maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[GCOV] Emit the writeout function as nested loops of global data. 

Summary:
Prior to this change, LLVM would in some cases emit *massive* writeout
functions with many 10s of 1000s of function calls in straight-line
code. This is a very wasteful way to represent what are fundamentally
loops and creates a number of scalability issues. Among other things,
register allocating these calls is extremely expensive. While D46127 makes this
less severe, we'll still run into scaling issues with this eventually. If not
in the compile time, just from the code size.

Now the pass builds up global data structures modeling the inputs to
these functions, and simply loops over the data structures calling the
relevant functions with those values. This ensures that the code size is
a fixed and only data size grows with larger amounts of coverage data.

A trivial change to IRBuilder is included to make it easier to build
the constants that make up the global data.

Reviewers: wmi, echristo

Subscribers: sanjoy, mcrosier, llvm-commits, hiraditya

Differential Revision: https://reviews.llvm.org/D46357

llvm-svn: 331407
This commit is contained in:
Chandler Carruth 2018-05-02 22:24:39 +00:00
parent fbe748ae86
commit 71c3a3fac5
3 changed files with 266 additions and 47 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
llvm/include/llvm/IR/IRBuilder.h
View File

@ -1476,14 +1476,19 @@ public:
return CreateConstInBoundsGEP2_32(Ty, Ptr, 0, Idx, Name);
}
Value *CreateStructGEP(Value *Ptr, unsigned Idx, const Twine &Name = "") {
return CreateConstInBoundsGEP2_32(nullptr, Ptr, 0, Idx, Name);
}
/// Same as CreateGlobalString, but return a pointer with "i8*" type
/// instead of a pointer to array of i8.
Value *CreateGlobalStringPtr(StringRef Str, const Twine &Name = "",
unsigned AddressSpace = 0) {
GlobalVariable *gv = CreateGlobalString(Str, Name, AddressSpace);
Value *zero = ConstantInt::get(Type::getInt32Ty(Context), 0);
Value *Args[] = { zero, zero };
return CreateInBoundsGEP(gv->getValueType(), gv, Args, Name);
Constant *CreateGlobalStringPtr(StringRef Str, const Twine &Name = "",
unsigned AddressSpace = 0) {
GlobalVariable *GV = CreateGlobalString(Str, Name, AddressSpace);
Constant *Zero = ConstantInt::get(Type::getInt32Ty(Context), 0);
Constant *Indices[] = {Zero, Zero};
return ConstantExpr::getInBoundsGetElementPtr(GV->getValueType(), GV,
Indices);
}
//===--------------------------------------------------------------------===//

227
llvm/lib/Transforms/Instrumentation/GCOVProfiling.cpp
View File

@ -17,6 +17,7 @@
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/Hashing.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/Sequence.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/ADT/StringExtras.h"
#include "llvm/ADT/StringMap.h"
@ -35,8 +36,8 @@
#include "llvm/Support/FileSystem.h"
#include "llvm/Support/Path.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Transforms/Instrumentation/GCOVProfiler.h"
#include "llvm/Transforms/Instrumentation.h"
#include "llvm/Transforms/Instrumentation/GCOVProfiler.h"
#include "llvm/Transforms/Utils/ModuleUtils.h"
#include <algorithm>
#include <memory>
@ -886,46 +887,196 @@ Function *GCOVProfiler::insertCounterWriteout(
Constant *SummaryInfo = getSummaryInfoFunc();
Constant *EndFile = getEndFileFunc();
NamedMDNode *CU_Nodes = M->getNamedMetadata("llvm.dbg.cu");
if (CU_Nodes) {
for (unsigned i = 0, e = CU_Nodes->getNumOperands(); i != e; ++i) {
auto *CU = cast<DICompileUnit>(CU_Nodes->getOperand(i));
// Skip module skeleton (and module) CUs.
if (CU->getDWOId())
continue;
std::string FilenameGcda = mangleName(CU, GCovFileType::GCDA);
uint32_t CfgChecksum = FileChecksums.empty() ? 0 : FileChecksums[i];
Builder.CreateCall(StartFile,
{Builder.CreateGlobalStringPtr(FilenameGcda),
Builder.CreateGlobalStringPtr(ReversedVersion),
Builder.getInt32(CfgChecksum)});
for (unsigned j = 0, e = CountersBySP.size(); j != e; ++j) {
auto *SP = cast_or_null<DISubprogram>(CountersBySP[j].second);
uint32_t FuncChecksum = Funcs.empty() ? 0 : Funcs[j]->getFuncChecksum();
Builder.CreateCall(
EmitFunction,
{Builder.getInt32(j),
Options.FunctionNamesInData
? Builder.CreateGlobalStringPtr(getFunctionName(SP))
: Constant::getNullValue(Builder.getInt8PtrTy()),
Builder.getInt32(FuncChecksum),
Builder.getInt8(Options.UseCfgChecksum),
Builder.getInt32(CfgChecksum)});
GlobalVariable *GV = CountersBySP[j].first;
unsigned Arcs =
cast<ArrayType>(GV->getValueType())->getNumElements();
Builder.CreateCall(EmitArcs, {Builder.getInt32(Arcs),
Builder.CreateConstGEP2_64(GV, 0, 0)});
}
Builder.CreateCall(SummaryInfo, {});
Builder.CreateCall(EndFile, {});
}
NamedMDNode *CUNodes = M->getNamedMetadata("llvm.dbg.cu");
if (!CUNodes) {
Builder.CreateRetVoid();
return WriteoutF;
}
// Collect the relevant data into a large constant data structure that we can
// walk to write out everything.
StructType *StartFileCallArgsTy = StructType::create(
{Builder.getInt8PtrTy(), Builder.getInt8PtrTy(), Builder.getInt32Ty()});
StructType *EmitFunctionCallArgsTy = StructType::create(
{Builder.getInt32Ty(), Builder.getInt8PtrTy(), Builder.getInt32Ty(),
Builder.getInt8Ty(), Builder.getInt32Ty()});
StructType *EmitArcsCallArgsTy = StructType::create(
{Builder.getInt32Ty(), Builder.getInt64Ty()->getPointerTo()});
StructType *FileInfoTy =
StructType::create({StartFileCallArgsTy, Builder.getInt32Ty(),
EmitFunctionCallArgsTy->getPointerTo(),
EmitArcsCallArgsTy->getPointerTo()});
Constant *Zero32 = Builder.getInt32(0);
SmallVector<Constant *, 8> FileInfos;
for (int i : llvm::seq<int>(0, CUNodes->getNumOperands())) {
auto *CU = cast<DICompileUnit>(CUNodes->getOperand(i));
// Skip module skeleton (and module) CUs.
if (CU->getDWOId())
continue;
std::string FilenameGcda = mangleName(CU, GCovFileType::GCDA);
uint32_t CfgChecksum = FileChecksums.empty() ? 0 : FileChecksums[i];
auto *StartFileCallArgs = ConstantStruct::get(
StartFileCallArgsTy, {Builder.CreateGlobalStringPtr(FilenameGcda),
Builder.CreateGlobalStringPtr(ReversedVersion),
Builder.getInt32(CfgChecksum)});
SmallVector<Constant *, 8> EmitFunctionCallArgsArray;
SmallVector<Constant *, 8> EmitArcsCallArgsArray;
for (int j : llvm::seq<int>(0, CountersBySP.size())) {
auto *SP = cast_or_null<DISubprogram>(CountersBySP[j].second);
uint32_t FuncChecksum = Funcs.empty() ? 0 : Funcs[j]->getFuncChecksum();
EmitFunctionCallArgsArray.push_back(ConstantStruct::get(
EmitFunctionCallArgsTy,
{Builder.getInt32(j),
Options.FunctionNamesInData
? Builder.CreateGlobalStringPtr(getFunctionName(SP))
: Constant::getNullValue(Builder.getInt8PtrTy()),
Builder.getInt32(FuncChecksum),
Builder.getInt8(Options.UseCfgChecksum),
Builder.getInt32(CfgChecksum)}));
GlobalVariable *GV = CountersBySP[j].first;
unsigned Arcs = cast<ArrayType>(GV->getValueType())->getNumElements();
EmitArcsCallArgsArray.push_back(ConstantStruct::get(
EmitArcsCallArgsTy,
{Builder.getInt32(Arcs),
ConstantExpr::getInBoundsGetElementPtr(
GV->getValueType(), GV,
makeArrayRef<Constant *>({Zero32, Zero32}))}));
}
// Create global arrays for the two emit calls.
int CountersSize = CountersBySP.size();
assert(CountersSize == (int)EmitFunctionCallArgsArray.size() &&
"Mismatched array size!");
assert(CountersSize == (int)EmitArcsCallArgsArray.size() &&
"Mismatched array size!");
auto *EmitFunctionCallArgsArrayTy =
ArrayType::get(EmitFunctionCallArgsTy, CountersSize);
auto *EmitFunctionCallArgsArrayGV = new GlobalVariable(
*M, EmitFunctionCallArgsArrayTy, /*isConstant*/ true,
GlobalValue::InternalLinkage,
ConstantArray::get(EmitFunctionCallArgsArrayTy,
EmitFunctionCallArgsArray),
Twine("__llvm_internal_gcov_emit_function_args.") + Twine(i));
auto *EmitArcsCallArgsArrayTy =
ArrayType::get(EmitArcsCallArgsTy, CountersSize);
EmitFunctionCallArgsArrayGV->setUnnamedAddr(
GlobalValue::UnnamedAddr::Global);
auto *EmitArcsCallArgsArrayGV = new GlobalVariable(
*M, EmitArcsCallArgsArrayTy, /*isConstant*/ true,
GlobalValue::InternalLinkage,
ConstantArray::get(EmitArcsCallArgsArrayTy, EmitArcsCallArgsArray),
Twine("__llvm_internal_gcov_emit_arcs_args.") + Twine(i));
EmitArcsCallArgsArrayGV->setUnnamedAddr(GlobalValue::UnnamedAddr::Global);
FileInfos.push_back(ConstantStruct::get(
FileInfoTy,
{StartFileCallArgs, Builder.getInt32(CountersSize),
ConstantExpr::getInBoundsGetElementPtr(
EmitFunctionCallArgsArrayTy, EmitFunctionCallArgsArrayGV,
makeArrayRef<Constant *>({Zero32, Zero32})),
ConstantExpr::getInBoundsGetElementPtr(
EmitArcsCallArgsArrayTy, EmitArcsCallArgsArrayGV,
makeArrayRef<Constant *>({Zero32, Zero32}))}));
}
// If we didn't find anything to actually emit, bail on out.
if (FileInfos.empty()) {
Builder.CreateRetVoid();
return WriteoutF;
}
// To simplify code, we cap the number of file infos we write out to fit
// easily in a 32-bit signed integer. This gives consistent behavior between
// 32-bit and 64-bit systems without requiring (potentially very slow) 64-bit
// operations on 32-bit systems. It also seems unreasonable to try to handle
// more than 2 billion files.
if ((int64_t)FileInfos.size() > (int64_t)INT_MAX)
FileInfos.resize(INT_MAX);
// Create a global for the entire data structure so we can walk it more
// easily.
auto *FileInfoArrayTy = ArrayType::get(FileInfoTy, FileInfos.size());
auto *FileInfoArrayGV = new GlobalVariable(
*M, FileInfoArrayTy, /*isConstant*/ true, GlobalValue::InternalLinkage,
ConstantArray::get(FileInfoArrayTy, FileInfos),
"__llvm_internal_gcov_emit_file_info");
FileInfoArrayGV->setUnnamedAddr(GlobalValue::UnnamedAddr::Global);
// Create the CFG for walking this data structure.
auto *FileLoopHeader =
BasicBlock::Create(*Ctx, "file.loop.header", WriteoutF);
auto *CounterLoopHeader =
BasicBlock::Create(*Ctx, "counter.loop.header", WriteoutF);
auto *FileLoopLatch = BasicBlock::Create(*Ctx, "file.loop.latch", WriteoutF);
auto *ExitBB = BasicBlock::Create(*Ctx, "exit", WriteoutF);
// We always have at least one file, so just branch to the header.
Builder.CreateBr(FileLoopHeader);
// The index into the files structure is our loop induction variable.
Builder.SetInsertPoint(FileLoopHeader);
PHINode *IV =
Builder.CreatePHI(Builder.getInt32Ty(), /*NumReservedValues*/ 2);
IV->addIncoming(Builder.getInt32(0), BB);
auto *FileInfoPtr =
Builder.CreateInBoundsGEP(FileInfoArrayGV, {Builder.getInt32(0), IV});
auto *StartFileCallArgsPtr = Builder.CreateStructGEP(FileInfoPtr, 0);
Builder.CreateCall(
StartFile,
{Builder.CreateLoad(Builder.CreateStructGEP(StartFileCallArgsPtr, 0)),
Builder.CreateLoad(Builder.CreateStructGEP(StartFileCallArgsPtr, 1)),
Builder.CreateLoad(Builder.CreateStructGEP(StartFileCallArgsPtr, 2))});
auto *NumCounters =
Builder.CreateLoad(Builder.CreateStructGEP(FileInfoPtr, 1));
auto *EmitFunctionCallArgsArray =
Builder.CreateLoad(Builder.CreateStructGEP(FileInfoPtr, 2));
auto *EmitArcsCallArgsArray =
Builder.CreateLoad(Builder.CreateStructGEP(FileInfoPtr, 3));
auto *EnterCounterLoopCond =
Builder.CreateICmpSLT(Builder.getInt32(0), NumCounters);
Builder.CreateCondBr(EnterCounterLoopCond, CounterLoopHeader, FileLoopLatch);
Builder.SetInsertPoint(CounterLoopHeader);
auto *JV = Builder.CreatePHI(Builder.getInt32Ty(), /*NumReservedValues*/ 2);
JV->addIncoming(Builder.getInt32(0), FileLoopHeader);
auto *EmitFunctionCallArgsPtr =
Builder.CreateInBoundsGEP(EmitFunctionCallArgsArray, {JV});
Builder.CreateCall(
EmitFunction,
{Builder.CreateLoad(Builder.CreateStructGEP(EmitFunctionCallArgsPtr, 0)),
Builder.CreateLoad(Builder.CreateStructGEP(EmitFunctionCallArgsPtr, 1)),
Builder.CreateLoad(Builder.CreateStructGEP(EmitFunctionCallArgsPtr, 2)),
Builder.CreateLoad(Builder.CreateStructGEP(EmitFunctionCallArgsPtr, 3)),
Builder.CreateLoad(
Builder.CreateStructGEP(EmitFunctionCallArgsPtr, 4))});
auto *EmitArcsCallArgsPtr =
Builder.CreateInBoundsGEP(EmitArcsCallArgsArray, {JV});
Builder.CreateCall(
EmitArcs,
{Builder.CreateLoad(Builder.CreateStructGEP(EmitArcsCallArgsPtr, 0)),
Builder.CreateLoad(Builder.CreateStructGEP(EmitArcsCallArgsPtr, 1))});
auto *NextJV = Builder.CreateAdd(JV, Builder.getInt32(1));
auto *CounterLoopCond = Builder.CreateICmpSLT(NextJV, NumCounters);
Builder.CreateCondBr(CounterLoopCond, CounterLoopHeader, FileLoopLatch);
JV->addIncoming(NextJV, CounterLoopHeader);
Builder.SetInsertPoint(FileLoopLatch);
Builder.CreateCall(SummaryInfo, {});
Builder.CreateCall(EndFile, {});
auto *NextIV = Builder.CreateAdd(IV, Builder.getInt32(1));
auto *FileLoopCond =
Builder.CreateICmpSLT(NextIV, Builder.getInt32(FileInfos.size()));
Builder.CreateCondBr(FileLoopCond, FileLoopHeader, ExitBB);
IV->addIncoming(NextIV, FileLoopLatch);
Builder.SetInsertPoint(ExitBB);
Builder.CreateRetVoid();
return WriteoutF;
}

69
llvm/test/Transforms/GCOVProfiling/function-numbering.ll
View File

@ -19,9 +19,72 @@ target triple = "x86_64-apple-macosx10.10.0"
; GCDA: @[[FOO:[0-9]+]] = private unnamed_addr constant [4 x i8] c"foo\00"
; GCDA-NOT: @{{[0-9]+}} = private unnamed_addr constant .* c"bar\00"
; GCDA: @[[BAZ:[0-9]+]] = private unnamed_addr constant [4 x i8] c"baz\00"
; GCDA: define internal void @__llvm_gcov_writeout()
; GCDA: call void @llvm_gcda_emit_function(i32 0, i8* getelementptr inbounds ([4 x i8], [4 x i8]* @[[FOO]]
; GCDA: call void @llvm_gcda_emit_function(i32 1, i8* getelementptr inbounds ([4 x i8], [4 x i8]* @[[BAZ]]
; GCDA: @__llvm_internal_gcov_emit_function_args.0 = internal unnamed_addr constant
; GCDA-SAME: { i32 0, i8* getelementptr inbounds ([4 x i8], [4 x i8]* @[[FOO]]
; GCDA-SAME: { i32 1, i8* getelementptr inbounds ([4 x i8], [4 x i8]* @[[BAZ]]
;
; GCDA-LABEL: define internal void @__llvm_gcov_writeout() {{.*}} {
; GCDA-NEXT: entry:
; GCDA-NEXT: br label %[[FILE_LOOP_HEADER:.*]]
;
; GCDA: [[FILE_LOOP_HEADER]]:
; GCDA-NEXT: %[[IV:.*]] = phi i32 [ 0, %entry ], [ %[[NEXT_IV:.*]], %[[FILE_LOOP_LATCH:.*]] ]
; GCDA-NEXT: %[[FILE_INFO:.*]] = getelementptr inbounds {{.*}}, {{.*}}* @__llvm_internal_gcov_emit_file_info, i32 0, i32 %[[IV]]
; GCDA-NEXT: %[[START_FILE_ARGS:.*]] = getelementptr inbounds {{.*}}, {{.*}}* %[[FILE_INFO]], i32 0, i32 0
; GCDA-NEXT: %[[START_FILE_ARG_0_PTR:.*]] = getelementptr inbounds {{.*}}, {{.*}}* %[[START_FILE_ARGS]], i32 0, i32 0
; GCDA-NEXT: %[[START_FILE_ARG_0:.*]] = load i8*, i8** %[[START_FILE_ARG_0_PTR]]
; GCDA-NEXT: %[[START_FILE_ARG_1_PTR:.*]] = getelementptr inbounds {{.*}}, {{.*}}* %[[START_FILE_ARGS]], i32 0, i32 1
; GCDA-NEXT: %[[START_FILE_ARG_1:.*]] = load i8*, i8** %[[START_FILE_ARG_1_PTR]]
; GCDA-NEXT: %[[START_FILE_ARG_2_PTR:.*]] = getelementptr inbounds {{.*}}, {{.*}}* %[[START_FILE_ARGS]], i32 0, i32 2
; GCDA-NEXT: %[[START_FILE_ARG_2:.*]] = load i32, i32* %[[START_FILE_ARG_2_PTR]]
; GCDA-NEXT: call void @llvm_gcda_start_file(i8* %[[START_FILE_ARG_0]], i8* %[[START_FILE_ARG_1]], i32 %[[START_FILE_ARG_2]])
; GCDA-NEXT: %[[NUM_COUNTERS_PTR:.*]] = getelementptr inbounds {{.*}}, {{.*}}* %[[FILE_INFO]], i32 0, i32 1
; GCDA-NEXT: %[[NUM_COUNTERS:.*]] = load i32, i32* %[[NUM_COUNTERS_PTR]]
; GCDA-NEXT: %[[EMIT_FUN_ARGS_ARRAY_PTR:.*]] = getelementptr inbounds {{.*}}, {{.*}}* %[[FILE_INFO]], i32 0, i32 2
; GCDA-NEXT: %[[EMIT_FUN_ARGS_ARRAY:.*]] = load {{.*}}*, {{.*}}** %[[EMIT_FUN_ARGS_ARRAY_PTR]]
; GCDA-NEXT: %[[EMIT_ARCS_ARGS_ARRAY_PTR:.*]] = getelementptr inbounds {{.*}}, {{.*}}* %[[FILE_INFO]], i32 0, i32 3
; GCDA-NEXT: %[[EMIT_ARCS_ARGS_ARRAY:.*]] = load {{.*}}*, {{.*}}** %[[EMIT_ARCS_ARGS_ARRAY_PTR]]
; GCDA-NEXT: %[[ENTER_COUNTER_LOOP_COND:.*]] = icmp slt i32 0, %[[NUM_COUNTERS]]
; GCDA-NEXT: br i1 %[[ENTER_COUNTER_LOOP_COND]], label %[[COUNTER_LOOP:.*]], label %[[FILE_LOOP_LATCH]]
;
; GCDA: [[COUNTER_LOOP]]:
; GCDA-NEXT: %[[JV:.*]] = phi i32 [ 0, %[[FILE_LOOP_HEADER]] ], [ %[[NEXT_JV:.*]], %[[COUNTER_LOOP]] ]
; GCDA-NEXT: %[[EMIT_FUN_ARGS:.*]] = getelementptr inbounds {{.*}}, {{.*}}* %[[EMIT_FUN_ARGS_ARRAY]], i32 %[[JV]]
; GCDA-NEXT: %[[EMIT_FUN_ARG_0_PTR:.*]] = getelementptr inbounds {{.*}}, {{.*}}* %[[EMIT_FUN_ARGS]], i32 0, i32 0
; GCDA-NEXT: %[[EMIT_FUN_ARG_0:.*]] = load i32, i32* %[[EMIT_FUN_ARG_0_PTR]]
; GCDA-NEXT: %[[EMIT_FUN_ARG_1_PTR:.*]] = getelementptr inbounds {{.*}}, {{.*}}* %[[EMIT_FUN_ARGS]], i32 0, i32 1
; GCDA-NEXT: %[[EMIT_FUN_ARG_1:.*]] = load i8*, i8** %[[EMIT_FUN_ARG_1_PTR]]
; GCDA-NEXT: %[[EMIT_FUN_ARG_2_PTR:.*]] = getelementptr inbounds {{.*}}, {{.*}}* %[[EMIT_FUN_ARGS]], i32 0, i32 2
; GCDA-NEXT: %[[EMIT_FUN_ARG_2:.*]] = load i32, i32* %[[EMIT_FUN_ARG_2_PTR]]
; GCDA-NEXT: %[[EMIT_FUN_ARG_3_PTR:.*]] = getelementptr inbounds {{.*}}, {{.*}}* %[[EMIT_FUN_ARGS]], i32 0, i32 3
; GCDA-NEXT: %[[EMIT_FUN_ARG_3:.*]] = load i8, i8* %[[EMIT_FUN_ARG_3_PTR]]
; GCDA-NEXT: %[[EMIT_FUN_ARG_4_PTR:.*]] = getelementptr inbounds {{.*}}, {{.*}}* %[[EMIT_FUN_ARGS]], i32 0, i32 4
; GCDA-NEXT: %[[EMIT_FUN_ARG_4:.*]] = load i32, i32* %[[EMIT_FUN_ARG_4_PTR]]
; GCDA-NEXT: call void @llvm_gcda_emit_function(i32 %[[EMIT_FUN_ARG_0]],
; GCDA-SAME: i8* %[[EMIT_FUN_ARG_1]],
; GCDA-SAME: i32 %[[EMIT_FUN_ARG_2]],
; GCDA-SAME: i8 %[[EMIT_FUN_ARG_3]],
; GCDA-SAME: i32 %[[EMIT_FUN_ARG_4]])
; GCDA-NEXT: %[[EMIT_ARCS_ARGS:.*]] = getelementptr inbounds {{.*}}, {{.*}}* %[[EMIT_ARCS_ARGS_ARRAY]], i32 %[[JV]]
; GCDA-NEXT: %[[EMIT_ARCS_ARG_0_PTR:.*]] = getelementptr inbounds {{.*}}, {{.*}}* %[[EMIT_ARCS_ARGS]], i32 0, i32 0
; GCDA-NEXT: %[[EMIT_ARCS_ARG_0:.*]] = load i32, i32* %[[EMIT_ARCS_ARG_0_PTR]]
; GCDA-NEXT: %[[EMIT_ARCS_ARG_1_PTR:.*]] = getelementptr inbounds {{.*}}, {{.*}}* %[[EMIT_ARCS_ARGS]], i32 0, i32 1
; GCDA-NEXT: %[[EMIT_ARCS_ARG_1:.*]] = load i64*, i64** %[[EMIT_ARCS_ARG_1_PTR]]
; GCDA-NEXT: call void @llvm_gcda_emit_arcs(i32 %[[EMIT_ARCS_ARG_0]],
; GCDA-SAME: i64* %[[EMIT_ARCS_ARG_1]])
; GCDA-NEXT: %[[NEXT_JV]] = add i32 %[[JV]], 1
; GCDA-NEXT: %[[COUNTER_LOOP_COND:.*]] = icmp slt i32 %[[NEXT_JV]], %[[NUM_COUNTERS]]
; GCDA-NEXT: br i1 %[[COUNTER_LOOP_COND]], label %[[COUNTER_LOOP]], label %[[FILE_LOOP_LATCH]]
;
; GCDA: [[FILE_LOOP_LATCH]]:
; GCDA-NEXT: call void @llvm_gcda_summary_info()
; GCDA-NEXT: call void @llvm_gcda_end_file()
; GCDA-NEXT: %[[NEXT_IV]] = add i32 %[[IV]], 1
; GCDA-NEXT: %[[FILE_LOOP_COND:.*]] = icmp slt i32 %[[NEXT_IV]], 1
; GCDA-NEXT: br i1 %[[FILE_LOOP_COND]], label %[[FILE_LOOP_HEADER]], label %[[EXIT:.*]]
;
; GCDA: [[EXIT]]:
; GCDA-NEXT: ret void
; GCNO: == foo (0) @
; GCNO-NOT: == bar ({{[0-9]+}}) @