[LTO] Record whether LTOUnit splitting is enabled in index

Summary:
Records in the module summary index whether the bitcode was compiled
with the option necessary to enable splitting the LTO unit
(e.g. -fsanitize=cfi, -fwhole-program-vtables, or -fsplit-lto-unit).

The information is passed down to the ModuleSummaryIndex builder via a
new module flag "EnableSplitLTOUnit", which is propagated onto a flag
on the summary index.

This is then used during the LTO link to check whether all linked
summaries were built with the same value of this flag. If not, an error
is issued when we detect a situation requiring whole program visibility
of the class hierarchy. This is the case when both of the following
conditions are met:
1) We are performing LowerTypeTests or Whole Program Devirtualization.
2) There are type tests or type checked loads in the code.

Note I have also changed the ThinLTOBitcodeWriter to also gate the
module splitting on the value of this flag.

Reviewers: pcc

Subscribers: ormris, mehdi_amini, Prazek, inglorion, eraman, steven_wu, dexonsmith, arphaman, dang, llvm-commits

Differential Revision: https://reviews.llvm.org/D53890

llvm-svn: 350948

include/llvm/Bitcode/BitcodeReader.h

@@ -51,6 +51,7 @@ class Module;
   struct BitcodeLTOInfo {
     bool IsThinLTO;
     bool HasSummary;
+    bool EnableSplitLTOUnit;
   };
 
   /// Represents a module in a bitcode file.

include/llvm/IR/ModuleSummaryIndex.h

@@ -831,6 +831,13 @@ private:
   /// union.
   bool HaveGVs;
 
+  // True if the index was created for a module compiled with -fsplit-lto-unit.
+  bool EnableSplitLTOUnit;
+
+  // True if some of the modules were compiled with -fsplit-lto-unit and
+  // some were not. Set when the combined index is created during the thin link.
+  bool PartiallySplitLTOUnits = false;
+
   std::set<std::string> CfiFunctionDefs;
   std::set<std::string> CfiFunctionDecls;
 
@@ -850,7 +857,9 @@ private:
 
 public:
   // See HaveGVs variable comment.
-  ModuleSummaryIndex(bool HaveGVs) : HaveGVs(HaveGVs), Saver(Alloc) {}
+  ModuleSummaryIndex(bool HaveGVs, bool EnableSplitLTOUnit = false)
+      : HaveGVs(HaveGVs), EnableSplitLTOUnit(EnableSplitLTOUnit), Saver(Alloc) {
+  }
 
   bool haveGVs() const { return HaveGVs; }
 
@@ -940,6 +949,12 @@ public:
     SkipModuleByDistributedBackend = true;
   }
 
+  bool enableSplitLTOUnit() const { return EnableSplitLTOUnit; }
+  void setEnableSplitLTOUnit() { EnableSplitLTOUnit = true; }
+
+  bool partiallySplitLTOUnits() const { return PartiallySplitLTOUnits; }
+  void setPartiallySplitLTOUnits() { PartiallySplitLTOUnits = true; }
+
   bool isGlobalValueLive(const GlobalValueSummary *GVS) const {
     return !WithGlobalValueDeadStripping || GVS->isLive();
   }

include/llvm/LTO/LTO.h

@@ -400,6 +400,9 @@ private:
   Error runThinLTO(AddStreamFn AddStream, NativeObjectCache Cache);
 
   mutable bool CalledGetMaxTasks = false;
+
+  // Use Optional to distinguish false from not yet initialized.
+  Optional<bool> EnableSplitLTOUnit;
 };
 
 /// The resolution for a symbol. The linker must provide a SymbolResolution for

lib/Analysis/ModuleSummaryAnalysis.cpp

@@ -457,7 +457,11 @@ ModuleSummaryIndex llvm::buildModuleSummaryIndex(
     std::function<BlockFrequencyInfo *(const Function &F)> GetBFICallback,
     ProfileSummaryInfo *PSI) {
   assert(PSI);
-  ModuleSummaryIndex Index(/*HaveGVs=*/true);
+  bool EnableSplitLTOUnit = false;
+  if (auto *MD = mdconst::extract_or_null<ConstantInt>(
+          M.getModuleFlag("EnableSplitLTOUnit")))
+    EnableSplitLTOUnit = MD->getZExtValue();
+  ModuleSummaryIndex Index(/*HaveGVs=*/true, EnableSplitLTOUnit);
 
   // Identify the local values in the llvm.used and llvm.compiler.used sets,
   // which should not be exported as they would then require renaming and

lib/Bitcode/Reader/BitcodeReader.cpp

@@ -5294,18 +5294,30 @@ Error ModuleSummaryIndexBitcodeReader::parseEntireSummary(unsigned ID) {
       break;
     case bitc::FS_FLAGS: {  // [flags]
       uint64_t Flags = Record[0];
-      // Scan flags (set only on the combined index).
+      // Scan flags.
-      assert(Flags <= 0x3 && "Unexpected bits in flag");
+      assert(Flags <= 0x1f && "Unexpected bits in flag");
 
       // 1 bit: WithGlobalValueDeadStripping flag.
+      // Set on combined index only.
       if (Flags & 0x1)
         TheIndex.setWithGlobalValueDeadStripping();
       // 1 bit: SkipModuleByDistributedBackend flag.
+      // Set on combined index only.
       if (Flags & 0x2)
         TheIndex.setSkipModuleByDistributedBackend();
       // 1 bit: HasSyntheticEntryCounts flag.
+      // Set on combined index only.
       if (Flags & 0x4)
         TheIndex.setHasSyntheticEntryCounts();
+      // 1 bit: DisableSplitLTOUnit flag.
+      // Set on per module indexes. It is up to the client to validate
+      // the consistency of this flag across modules being linked.
+      if (Flags & 0x8)
+        TheIndex.setEnableSplitLTOUnit();
+      // 1 bit: PartiallySplitLTOUnits flag.
+      // Set on combined index only.
+      if (Flags & 0x10)
+        TheIndex.setPartiallySplitLTOUnits();
       break;
     }
     case bitc::FS_VALUE_GUID: { // [valueid, refguid]
@@ -5917,6 +5929,46 @@ Expected<std::unique_ptr<ModuleSummaryIndex>> BitcodeModule::getSummary() {
   return std::move(Index);
 }
 
+static Expected<bool> getEnableSplitLTOUnitFlag(BitstreamCursor &Stream,
+                                                unsigned ID) {
+  if (Stream.EnterSubBlock(ID))
+    return error("Invalid record");
+  SmallVector<uint64_t, 64> Record;
+
+  while (true) {
+    BitstreamEntry Entry = Stream.advanceSkippingSubblocks();
+
+    switch (Entry.Kind) {
+    case BitstreamEntry::SubBlock: // Handled for us already.
+    case BitstreamEntry::Error:
+      return error("Malformed block");
+    case BitstreamEntry::EndBlock:
+      // If no flags record found, conservatively return true to mimic
+      // behavior before this flag was added.
+      return true;
+    case BitstreamEntry::Record:
+      // The interesting case.
+      break;
+    }
+
+    // Look for the FS_FLAGS record.
+    Record.clear();
+    auto BitCode = Stream.readRecord(Entry.ID, Record);
+    switch (BitCode) {
+    default: // Default behavior: ignore.
+      break;
+    case bitc::FS_FLAGS: { // [flags]
+      uint64_t Flags = Record[0];
+      // Scan flags.
+      assert(Flags <= 0x1f && "Unexpected bits in flag");
+
+      return Flags & 0x8;
+    }
+    }
+  }
+  llvm_unreachable("Exit infinite loop");
+}
+
 // Check if the given bitcode buffer contains a global value summary block.
 Expected<BitcodeLTOInfo> BitcodeModule::getLTOInfo() {
   BitstreamCursor Stream(Buffer);
@@ -5932,14 +5984,27 @@ Expected<BitcodeLTOInfo> BitcodeModule::getLTOInfo() {
     case BitstreamEntry::Error:
       return error("Malformed block");
     case BitstreamEntry::EndBlock:
-      return BitcodeLTOInfo{/*IsThinLTO=*/false, /*HasSummary=*/false};
+      return BitcodeLTOInfo{/*IsThinLTO=*/false, /*HasSummary=*/false,
+                            /*EnableSplitLTOUnit=*/false};
 
     case BitstreamEntry::SubBlock:
-      if (Entry.ID == bitc::GLOBALVAL_SUMMARY_BLOCK_ID)
+      if (Entry.ID == bitc::GLOBALVAL_SUMMARY_BLOCK_ID) {
-        return BitcodeLTOInfo{/*IsThinLTO=*/true, /*HasSummary=*/true};
+        Expected<bool> EnableSplitLTOUnit =
+            getEnableSplitLTOUnitFlag(Stream, Entry.ID);
+        if (!EnableSplitLTOUnit)
+          return EnableSplitLTOUnit.takeError();
+        return BitcodeLTOInfo{/*IsThinLTO=*/true, /*HasSummary=*/true,
+                              *EnableSplitLTOUnit};
+      }
 
-      if (Entry.ID == bitc::FULL_LTO_GLOBALVAL_SUMMARY_BLOCK_ID)
+      if (Entry.ID == bitc::FULL_LTO_GLOBALVAL_SUMMARY_BLOCK_ID) {
-        return BitcodeLTOInfo{/*IsThinLTO=*/false, /*HasSummary=*/true};
+        Expected<bool> EnableSplitLTOUnit =
+            getEnableSplitLTOUnitFlag(Stream, Entry.ID);
+        if (!EnableSplitLTOUnit)
+          return EnableSplitLTOUnit.takeError();
+        return BitcodeLTOInfo{/*IsThinLTO=*/false, /*HasSummary=*/true,
+                              *EnableSplitLTOUnit};
+      }
 
       // Ignore other sub-blocks.
       if (Stream.SkipBlock())

lib/Bitcode/Writer/BitcodeWriter.cpp

@@ -3618,6 +3618,13 @@ void ModuleBitcodeWriterBase::writePerModuleGlobalValueSummary() {
 
   Stream.EmitRecord(bitc::FS_VERSION, ArrayRef<uint64_t>{INDEX_VERSION});
 
+  // Write the index flags.
+  uint64_t Flags = 0;
+  // Bits 1-3 are set only in the combined index, skip them.
+  if (Index->enableSplitLTOUnit())
+    Flags |= 0x8;
+  Stream.EmitRecord(bitc::FS_FLAGS, ArrayRef<uint64_t>{Flags});
+
   if (Index->begin() == Index->end()) {
     Stream.ExitBlock();
     return;
@@ -3734,6 +3741,10 @@ void IndexBitcodeWriter::writeCombinedGlobalValueSummary() {
     Flags |= 0x2;
   if (Index.hasSyntheticEntryCounts())
     Flags |= 0x4;
+  if (Index.enableSplitLTOUnit())
+    Flags |= 0x8;
+  if (Index.partiallySplitLTOUnits())
+    Flags |= 0x10;
   Stream.EmitRecord(bitc::FS_FLAGS, ArrayRef<uint64_t>{Flags});
 
   for (const auto &GVI : valueIds()) {

lib/LTO/LTO.cpp

@@ -546,6 +546,15 @@ Error LTO::addModule(InputFile &Input, unsigned ModI,
   if (!LTOInfo)
     return LTOInfo.takeError();
 
+  if (EnableSplitLTOUnit.hasValue()) {
+    // If only some modules were split, flag this in the index so that
+    // we can skip or error on optimizations that need consistently split
+    // modules (whole program devirt and lower type tests).
+    if (EnableSplitLTOUnit.getValue() != LTOInfo->EnableSplitLTOUnit)
+      ThinLTO.CombinedIndex.setPartiallySplitLTOUnits();
+  } else
+    EnableSplitLTOUnit = LTOInfo->EnableSplitLTOUnit;
+
   BitcodeModule BM = Input.Mods[ModI];
   auto ModSyms = Input.module_symbols(ModI);
   addModuleToGlobalRes(ModSyms, {ResI, ResE},

lib/Transforms/IPO/LowerTypeTests.cpp

@@ -1702,6 +1702,13 @@ bool LowerTypeTestsModule::lower() {
       !ExportSummary && !ImportSummary)
     return false;
 
+  // If only some of the modules were split, we cannot correctly handle
+  // code that contains type tests.
+  if (TypeTestFunc && !TypeTestFunc->use_empty() &&
+      ((ExportSummary && ExportSummary->partiallySplitLTOUnits()) ||
+       (ImportSummary && ImportSummary->partiallySplitLTOUnits())))
+    report_fatal_error("inconsistent LTO Unit splitting with llvm.type.test");
+
   if (ImportSummary) {
     if (TypeTestFunc) {
       for (auto UI = TypeTestFunc->use_begin(), UE = TypeTestFunc->use_end();

lib/Transforms/IPO/ThinLTOBitcodeWriter.cpp

@@ -418,8 +418,18 @@ void splitAndWriteThinLTOBitcode(
   }
 }
 
-// Returns whether this module needs to be split because it uses type metadata.
+// Returns whether this module needs to be split because splitting is
+// enabled and it uses type metadata.
 bool requiresSplit(Module &M) {
+  // First check if the LTO Unit splitting has been enabled.
+  bool EnableSplitLTOUnit = false;
+  if (auto *MD = mdconst::extract_or_null<ConstantInt>(
+          M.getModuleFlag("EnableSplitLTOUnit")))
+    EnableSplitLTOUnit = MD->getZExtValue();
+  if (!EnableSplitLTOUnit)
+    return false;
+
+  // Module only needs to be split if it contains type metadata.
   for (auto &GO : M.global_objects()) {
     if (GO.hasMetadata(LLVMContext::MD_type))
       return true;
@@ -431,7 +441,7 @@ bool requiresSplit(Module &M) {
 void writeThinLTOBitcode(raw_ostream &OS, raw_ostream *ThinLinkOS,
                          function_ref<AAResults &(Function &)> AARGetter,
                          Module &M, const ModuleSummaryIndex *Index) {
-  // See if this module has any type metadata. If so, we need to split it.
+  // Split module if splitting is enabled and it contains any type metadata.
   if (requiresSplit(M))
     return splitAndWriteThinLTOBitcode(OS, ThinLinkOS, AARGetter, M);
 

lib/Transforms/IPO/WholeProgramDevirt.cpp

@@ -1563,6 +1563,17 @@ bool DevirtModule::run() {
       M.getFunction(Intrinsic::getName(Intrinsic::type_checked_load));
   Function *AssumeFunc = M.getFunction(Intrinsic::getName(Intrinsic::assume));
 
+  // If only some of the modules were split, we cannot correctly handle
+  // code that contains type tests or type checked loads.
+  if ((ExportSummary && ExportSummary->partiallySplitLTOUnits()) ||
+      (ImportSummary && ImportSummary->partiallySplitLTOUnits())) {
+    if ((TypeTestFunc && !TypeTestFunc->use_empty()) ||
+        (TypeCheckedLoadFunc && !TypeCheckedLoadFunc->use_empty()))
+      report_fatal_error("inconsistent LTO Unit splitting with llvm.type.test "
+                         "or llvm.type.checked.load");
+    return false;
+  }
+
   // Normally if there are no users of the devirtualization intrinsics in the
   // module, this pass has nothing to do. But if we are exporting, we also need
   // to handle any users that appear only in the function summaries.

test/Bitcode/thinlto-alias.ll

@@ -18,6 +18,7 @@
 ; CHECK-NEXT: <FUNCTION op0=4 op1=7
 ; CHECK:       <GLOBALVAL_SUMMARY_BLOCK
 ; CHECK-NEXT:    <VERSION
+; CHECK-NEXT:    <FLAGS
 ; See if the call to func is registered.
 ; The value id 1 matches the second FUNCTION record above.
 ; CHECK-NEXT:    <PERMODULE {{.*}} op6=1/>

test/Bitcode/thinlto-alias2.ll

@@ -4,6 +4,7 @@
 
 ; CHECK:       <GLOBALVAL_SUMMARY_BLOCK
 ; CHECK-NEXT:    <VERSION
+; CHECK-NEXT:    <FLAGS
 ; CHECK-NEXT:    <PERMODULE {{.*}} op4=0 op5=0 op6=[[ALIASID:[0-9]+]]/>
 ; CHECK-NEXT:    <PERMODULE {{.*}} op0=[[ALIASEEID:[0-9]+]]
 ; CHECK-NEXT:    <ALIAS {{.*}} op0=[[ALIASID]] {{.*}} op2=[[ALIASEEID]]/>

test/Bitcode/thinlto-function-summary-callgraph-cast.ll

@@ -5,6 +5,7 @@
 
 ; CHECK:       <GLOBALVAL_SUMMARY_BLOCK
 ; CHECK-NEXT:    <VERSION
+; CHECK-NEXT:    <FLAGS
 ; "op7" is a call to "callee" function.
 ; CHECK-NEXT:    <PERMODULE {{.*}} op8=3 op9=[[ALIASID:[0-9]+]]/>
 ; "another_caller" has only references but no calls.

test/Bitcode/thinlto-function-summary-callgraph-pgo.ll

@@ -16,6 +16,7 @@
 ; CHECK-NEXT: <FUNCTION op0=4 op1=4
 ; CHECK:       <GLOBALVAL_SUMMARY_BLOCK
 ; CHECK-NEXT:    <VERSION
+; CHECK-NEXT:    <FLAGS
 ; See if the call to func is registered, using the expected hotness type.
 ; CHECK-NEXT:    <PERMODULE_PROFILE {{.*}} op6=1 op7=2/>
 ; CHECK-NEXT:  </GLOBALVAL_SUMMARY_BLOCK>

test/Bitcode/thinlto-function-summary-callgraph-profile-summary.ll

@@ -46,6 +46,7 @@
 ; CHECK-NEXT: <FUNCTION op0=42 op1=5
 ; CHECK-LABEL:       <GLOBALVAL_SUMMARY_BLOCK
 ; CHECK-NEXT:    <VERSION
+; CHECK-NEXT:    <FLAGS
 ; CHECK-NEXT:    <VALUE_GUID op0=25 op1=123/>
 ; op4=hot1 op6=cold op8=hot2 op10=hot4 op12=none1 op14=hot3 op16=none2 op18=none3 op20=123
 ; CHECK-NEXT:    <PERMODULE_PROFILE {{.*}} op6=1 op7=3 op8=5 op9=1 op10=2 op11=3 op12=4 op13=1 op14=6 op15=2 op16=3 op17=3 op18=7 op19=2 op20=8 op21=2 op22=25 op23=4/>

test/Bitcode/thinlto-function-summary-callgraph-relbf.ll

@@ -12,6 +12,7 @@
 ; CHECK-NEXT: <FUNCTION op0=17 op1=4
 ; CHECK:       <GLOBALVAL_SUMMARY_BLOCK
 ; CHECK-NEXT:    <VERSION
+; CHECK-NEXT:    <FLAGS
 ; See if the call to func is registered.
 ; CHECK-NEXT:    <PERMODULE_RELBF {{.*}} op4=1 {{.*}} op8=256
 ; CHECK-NEXT:  </GLOBALVAL_SUMMARY_BLOCK>

test/Bitcode/thinlto-function-summary-callgraph-sample-profile-summary.ll

@@ -29,6 +29,7 @@
 
 ; CHECK-LABEL:       <GLOBALVAL_SUMMARY_BLOCK
 ; CHECK-NEXT:    <VERSION
+; CHECK-NEXT:    <FLAGS
 ; CHECK-NEXT:    <VALUE_GUID op0=26 op1=123/>
 ; op4=none1 op6=hot1 op8=cold1 op10=none2 op12=hot2 op14=cold2 op16=none3 op18=hot3 op20=cold3 op22=123
 ; CHECK-NEXT:    <PERMODULE_PROFILE {{.*}} op6=7 op7=0 op8=1 op9=3 op10=4 op11=1 op12=8 op13=0 op14=2 op15=3 op16=5 op17=1 op18=9 op19=0 op20=3 op21=3 op22=6 op23=1 op24=26 op25=4/>

test/Bitcode/thinlto-function-summary-callgraph.ll

@@ -17,6 +17,7 @@
 ; CHECK-NEXT: <FUNCTION op0=17 op1=4
 ; CHECK:       <GLOBALVAL_SUMMARY_BLOCK
 ; CHECK-NEXT:    <VERSION
+; CHECK-NEXT:    <FLAGS
 ; See if the call to func is registered
 ; CHECK-NEXT:    <PERMODULE {{.*}} op4=1
 ; CHECK-NEXT:  </GLOBALVAL_SUMMARY_BLOCK>

test/Bitcode/thinlto-function-summary.ll

@@ -19,6 +19,7 @@
 ; BC-NEXT: <ALIAS op0=67 op1=1
 ; BC: <GLOBALVAL_SUMMARY_BLOCK
 ; BC-NEXT: <VERSION
+; BC-NEXT: <FLAGS
 ; BC-NEXT: <PERMODULE {{.*}} op0=1 op1=0
 ; BC-NEXT: <PERMODULE {{.*}} op0=2 op1=0
 ; BC-NEXT: <PERMODULE {{.*}} op0=3 op1=7

test/LTO/Resolution/X86/export-jumptable.ll

@@ -2,7 +2,7 @@
 ; the full LTO object file; any such functions will be referenced by the jump
 ; table.
 
-; RUN: opt -thinlto-bc -o %t %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t %s
 ; RUN: llvm-lto2 run -o %t2 -r %t,f1,p -r %t,f2,p -r %t,_start,px %t -save-temps
 ; RUN: llvm-dis %t2.1.2.internalize.bc -o - | FileCheck %s
 

test/LTO/Resolution/X86/local-def-dllimport.ll

@@ -1,4 +1,4 @@
-; RUN: opt -thinlto-bc -o %t0.bc %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t0.bc %s
 ; RUN: llvm-lto2 run -r %t0.bc,__imp_f,l \
 ; RUN:               -r %t0.bc,g,p \
 ; RUN:               -r %t0.bc,g,l \

test/LTO/Resolution/X86/lowertypetests.ll

@@ -1,4 +1,4 @@
-; RUN: opt -thinlto-bc -o %t %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t %s
 ; RUN: llvm-lto2 run -r %t,f,plx -r %t,g_alias,plx -r %t,foo,lx -r %t,foo,plx -r %t,bar,lx -r %t,bar,plx -o %t1 %t
 ; RUN: llvm-nm %t1.0 | FileCheck --check-prefix=MERGED %s
 ; RUN: llvm-nm %t1.1 | FileCheck %s

test/LTO/Resolution/X86/lto-unit-check.ll

@@ -0,0 +1,55 @@
+; Test to ensure that the Enable Split LTO Unit flag is set properly in the
+; summary, and that we correctly silently handle linking bitcode files with
+; different values of this flag.
+
+; Linking bitcode both with EnableSplitLTOUnit set should work
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t1 %s
+; RUN: llvm-bcanalyzer -dump %t1 | FileCheck %s --check-prefix=SPLITLTOUNIT
+; RUN: llvm-dis -o - %t1 | FileCheck %s --check-prefix=ENABLESPLITFLAG
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t2 %s
+; RUN: llvm-bcanalyzer -dump %t2 | FileCheck %s --check-prefix=SPLITLTOUNIT
+; RUN: llvm-dis -o - %t2 | FileCheck %s --check-prefix=ENABLESPLITFLAG
+; RUN: llvm-lto2 run -o %t3 %t1 %t2
+
+; Linking bitcode both without EnableSplitLTOUnit set should work
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit=false -o %t1 %s
+; RUN: llvm-bcanalyzer -dump %t1 | FileCheck %s --check-prefix=NOSPLITLTOUNIT
+; RUN: llvm-dis -o - %t1 | FileCheck %s --check-prefix=NOENABLESPLITFLAG
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit=false -o %t2 %s
+; RUN: llvm-bcanalyzer -dump %t2 | FileCheck %s --check-prefix=NOSPLITLTOUNIT
+; RUN: llvm-dis -o - %t2 | FileCheck %s --check-prefix=NOENABLESPLITFLAG
+; RUN: llvm-lto2 run -o %t3 %t1 %t2
+
+; Linking bitcode with different values of EnableSplitLTOUnit should succeed
+; (silently skipping any optimizations like whole program devirt that rely
+; on all modules being split).
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t1 %s
+; RUN: llvm-bcanalyzer -dump %t1 | FileCheck %s --check-prefix=SPLITLTOUNIT
+; RUN: llvm-dis -o - %t1 | FileCheck %s --check-prefix=ENABLESPLITFLAG
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit=false -o %t2 %s
+; RUN: llvm-bcanalyzer -dump %t2 | FileCheck %s --check-prefix=NOSPLITLTOUNIT
+; RUN: llvm-dis -o - %t2 | FileCheck %s --check-prefix=NOENABLESPLITFLAG
+; RUN: llvm-lto2 run -o %t3 %t1 %t2
+
+; Linking bitcode with different values of EnableSplitLTOUnit (reverse order)
+; should succeed (silently skipping any optimizations like whole program devirt
+; that rely on all modules being split).
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit=false -o %t1 %s
+; RUN: llvm-bcanalyzer -dump %t1 | FileCheck %s --check-prefix=NOSPLITLTOUNIT
+; RUN: llvm-dis -o - %t1 | FileCheck %s --check-prefix=NOENABLESPLITFLAG
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t2 %s
+; RUN: llvm-bcanalyzer -dump %t2 | FileCheck %s --check-prefix=SPLITLTOUNIT
+; RUN: llvm-dis -o - %t2 | FileCheck %s --check-prefix=ENABLESPLITFLAG
+; RUN: llvm-lto2 run -o %t3 %t1 %t2
+
+; The flag should be set when splitting is disabled (for backwards compatibility
+; with older bitcode where it was always enabled).
+; SPLITLTOUNIT: <FLAGS op0=8/>
+; NOSPLITLTOUNIT: <FLAGS op0=0/>
+
+; Check that the corresponding module flag is set when expected.
+; ENABLESPLITFLAG: !{i32 1, !"EnableSplitLTOUnit", i32 1}
+; NOENABLESPLITFLAG-NOT: !{i32 1, !"EnableSplitLTOUnit", i32 1}
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"

test/ThinLTO/X86/cache-icall.ll

@@ -4,8 +4,8 @@
 ; This affects code generated for any users of f(). Make sure that we don't pull a stale object
 ; file for %t.o from the cache.
 
-; RUN: opt -module-hash -module-summary -thinlto-bc %s -o %t.bc
+; RUN: opt -module-hash -module-summary -thinlto-bc -thinlto-split-lto-unit %s -o %t.bc
-; RUN: opt -module-hash -module-summary -thinlto-bc %p/Inputs/cache-icall.ll -o %t2.bc
+; RUN: opt -module-hash -module-summary -thinlto-bc -thinlto-split-lto-unit %p/Inputs/cache-icall.ll -o %t2.bc
 
 ; RUN: rm -Rf %t.cache && mkdir %t.cache
 

test/ThinLTO/X86/cfi-devirt.ll

@@ -2,7 +2,7 @@
 
 ; Test CFI devirtualization through the thin link and backend.
 
-; RUN: opt -thinlto-bc -o %t.o %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t.o %s
 
 ; Legacy PM
 ; FIXME: Fix machine verifier issues and remove -verify-machineinstrs=0. PR39436.
@@ -44,6 +44,27 @@
 
 ; REMARK: single-impl: devirtualized a call to _ZN1A1nEi
 
+; Next check that we emit an error when trying to LTO link this module
+; containing an llvm.type.checked.load (with a split LTO Unit) with one
+; that does not have a split LTO Unit.
+; RUN: opt -thinlto-bc -o %t2.o %S/Inputs/empty.ll
+; RUN: not llvm-lto2 run %t.o %t2.o -save-temps -pass-remarks=. \
+; RUN:   -verify-machineinstrs=0 \
+; RUN:   -o %t3 \
+; RUN:   -r=%t.o,test,px \
+; RUN:   -r=%t.o,_ZN1A1nEi,p \
+; RUN:   -r=%t.o,_ZN1B1fEi,p \
+; RUN:   -r=%t.o,_ZN1C1fEi,p \
+; RUN:   -r=%t.o,empty,p \
+; RUN:   -r=%t.o,_ZTV1B, \
+; RUN:   -r=%t.o,_ZTV1C, \
+; RUN:   -r=%t.o,_ZN1A1nEi, \
+; RUN:   -r=%t.o,_ZN1B1fEi, \
+; RUN:   -r=%t.o,_ZN1C1fEi, \
+; RUN:   -r=%t.o,_ZTV1B,px \
+; RUN:   -r=%t.o,_ZTV1C,px 2>&1 | FileCheck %s --check-prefix=ERROR
+; ERROR: LLVM ERROR: inconsistent LTO Unit splitting with llvm.type.test or llvm.type.checked.load
+
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64-grtev4-linux-gnu"
 

test/ThinLTO/X86/cfi-distributed.ll

@@ -3,8 +3,8 @@
 ; Test to ensure that only referenced type ID records are emitted into
 ; each distributed index file.
 
-; RUN: opt -thinlto-bc -o %t1.o %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t1.o %s
-; RUN: opt -thinlto-bc -o %t2.o %p/Inputs/cfi-distributed.ll
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t2.o %p/Inputs/cfi-distributed.ll
 
 ; RUN: llvm-lto2 run -thinlto-distributed-indexes %t1.o %t2.o \
 ; RUN:   -o %t3 \

test/ThinLTO/X86/cfi-icall.ll

@@ -1,4 +1,4 @@
-; RUN: opt -thinlto-bc %s -o %t1.bc
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit %s -o %t1.bc
 ; RUN: llvm-lto2 run  -thinlto-distributed-indexes %t1.bc -o %t.out -save-temps \
 ; RUN:   -r %t1.bc,foo,plx \
 ; RUN:   -r %t1.bc,bar,x \

test/ThinLTO/X86/cfi.ll

@@ -2,7 +2,7 @@
 
 ; Test CFI through the thin link and backend.
 
-; RUN: opt -thinlto-bc -o %t.o %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t.o %s
 
 ; Legacy PM
 ; RUN: llvm-lto2 run -save-temps %t.o \

test/ThinLTO/X86/devirt-after-icp.ll

@@ -42,7 +42,7 @@
 ; will use the same vtable pointer. Without a dominance check, we could
 ; incorrectly devirtualize a->foo() to B::foo();
 
-; RUN: opt -thinlto-bc -o %t.o %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t.o %s
 
 ; Legacy PM
 ; FIXME: Fix machine verifier issues and remove -verify-machineinstrs=0. PR39436.

test/Transforms/ThinLTOBitcodeWriter/circular-reference.ll

@@ -1,4 +1,4 @@
-; RUN: opt -thinlto-bc -o %t %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t %s
 ; RUN: llvm-modextract -b -n 0 -o - %t | llvm-dis | FileCheck --check-prefix=M0 %s
 ; RUN: llvm-modextract -b -n 1 -o - %t | llvm-dis | FileCheck --check-prefix=M1 %s
 

test/Transforms/ThinLTOBitcodeWriter/comdat.ll

@@ -1,4 +1,4 @@
-; RUN: opt -thinlto-bc -o %t %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t %s
 ; RUN: llvm-modextract -n 0 -o - %t | llvm-dis | FileCheck --check-prefix=THIN %s
 ; RUN: llvm-modextract -n 1 -o - %t | llvm-dis | FileCheck --check-prefix=MERGED %s
 

test/Transforms/ThinLTOBitcodeWriter/filter-alias.ll

@@ -1,4 +1,4 @@
-; RUN: opt -thinlto-bc -o %t %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t %s
 ; RUN: llvm-modextract -n 0 -o - %t | llvm-dis | FileCheck --check-prefix=CHECK0 %s
 ; RUN: llvm-modextract -n 1 -o - %t | llvm-dis | FileCheck --check-prefix=CHECK1 %s
 ; CHECK0-NOT: @{{.*}}anon{{.*}}=

test/Transforms/ThinLTOBitcodeWriter/function-alias.ll

@@ -1,4 +1,4 @@
-; RUN: opt -thinlto-bc -o %t %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t %s
 ; RUN: llvm-modextract -n 1 -o - %t | llvm-dis | FileCheck --check-prefix=CHECK1 %s
 
 target triple = "x86_64-unknown-linux-gnu"

test/Transforms/ThinLTOBitcodeWriter/pr33536.ll

@@ -1,7 +1,7 @@
 ; Test for a bug specific to the new pass manager where we may build a domtree
 ; to make more precise AA queries for functions.
 ;
-; RUN: opt -aa-pipeline=default -passes='no-op-module' -debug-pass-manager -thinlto-bc -o %t %s
+; RUN: opt -aa-pipeline=default -passes='no-op-module' -debug-pass-manager -thinlto-bc -thinlto-split-lto-unit -o %t %s
 ; RUN: llvm-modextract -b -n 0 -o - %t | llvm-dis | FileCheck --check-prefix=M0 %s
 ; RUN: llvm-modextract -b -n 1 -o - %t | llvm-dis | FileCheck --check-prefix=M1 %s
 

test/Transforms/ThinLTOBitcodeWriter/split-internal-typeid.ll

@@ -1,4 +1,4 @@
-; RUN: opt -thinlto-bc -o %t %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t %s
 ; RUN: llvm-modextract -b -n 0 -o %t0 %t
 ; RUN: llvm-modextract -b -n 1 -o %t1 %t
 ; RUN: not llvm-modextract -b -n 2 -o - %t 2>&1 | FileCheck --check-prefix=ERROR %s

test/Transforms/ThinLTOBitcodeWriter/split-internal1.ll

@@ -1,4 +1,4 @@
-; RUN: opt -thinlto-bc -o %t %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t %s
 ; RUN: llvm-modextract -b -n 0 -o %t0 %t
 ; RUN: llvm-modextract -b -n 1 -o %t1 %t
 ; RUN: not llvm-modextract -b -n 2 -o - %t 2>&1 | FileCheck --check-prefix=ERROR %s

test/Transforms/ThinLTOBitcodeWriter/split-internal2.ll

@@ -1,4 +1,4 @@
-; RUN: opt -thinlto-bc -o %t %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t %s
 ; RUN: llvm-modextract -b -n 0 -o %t0 %t
 ; RUN: llvm-modextract -b -n 1 -o %t1 %t
 ; RUN: not llvm-modextract -b -n 2 -o - %t 2>&1 | FileCheck --check-prefix=ERROR %s

test/Transforms/ThinLTOBitcodeWriter/split-vfunc-internal.ll

@@ -1,4 +1,4 @@
-; RUN: opt -thinlto-bc -o %t %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t %s
 ; RUN: llvm-modextract -b -n 0 -o - %t | llvm-dis | FileCheck --check-prefix=M0 %s
 ; RUN: llvm-modextract -b -n 1 -o - %t | llvm-dis | FileCheck --check-prefix=M1 %s
 

test/Transforms/ThinLTOBitcodeWriter/split-vfunc.ll

@@ -1,4 +1,4 @@
-; RUN: opt -thinlto-bc -o %t %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t %s
 ; RUN: llvm-modextract -b -n 0 -o - %t | llvm-dis | FileCheck --check-prefix=M0 %s
 ; RUN: llvm-modextract -b -n 1 -o - %t | llvm-dis | FileCheck --check-prefix=M1 %s
 

test/Transforms/ThinLTOBitcodeWriter/split.ll

@@ -1,6 +1,6 @@
 ; Generate bitcode files with summary, as well as minimized bitcode without
 ; the debug metadata for the thin link.
-; RUN: opt -thinlto-bc -thin-link-bitcode-file=%t2 -o %t %s
+; RUN: opt -thinlto-bc -thin-link-bitcode-file=%t2 -thinlto-split-lto-unit -o %t %s
 ; RUN: llvm-modextract -b -n 0 -o %t0.bc %t
 ; RUN: llvm-modextract -b -n 1 -o %t1.bc %t
 ; RUN: llvm-modextract -b -n 0 -o %t0.thinlink.bc %t2

test/Transforms/ThinLTOBitcodeWriter/symver.ll

@@ -1,4 +1,4 @@
-; RUN: opt -thinlto-bc -o %t %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t %s
 ; RUN: llvm-modextract -n 1 -o - %t | llvm-dis | FileCheck %s
 
 ; The target assembly parser is required to parse the symver directives

test/Transforms/ThinLTOBitcodeWriter/unsplittable.ll

@@ -1,4 +1,4 @@
-; RUN: opt -thinlto-bc -thin-link-bitcode-file=%t2 -o %t %s
+; RUN: opt -thinlto-bc -thin-link-bitcode-file=%t2 -thinlto-split-lto-unit -o %t %s
 ; RUN: llvm-dis -o - %t | FileCheck %s
 ; RUN: llvm-bcanalyzer -dump %t | FileCheck --check-prefix=BCA %s
 ; When not splitting the module, the thin link bitcode file should simply be a
@@ -28,7 +28,8 @@ define void @h() comdat {
   ret void
 }
 
-; CHECK: !llvm.module.flags = !{![[FLAG:[0-9]+]]}
+; CHECK: !llvm.module.flags = !{![[FLAG1:[0-9]+]], ![[FLAG2:[0-9]+]]}
-; CHECK: ![[FLAG]] = !{i32 1, !"ThinLTO", i32 0}
+; CHECK: ![[FLAG1]] = !{i32 1, !"EnableSplitLTOUnit", i32 1}
+; CHECK: ![[FLAG2]] = !{i32 1, !"ThinLTO", i32 0}
 
 !0 = !{i32 0, !"typeid"}

test/Transforms/ThinLTOBitcodeWriter/x86/module-asm.ll

@@ -1,4 +1,4 @@
-; RUN: opt -thinlto-bc -o %t %s
+; RUN: opt -thinlto-bc -thinlto-split-lto-unit -o %t %s
 ; RUN: llvm-modextract -b -n 0 -o - %t | llvm-dis | FileCheck --check-prefix=M0 %s
 ; RUN: llvm-modextract -b -n 1 -o - %t | llvm-dis | FileCheck --check-prefix=M1 %s
 

tools/opt/opt.cpp

@@ -103,6 +103,10 @@ static cl::opt<bool>
     OutputThinLTOBC("thinlto-bc",
                     cl::desc("Write output as ThinLTO-ready bitcode"));
 
+static cl::opt<bool>
+    SplitLTOUnit("thinlto-split-lto-unit",
+                 cl::desc("Enable splitting of a ThinLTO LTOUnit"));
+
 static cl::opt<std::string> ThinLinkBitcodeFile(
     "thin-link-bitcode-file", cl::value_desc("filename"),
     cl::desc(
@@ -596,6 +600,9 @@ int main(int argc, char **argv) {
     if (CheckBitcodeOutputToConsole(Out->os(), !Quiet))
       NoOutput = true;
 
+  if (OutputThinLTOBC)
+    M->addModuleFlag(Module::Error, "EnableSplitLTOUnit", SplitLTOUnit);
+
   if (PassPipeline.getNumOccurrences() > 0) {
     OutputKind OK = OK_NoOutput;
     if (!NoOutput)