[ThinLTO] Only emit referenced type id records in index files

Summary:
Currently all type ids are emitted into the index file when it is
written. For distributed ThinLTO, that meant that all type ids were
being duplicated into every single distributed index file, regardless of
whether they were referenced, leading to huge amounts of unnecessary
duplication and size bloat.

Keep track of the type id GUIDs actually referenced by the GV summary
records being emitted, and only emit those type IDs.

Add a new test, and fix test/Assembler/thinlto-summary.ll so that all
type ids are referenced to prevent deletion in that test.

Reviewers: pcc

Subscribers: mehdi_amini, inglorion, eraman, steven_wu, dexonsmith, vitalybuka, llvm-commits

Differential Revision: https://reviews.llvm.org/D49565

llvm-svn: 337503

4 files changed, 109 insertions, 7 deletions

diff --git a/llvm/lib/Bitcode/Writer/BitcodeWriter.cpp b/llvm/lib/Bitcode/Writer/BitcodeWriter.cpp
index 9072be68372..be75df0820d 100644
--- a/llvm/lib/Bitcode/Writer/BitcodeWriter.cpp
+++ b/llvm/lib/Bitcode/Writer/BitcodeWriter.cpp
@@ -3353,10 +3353,14 @@ void IndexBitcodeWriter::writeModStrings() {
 
 /// Write the function type metadata related records that need to appear before
 /// a function summary entry (whether per-module or combined).
-static void writeFunctionTypeMetadataRecords(BitstreamWriter &Stream,
-                                             FunctionSummary *FS) {
-  if (!FS->type_tests().empty())
+static void writeFunctionTypeMetadataRecords(
+    BitstreamWriter &Stream, FunctionSummary *FS,
+    std::set<GlobalValue::GUID> &ReferencedTypeIds) {
+  if (!FS->type_tests().empty()) {
     Stream.EmitRecord(bitc::FS_TYPE_TESTS, FS->type_tests());
+    for (auto &TT : FS->type_tests())
+      ReferencedTypeIds.insert(TT);
+  }
 
   SmallVector<uint64_t, 64> Record;
 
@@ -3368,6 +3372,7 @@ static void writeFunctionTypeMetadataRecords(BitstreamWriter &Stream,
     for (auto &VF : VFs) {
       Record.push_back(VF.GUID);
       Record.push_back(VF.Offset);
+      ReferencedTypeIds.insert(VF.GUID);
     }
     Stream.EmitRecord(Ty, Record);
   };
@@ -3382,6 +3387,7 @@ static void writeFunctionTypeMetadataRecords(BitstreamWriter &Stream,
     for (auto &VC : VCs) {
       Record.clear();
       Record.push_back(VC.VFunc.GUID);
+      ReferencedTypeIds.insert(VC.VFunc.GUID);
       Record.push_back(VC.VFunc.Offset);
       Record.insert(Record.end(), VC.Args.begin(), VC.Args.end());
       Stream.EmitRecord(Ty, Record);
@@ -3447,7 +3453,8 @@ void ModuleBitcodeWriterBase::writePerModuleFunctionSummaryRecord(
   NameVals.push_back(ValueID);
 
   FunctionSummary *FS = cast<FunctionSummary>(Summary);
-  writeFunctionTypeMetadataRecords(Stream, FS);
+  std::set<GlobalValue::GUID> ReferencedTypeIds;
+  writeFunctionTypeMetadataRecords(Stream, FS, ReferencedTypeIds);
 
   NameVals.push_back(getEncodedGVSummaryFlags(FS->flags()));
   NameVals.push_back(FS->instCount());
@@ -3702,6 +3709,10 @@ void IndexBitcodeWriter::writeCombinedGlobalValueSummary() {
 
   SmallVector<uint64_t, 64> NameVals;
 
+  // Set that will be populated during call to writeFunctionTypeMetadataRecords
+  // with the type ids referenced by this index file.
+  std::set<GlobalValue::GUID> ReferencedTypeIds;
+
   // For local linkage, we also emit the original name separately
   // immediately after the record.
   auto MaybeEmitOriginalName = [&](GlobalValueSummary &S) {
@@ -3753,7 +3764,7 @@ void IndexBitcodeWriter::writeCombinedGlobalValueSummary() {
     }
 
     auto *FS = cast<FunctionSummary>(S);
-    writeFunctionTypeMetadataRecords(Stream, FS);
+    writeFunctionTypeMetadataRecords(Stream, FS, ReferencedTypeIds);
 
     NameVals.push_back(*ValueId);
     NameVals.push_back(Index.getModuleId(FS->modulePath()));
@@ -3862,6 +3873,9 @@ void IndexBitcodeWriter::writeCombinedGlobalValueSummary() {
 
   if (!Index.typeIds().empty()) {
     for (auto &S : Index.typeIds()) {
+      // Skip if not referenced in any GV summary within this index file.
+      if (!ReferencedTypeIds.count(GlobalValue::getGUID(S.first)))
+        continue;
       writeTypeIdSummaryRecord(NameVals, StrtabBuilder, S.first, S.second);
       Stream.EmitRecord(bitc::FS_TYPE_ID, NameVals);
       NameVals.clear();
diff --git a/llvm/test/Assembler/thinlto-summary.ll b/llvm/test/Assembler/thinlto-summary.ll
index 66a5bef015e..6722fb1a8b7 100644
--- a/llvm/test/Assembler/thinlto-summary.ll
+++ b/llvm/test/Assembler/thinlto-summary.ll
@@ -50,7 +50,7 @@
 ^20 = gv: (guid: 19, summaries: (function: (module: ^0, flags: (linkage: external, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 8, typeIdInfo: (typeTestAssumeVCalls: (vFuncId: (^27, offset: 16))))))
 ^21 = gv: (guid: 20, summaries: (function: (module: ^0, flags: (linkage: external, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 5, typeIdInfo: (typeCheckedLoadVCalls: (vFuncId: (^25, offset: 16))))))
 ^22 = gv: (guid: 21, summaries: (function: (module: ^0, flags: (linkage: external, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 15, typeIdInfo: (typeTestAssumeConstVCalls: (vFuncId: (^27, offset: 16), args: (42), vFuncId: (^27, offset: 24), args: (43))))))
-^23 = gv: (guid: 22, summaries: (function: (module: ^0, flags: (linkage: external, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 5, typeIdInfo: (typeCheckedLoadConstVCalls: (vFuncId: (^27, offset: 16), args: (42))))))
+^23 = gv: (guid: 22, summaries: (function: (module: ^0, flags: (linkage: external, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 5, typeIdInfo: (typeCheckedLoadConstVCalls: (vFuncId: (^28, offset: 16), args: (42))))))
 
 ; Test TypeId summaries:
 
@@ -88,7 +88,7 @@
 ; CHECK: ^20 = gv: (guid: 19, summaries: (function: (module: ^0, flags: (linkage: external, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 8, typeIdInfo: (typeTestAssumeVCalls: (vFuncId: (^27, offset: 16))))))
 ; CHECK: ^21 = gv: (guid: 20, summaries: (function: (module: ^0, flags: (linkage: external, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 5, typeIdInfo: (typeCheckedLoadVCalls: (vFuncId: (^25, offset: 16))))))
 ; CHECK: ^22 = gv: (guid: 21, summaries: (function: (module: ^0, flags: (linkage: external, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 15, typeIdInfo: (typeTestAssumeConstVCalls: (vFuncId: (^27, offset: 16), args: (42), vFuncId: (^27, offset: 24), args: (43))))))
-; CHECK: ^23 = gv: (guid: 22, summaries: (function: (module: ^0, flags: (linkage: external, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 5, typeIdInfo: (typeCheckedLoadConstVCalls: (vFuncId: (^27, offset: 16), args: (42))))))
+; CHECK: ^23 = gv: (guid: 22, summaries: (function: (module: ^0, flags: (linkage: external, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 5, typeIdInfo: (typeCheckedLoadConstVCalls: (vFuncId: (^28, offset: 16), args: (42))))))
 ; CHECK: ^24 = typeid: (name: "_ZTS1A", summary: (typeTestRes: (kind: allOnes, sizeM1BitWidth: 7), wpdResolutions: ((offset: 0, wpdRes: (kind: branchFunnel)), (offset: 8, wpdRes: (kind: singleImpl, singleImplName: "_ZN1A1nEi")), (offset: 16, wpdRes: (kind: indir, resByArg: (args: (1, 2), byArg: (kind: indir, byte: 2, bit: 3), args: (3), byArg: (kind: uniformRetVal, info: 1), args: (4), byArg: (kind: uniqueRetVal, info: 1), args: (5), byArg: (kind: virtualConstProp))))))) ; guid = 7004155349499253778
 ; CHECK: ^25 = typeid: (name: "_ZTS1B", summary: (typeTestRes: (kind: inline, sizeM1BitWidth: 0, alignLog2: 1, sizeM1: 2, bitMask: 3, inlineBits: 4))) ; guid = 6203814149063363976
 ; CHECK: ^26 = typeid: (name: "_ZTS1C", summary: (typeTestRes: (kind: single, sizeM1BitWidth: 0))) ; guid = 1884921850105019584
diff --git a/llvm/test/ThinLTO/X86/Inputs/cfi-distributed.ll b/llvm/test/ThinLTO/X86/Inputs/cfi-distributed.ll
new file mode 100644
index 00000000000..376c3f5c157
--- /dev/null
+++ b/llvm/test/ThinLTO/X86/Inputs/cfi-distributed.ll
@@ -0,0 +1,28 @@
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-grtev4-linux-gnu"
+
+%struct.B2 = type { %struct.A2 }
+%struct.A2 = type { i32 (...)** }
+
+@_ZTV1B2 = constant { [3 x i8*] } { [3 x i8*] [i8* undef, i8* undef, i8* undef] }, !type !0
+
+define void @test2(i8* %b) {
+entry:
+  %0 = bitcast i8* %b to i8**
+  %vtable2 = load i8*, i8** %0
+  %1 = tail call i1 @llvm.type.test(i8* %vtable2, metadata !"_ZTS1A2")
+  br i1 %1, label %cont, label %trap
+
+trap:
+  tail call void @llvm.trap()
+  unreachable
+
+cont:
+  ret void
+}
+
+declare i1 @llvm.type.test(i8*, metadata)
+declare void @llvm.trap()
+
+!0 = !{i64 16, !"_ZTS1A2"}
+!1 = !{i64 16, !"_ZTS1B2"}
diff --git a/llvm/test/ThinLTO/X86/cfi-distributed.ll b/llvm/test/ThinLTO/X86/cfi-distributed.ll
new file mode 100644
index 00000000000..40753fd2bff
--- /dev/null
+++ b/llvm/test/ThinLTO/X86/cfi-distributed.ll
@@ -0,0 +1,60 @@
+; REQUIRES: x86-registered-target
+
+; 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 -o %t2.o %p/Inputs/cfi-distributed.ll
+
+; RUN: llvm-lto2 run -thinlto-distributed-indexes %t1.o %t2.o \
+; RUN:   -o %t3 \
+; RUN:   -r=%t1.o,test,px \
+; RUN:   -r=%t1.o,_ZTV1B, \
+; RUN:   -r=%t1.o,_ZTV1B,px \
+; RUN:   -r=%t1.o,test2, \
+; RUN:   -r=%t2.o,test2,px \
+; RUN:   -r=%t2.o,_ZTV1B2, \
+; RUN:   -r=%t2.o,_ZTV1B2,px
+
+; Since @test calls @test2, the latter should be imported here and the
+; first index file should reference both type ids.
+; RUN: llvm-dis %t1.o.thinlto.bc -o - | FileCheck %s --check-prefix=INDEX1
+; INDEX1: typeid: (name: "_ZTS1A"
+; INDEX1: typeid: (name: "_ZTS1A2"
+
+; The second index file, corresponding to @test2, should only contain the
+; typeid for _ZTS1A.
+; RUN: llvm-dis %t2.o.thinlto.bc -o - | FileCheck %s --check-prefix=INDEX2
+; INDEX2-NOT: typeid: (name: "_ZTS1A"
+; INDEX2: typeid: (name: "_ZTS1A2"
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-grtev4-linux-gnu"
+
+%struct.B = type { %struct.A }
+%struct.A = type { i32 (...)** }
+
+@_ZTV1B = constant { [3 x i8*] } { [3 x i8*] [i8* undef, i8* undef, i8* undef] }, !type !0
+
+define void @test(i8* %b) {
+entry:
+  tail call void @test2(i8* %b)
+  %0 = bitcast i8* %b to i8**
+  %vtable2 = load i8*, i8** %0
+  %1 = tail call i1 @llvm.type.test(i8* %vtable2, metadata !"_ZTS1A")
+  br i1 %1, label %cont, label %trap
+
+trap:
+  tail call void @llvm.trap()
+  unreachable
+
+cont:
+  ret void
+}
+
+declare void @test2(i8*)
+declare i1 @llvm.type.test(i8*, metadata)
+declare void @llvm.trap()
+
+!0 = !{i64 16, !"_ZTS1A"}
+!1 = !{i64 16, !"_ZTS1B"}