriteDXILPass(raw_ostream &o) : ModulePass(ID), OS(o) {
+    initializeWriteDXILPassPass(*PassRegistry::getPassRegistry());
+  }
+
+  StringRef getPassName() const override { return "Bitcode Writer"; }
+
+  bool runOnModule(Module &M) override {
+    WriteDXILToFile(M, OS);
+    return false;
+  }
+  void getAnalysisUsage(AnalysisUsage &AU) const override {
+    AU.setPreservesAll();
+  }
+};
+} // namespace
+
+char WriteDXILPass::ID = 0;
+INITIALIZE_PASS_BEGIN(WriteDXILPass, "write-bitcode", "Write Bitcode", false,
+                      true)
+INITIALIZE_PASS_DEPENDENCY(ModuleSummaryIndexWrapperPass)
+INITIALIZE_PASS_END(WriteDXILPass, "write-bitcode", "Write Bitcode", false,
+                    true)
+
+ModulePass *llvm::createDXILWriterPass(raw_ostream &Str) {
+  return new WriteDXILPass(Str);
+}

llvm/lib/Target/DirectX/DirectX.h

 
 #ifndef LLVM_LIB_TARGET_DIRECTX_DIRECTX_H
 #define LLVM_LIB_TARGET_DIRECTX_DIRECTX_H
 
 namespace llvm {
 class ModulePass;
 class PassRegistry;
 
+/// Initializer for dxil writer pass
+void initializeWriteDXILPassPass(PassRegistry &);
+
 /// Initializer for DXIL-prepare
 void initializeDXILPrepareModulePass(PassRegistry &);
 
 /// Pass to convert modules into DXIL-compatable modules
 ModulePass *createDXILPrepareModulePass();
 } // namespace llvm
 
 #endif // LLVM_LIB_TARGET_DIRECTX_DIRECTX_H

llvm/lib/Target/DirectX/DirectXTargetMachine.cpp

 //===- DirectXTargetMachine.cpp - DirectX Target Implementation -*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 ///
 /// \file
 /// This file contains DirectX target initializer.
 ///
 //===----------------------------------------------------------------------===//
 
 #include "DirectXTargetMachine.h"
+#include "DXILWriter/DXILWriterPass.h"
 #include "DirectX.h"
 #include "DirectXSubtarget.h"
 #include "DirectXTargetTransformInfo.h"
 #include "TargetInfo/DirectXTargetInfo.h"
-#include "llvm/Bitcode/BitcodeWriterPass.h"
 #include "llvm/CodeGen/Passes.h"
 #include "llvm/CodeGen/TargetPassConfig.h"
 #include "llvm/IR/IRPrintingPasses.h"
 #include "llvm/IR/LegacyPassManager.h"
 #include "llvm/MC/SectionKind.h"
 #include "llvm/MC/TargetRegistry.h"
 #include "llvm/Support/CodeGen.h"
 #include "llvm/Support/Compiler.h"
bool DirectXTargetMachine::addPassesToEmitFile(
     CodeGenFileType FileType, bool DisableVerify,
     MachineModuleInfoWrapperPass *MMIWP) {
   PM.add(createDXILPrepareModulePass());
   switch (FileType) {
   case CGFT_AssemblyFile:
     PM.add(createPrintModulePass(Out, "", true));
     break;
   case CGFT_ObjectFile:
-    // TODO: Write DXIL instead of bitcode
-    PM.add(createBitcodeWriterPass(Out, true, false, false));
+    // TODO: Use MC Object streamer to write DXContainer
+    PM.add(createDXILWriterPass(Out));
     break;
   case CGFT_Null:
     break;
   }
   return false;
 }
 
 bool DirectXTargetMachine::addPassesToEmitMC(PassManagerBase &PM,

llvm/test/CMakeLists.txt

llvm_canonicalize_cmake_booleans(
   LLVM_BUILD_EXAMPLES
   LLVM_ENABLE_PLUGINS
   LLVM_BYE_LINK_INTO_TOOLS
   LLVM_HAVE_TF_AOT
   LLVM_HAVE_TF_API
   LLVM_INLINER_MODEL_AUTOGENERATED
   LLVM_RAEVICT_MODEL_AUTOGENERATED
   LLVM_ENABLE_EXPENSIVE_CHECKS
+  LLVM_INCLUDE_DXIL_TESTS
   )
 
 configure_lit_site_cfg(
   ${CMAKE_CURRENT_SOURCE_DIR}/lit.site.cfg.py.in
   ${CMAKE_CURRENT_BINARY_DIR}/lit.site.cfg.py
   MAIN_CONFIG
   ${CMAKE_CURRENT_SOURCE_DIR}/lit.cfg.py
   PATHS
set(LLVM_TEST_DEPENDS ${LLVM_TEST_DEPENDS}
           ocaml_llvm_ipo
           ocaml_llvm_passmgr_builder
           ocaml_llvm_scalar_opts
           ocaml_llvm_transform_utils
           ocaml_llvm_vectorize
         )
 endif()
 
+if (LLVM_INCLUDE_DXIL_TESTS)
+  list(APPEND LLVM_TEST_DEPENDS dxil-dis)
+endif()
+
 add_custom_target(llvm-test-depends DEPENDS ${LLVM_TEST_DEPENDS})
 set_target_properties(llvm-test-depends PROPERTIES FOLDER "Tests")
 
 if(LLVM_BUILD_TOOLS)
   set(exclude_from_check_all "")
 else()
   set(exclude_from_check_all "EXCLUDE_FROM_CHECK_ALL")
 endif()

llvm/test/lit.cfg.py

tools.extend([
     ToolSubst('Kaleidoscope-Ch8', unresolved='ignore'),
     ToolSubst('LLJITWithThinLTOSummaries', unresolved='ignore'),
     ToolSubst('LLJITWithRemoteDebugging', unresolved='ignore'),
     ToolSubst('OrcV2CBindingsBasicUsage', unresolved='ignore'),
     ToolSubst('OrcV2CBindingsAddObjectFile', unresolved='ignore'),
     ToolSubst('OrcV2CBindingsRemovableCode', unresolved='ignore'),
     ToolSubst('OrcV2CBindingsReflectProcessSymbols', unresolved='ignore'),
     ToolSubst('OrcV2CBindingsLazy', unresolved='ignore'),
-    ToolSubst('OrcV2CBindingsVeryLazy', unresolved='ignore')])
+    ToolSubst('OrcV2CBindingsVeryLazy', unresolved='ignore'),
+    ToolSubst('dxil-dis', unresolved='ignore')])
 
 llvm_config.add_tool_substitutions(tools, config.llvm_tools_dir)
 
 # Targets
 
 config.targets = frozenset(config.targets_to_build.split())
 
 for arch in config.targets_to_build.split():

llvm/test/lit.site.cfg.py.in

 config.libcxx_used = @LLVM_LIBCXX_USED@
 config.has_plugins = @LLVM_ENABLE_PLUGINS@
 config.linked_bye_extension = @LLVM_BYE_LINK_INTO_TOOLS@
 config.have_tf_aot = @LLVM_HAVE_TF_AOT@
 config.have_tf_api = @LLVM_HAVE_TF_API@
 config.llvm_inliner_model_autogenerated = @LLVM_INLINER_MODEL_AUTOGENERATED@
 config.llvm_raevict_model_autogenerated = @LLVM_RAEVICT_MODEL_AUTOGENERATED@
 config.expensive_checks = @LLVM_ENABLE_EXPENSIVE_CHECKS@
+config.dxil_tests = @LLVM_INCLUDE_DXIL_TESTS@
 
 import lit.llvm
 lit.llvm.initialize(lit_config, config)
 
 # Let the main config do the real work.
 lit_config.load_config(
     config, os.path.join(config.llvm_src_root, "test/lit.cfg.py"))

llvm/test/tools/dxil-dis/BasicIR.ll

+; RUN: llc --filetype=obj %s -o - | dxil-dis -o - | FileCheck %s
+
+; CHECK: define i32 @foo(i32 %X, i32 %Y) {
+; CHECK:   %Z = sub i32 %X, %Y
+; CHECK:   %Q = add i32 %Z, %Y
+; CHECK:   ret i32 %Q
+; CHECK: }
+
+target triple = "dxil-unknown-unknown"
+
+define i32 @foo(i32 %X, i32 %Y) {
+  %Z = sub i32 %X, %Y
+  %Q = add i32 %Z, %Y
+  ret i32 %Q
+}

llvm/test/tools/dxil-dis/attribute-filter.ll

+; RUN: llc %s --filetype=obj -o - | dxil-dis -o - | FileCheck %s
+
+; CHECK: target triple = "dxil-unknown-unknown"
+target triple = "dxil-unknown-unknown"
+
+; CHECK: Function Attrs: nounwind readnone
+; Function Attrs: norecurse nounwind readnone willreturn
+define float @fma(float %0, float %1, float %2) #0 {
+  %4 = fmul float %0, %1
+  %5 = fadd float %4, %2
+  ret float %5
+}
+
+; CHECK: attributes #0 = { nounwind readnone "disable-tail-calls"="false" }
+attributes #0 = { norecurse nounwind readnone willreturn "disable-tail-calls"="false" }

llvm/test/tools/dxil-dis/debug-info.ll

+; RUN: llc --filetype=obj %s -o - | dxil-dis -o - | FileCheck %s
+target triple = "dxil-unknown-unknown"
+target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"
+
+; CHECK: define float @fma(float, float, float) unnamed_addr #0 !dbg !6
+; Function Attrs: norecurse nounwind readnone willreturn
+define dso_local float @fma(float %0, float %1, float %2) local_unnamed_addr #0 !dbg !6 {
+; CHECK-NEXT: call void @llvm.dbg.value(metadata float %0, metadata !11, metadata !14), !dbg !15
+; CHECK-NEXT: call void @llvm.dbg.value(metadata float %1, metadata !12, metadata !14), !dbg !15
+; CHECK-NEXT: call void @llvm.dbg.value(metadata float %2, metadata !13, metadata !14), !dbg !15
+  call void @llvm.dbg.value(metadata float %0, metadata !11, metadata !DIExpression()), !dbg !14
+  call void @llvm.dbg.value(metadata float %1, metadata !12, metadata !DIExpression()), !dbg !14
+  call void @llvm.dbg.value(metadata float %2, metadata !13, metadata !DIExpression()), !dbg !14
+; CHECK-NEXT: %4 = fmul float %0, %1, !dbg !16
+; CHECK-NEXT: %5 = fadd float %4, %2, !dbg !17
+  %4 = fmul float %0, %1, !dbg !15
+  %5 = fadd float %4, %2, !dbg !16
+  ret float %5, !dbg !17
+}
+
+; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
+declare void @llvm.dbg.value(metadata, metadata, metadata) #1
+
+attributes #0 = { norecurse nounwind readnone willreturn }
+attributes #1 = { nofree nosync nounwind readnone speculatable willreturn }
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4}
+!llvm.ident = !{!5}
+
+; Other tests verify that we come back with reasonable structure for the debug
+; info types, this test just needs to ensure they are there.
+; The patch this is paired with fixes a bug where function debug info wasn't
+; being emitted correctly even though other tests verified the MD would be
+; emitted if it was referenced as module metadata.
+
+; CHECK:      !0 = distinct !DICompileUnit
+; CHECK-NEXT: !1 = !DIFile(filename:
+; CHECK:      !6 = distinct !DISubprogram(name: "fma", 
+; CHECK:      !11 = !DILocalVariable(tag:
+; CHECK-NEXT: !12 = !DILocalVariable(tag:
+; CHECK-NEXT: !13 = !DILocalVariable(tag:
+; CHECK-NEXT: !14 = !DIExpression()
+; CHECK-NEXT: !15 = !DILocation(line:
+; CHECK-NEXT: !16 = !DILocation(line:
+; CHECK-NEXT: !17 = !DILocation(line:
+; CHECK-NEXT: !18 = !DILocation(line:
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
+!1 = !DIFile(filename: "in.c", directory: "dir")
+!2 = !{}
+!3 = !{i32 7, !"Dwarf Version", i32 2}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+!5 = !{!"Some Compiler"}
+!6 = distinct !DISubprogram(name: "fma", scope: !1, file: !1, line: 1, type: !7, scopeLine: 1, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !10)
+!7 = !DISubroutineType(types: !8)
+!8 = !{!9, !9, !9, !9}
+!9 = !DIBasicType(name: "float", size: 32, encoding: DW_ATE_float)
+!10 = !{!11, !12, !13}
+!11 = !DILocalVariable(name: "x", arg: 1, scope: !6, file: !1, line: 1, type: !9)
+!12 = !DILocalVariable(name: "y", arg: 2, scope: !6, file: !1, line: 1, type: !9)
+!13 = !DILocalVariable(name: "z", arg: 3, scope: !6, file: !1, line: 1, type: !9)
+!14 = !DILocation(line: 0, scope: !6)
+!15 = !DILocation(line: 2, column: 12, scope: !6)
+!16 = !DILocation(line: 2, column: 16, scope: !6)
+!17 = !DILocation(line: 2, column: 3, scope: !6)

llvm/test/tools/dxil-dis/di-compile-unit.ll

+; RUN: llc --filetype=obj %s -o - | dxil-dis -o - | FileCheck %s
+target triple = "dxil-unknown-unknown"
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "Some Compiler", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
+!1 = !DIFile(filename: "di-compile-unit.src", directory: "/some-path")
+!2 = !{}
+!3 = !{i32 7, !"Dwarf Version", i32 2}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+
+; CHECK: !0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "Some Compiler", isOptimized: true, runtimeVersion: 0, emissionKind: 1, enums: !2)
+; CHECK: !1 = !DIFile(filename: "di-compile-unit.src", directory: "/some-path")
+; CHECK: !2 = !{}
+; CHECK: !3 = !{i32 7, !"Dwarf Version", i32 2}
+; CHECK: !4 = !{i32 2, !"Debug Info Version", i32 3}

llvm/test/tools/dxil-dis/di-subprogram.ll

+; RUN: llc --filetype=obj %s -o - | dxil-dis -o - | FileCheck %s
+target triple = "dxil-unknown-unknown"
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4}
+!llvm.used = !{!5}
+!llvm.lines = !{!13, !14, !15, !16}
+
+; CHECK: !0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "Some Compiler", isOptimized: true, runtimeVersion: 0, emissionKind: 1, enums: !2)
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "Some Compiler", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
+; CHECK: !1 = !DIFile(filename: "some-source", directory: "some-path")
+!1 = !DIFile(filename: "some-source", directory: "some-path")
+!2 = !{}
+
+; CHECK: !3 = !{i32 7, !"Dwarf Version", i32 2}
+!3 = !{i32 7, !"Dwarf Version", i32 2}
+; CHECK: !4 = !{i32 2, !"Debug Info Version", i32 3}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+
+; CHECK: !5 = distinct !DISubprogram(name: "fma", scope: !1, file: !1, line: 1, type: !6, isLocal: false, isDefinition: true, scopeLine: 1, flags: DIFlagPrototyped, isOptimized: true, function: !0, variables: !9)
+!5 = distinct !DISubprogram(name: "fma", scope: !1, file: !1, line: 1, type: !6, scopeLine: 1, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !9)
+
+; CHECK: !6 = !DISubroutineType(types: !7)
+!6 = !DISubroutineType(types: !7)
+
+; CHECK: !7 = !{!8, !8, !8, !8}
+!7 = !{!8, !8, !8, !8}
+
+; CHECK: !8 = !DIBasicType(name: "float", size: 32, encoding: DW_ATE_float)
+!8 = !DIBasicType(name: "float", size: 32, encoding: DW_ATE_float)
+
+; CHECK: !9 = !{!10, !11, !12}
+!9 = !{!10, !11, !12}
+
+; CHECK: !10 = !DILocalVariable(tag: DW_TAG_variable, name: "x", arg: 1, scope: !5, file: !1, line: 1, type: !8)
+!10 = !DILocalVariable(name: "x", arg: 1, scope: !5, file: !1, line: 1, type: !8)
+
+; CHECK: !11 = !DILocalVariable(tag: DW_TAG_variable, name: "y", arg: 2, scope: !5, file: !1, line: 1, type: !8)
+!11 = !DILocalVariable(name: "y", arg: 2, scope: !5, file: !1, line: 1, type: !8)
+
+; CHECK: !12 = !DILocalVariable(tag: DW_TAG_variable, name: "z", arg: 3, scope: !5, file: !1, line: 1, type: !8)
+!12 = !DILocalVariable(name: "z", arg: 3, scope: !5, file: !1, line: 1, type: !8)
+
+
+; CHECK: !13 = !DILocation(line: 0, scope: !5)
+; CHECK: !14 = !DILocation(line: 2, column: 12, scope: !5)
+; CHECK: !15 = !DILocation(line: 2, column: 16, scope: !5)
+; CHECK: !16 = !DILocation(line: 2, column: 3, scope: !5)
+
+!13 = !DILocation(line: 0, scope: !5)
+!14 = !DILocation(line: 2, column: 12, scope: !5)
+!15 = !DILocation(line: 2, column: 16, scope: !5)
+!16 = !DILocation(line: 2, column: 3, scope: !5)

llvm/test/tools/dxil-dis/di-subrotine.ll

+; RUN: llc --filetype=obj %s -o - | dxil-dis -o - | FileCheck %s
+target triple = "dxil-unknown-unknown"
+
+!llvm.used = !{!0}
+
+!0 = !DISubroutineType(types: !1)
+!1 = !{!2, !2, !2, !2}
+!2 = !DIBasicType(name: "float", size: 32, encoding: DW_ATE_float)
+
+; CHECK: !0 = !DISubroutineType(types: !1)
+; CHECK: !1 = !{!2, !2, !2, !2}
+; CHECK: !2 = !DIBasicType(name: "float", size: 32, encoding: DW_ATE_float)

llvm/test/tools/dxil-dis/lit.local.cfg

+if not config.dxil_tests:
+  config.unsupported = True
+config.suffixes = ['.ll']

llvm/test/tools/dxil-dis/metadata.ll

+; RUN: llc --filetype=obj %s -o - | dxil-dis 
+target triple = "dxil-unknown-unknown"
+
+!llvm.foo = !{!0}
+!llvm.bar = !{!1}
+
+!0 = !{i32 42}
+!1 = !{!"Some MDString"}
+
+; CHECK: !llvm.foo = !{!0}
+; CHECK: !llvm.bar = !{!1}
+; CHECK: !0 = !{i32 42}
+; CHECK: !1 = !{!"Some MDString"}

llvm/tools/dxil-dis/CMakeLists.txt

+option(LLVM_INCLUDE_DXIL_TESTS "Include DXIL tests" Off)
+mark_as_advanced(LLVM_INCLUDE_DXIL_TESTS)
+
+if (NOT LLVM_INCLUDE_DXIL_TESTS)
+  return()
+endif ()
+
+if (NOT "DirectX" IN_LIST LLVM_TARGETS_TO_BUILD)
+  message(FATAL_ERROR "Building dxil-dis tests is unsupported without the DirectX target")
+endif ()
+
+if (CMAKE_HOST_UNIX)
+  set(LLVM_LINK_OR_COPY create_symlink)
+else ()
+  set(LLVM_LINK_OR_COPY copy)
+endif ()
+
+if (DXIL_DIS)
+  add_custom_target(dxil-dis
+    COMMAND ${CMAKE_COMMAND} -E ${LLVM_LINK_OR_COPY} "${DXIL_DIS}" "${LLVM_RUNTIME_OUTPUT_INTDIR}/dxil-dis")
+  return()
+endif ()
+
+include(ExternalProject)
+
+set(SOURCE_DIR ${CMAKE_CURRENT_BINARY_DIR}/DXC-src)
+set(BINARY_DIR ${CMAKE_CURRENT_BINARY_DIR}/DXC-bins)
+set(GIT_SETTINGS GIT_REPOSITORY https://github.com/microsoft/DirectXShaderCompiler.git)
+
+if (DXC_SOURCE_DIR)
+  set(SOURCE_DIR ${DXC_SOURCE_DIR})
+  unset(GIT_SETTINGS)
+endif ()
+
+ExternalProject_Add(DXC
+                    ${GIT_SETTINGS}
+                    SOURCE_DIR ${SOURCE_DIR}
+                    BINARY_DIR ${BINARY_DIR}
+                    CMAKE_ARGS -C ${SOURCE_DIR}/cmake/caches/PredefinedParams.cmake -DLLVM_INCLUDE_TESTS=On
+                    BUILD_COMMAND ${CMAKE_COMMAND} --build ${BINARY_DIR} --target llvm-dis
+                    BUILD_BYPRODUCTS ${BINARY_DIR}/bin/llvm-dis
+                    INSTALL_COMMAND ""
+                    )
+
+add_custom_target(dxil-dis
+                  COMMAND ${CMAKE_COMMAND} -E ${LLVM_LINK_OR_COPY} "${BINARY_DIR}/bin/llvm-dis${CMAKE_EXECUTABLE_SUFFIX}" "${LLVM_RUNTIME_OUTPUT_INTDIR}/dxil-dis${CMAKE_EXECUTABLE_SUFFIX}"
+                  DEPENDS DXC
+                  )