blob: 1d55a8ea2c64cfa6a44ade013ed4e506779d3d1b [file] [log] [blame]
//===- InstrProf.h - Instrumented profiling format support ------*- C++ -*-===//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
// Instrumentation-based profiling data is generated by instrumented
// binaries through library functions in compiler-rt, and read by the clang
// frontend to feed PGO.
#include "llvm/ADT/ArrayRef.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/StringRef.h"
#include "llvm/ADT/StringSet.h"
#include "llvm/ADT/Triple.h"
#include "llvm/IR/GlobalValue.h"
#include "llvm/IR/ProfileSummary.h"
#include "llvm/ProfileData/"
#include "llvm/Support/Compiler.h"
#include "llvm/Support/Endian.h"
#include "llvm/Support/Error.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/Host.h"
#include "llvm/Support/MD5.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/raw_ostream.h"
#include <algorithm>
#include <cassert>
#include <cstddef>
#include <cstdint>
#include <cstring>
#include <list>
#include <memory>
#include <string>
#include <system_error>
#include <utility>
#include <vector>
namespace llvm {
class Function;
class GlobalVariable;
struct InstrProfRecord;
class InstrProfSymtab;
class Instruction;
class MDNode;
class Module;
enum InstrProfSectKind {
#define INSTR_PROF_SECT_ENTRY(Kind, SectNameCommon, SectNameCoff, Prefix) Kind,
#include "llvm/ProfileData/"
/// Return the name of the profile section corresponding to \p IPSK.
/// The name of the section depends on the object format type \p OF. If
/// \p AddSegmentInfo is true, a segment prefix and additional linker hints may
/// be added to the section name (this is the default).
std::string getInstrProfSectionName(InstrProfSectKind IPSK,
Triple::ObjectFormatType OF,
bool AddSegmentInfo = true);
/// Return the name profile runtime entry point to do value profiling
/// for a given site.
inline StringRef getInstrProfValueProfFuncName() {
/// Return the name profile runtime entry point to do value range profiling.
inline StringRef getInstrProfValueRangeProfFuncName() {
/// Return the name prefix of variables containing instrumented function names.
inline StringRef getInstrProfNameVarPrefix() { return "__profn_"; }
/// Return the name prefix of variables containing per-function control data.
inline StringRef getInstrProfDataVarPrefix() { return "__profd_"; }
/// Return the name prefix of profile counter variables.
inline StringRef getInstrProfCountersVarPrefix() { return "__profc_"; }
/// Return the name prefix of value profile variables.
inline StringRef getInstrProfValuesVarPrefix() { return "__profvp_"; }
/// Return the name of value profile node array variables:
inline StringRef getInstrProfVNodesVarName() { return "__llvm_prf_vnodes"; }
/// Return the name prefix of the COMDAT group for instrumentation variables
/// associated with a COMDAT function.
inline StringRef getInstrProfComdatPrefix() { return "__profv_"; }
/// Return the name of the variable holding the strings (possibly compressed)
/// of all function's PGO names.
inline StringRef getInstrProfNamesVarName() {
return "__llvm_prf_nm";
/// Return the name of a covarage mapping variable (internal linkage)
/// for each instrumented source module. Such variables are allocated
/// in the __llvm_covmap section.
inline StringRef getCoverageMappingVarName() {
return "__llvm_coverage_mapping";
/// Return the name of the internal variable recording the array
/// of PGO name vars referenced by the coverage mapping. The owning
/// functions of those names are not emitted by FE (e.g, unused inline
/// functions.)
inline StringRef getCoverageUnusedNamesVarName() {
return "__llvm_coverage_names";
/// Return the name of function that registers all the per-function control
/// data at program startup time by calling __llvm_register_function. This
/// function has internal linkage and is called by __llvm_profile_init
/// runtime method. This function is not generated for these platforms:
/// Darwin, Linux, and FreeBSD.
inline StringRef getInstrProfRegFuncsName() {
return "__llvm_profile_register_functions";
/// Return the name of the runtime interface that registers per-function control
/// data for one instrumented function.
inline StringRef getInstrProfRegFuncName() {
return "__llvm_profile_register_function";
/// Return the name of the runtime interface that registers the PGO name strings.
inline StringRef getInstrProfNamesRegFuncName() {
return "__llvm_profile_register_names_function";
/// Return the name of the runtime initialization method that is generated by
/// the compiler. The function calls __llvm_profile_register_functions and
/// __llvm_profile_override_default_filename functions if needed. This function
/// has internal linkage and invoked at startup time via init_array.
inline StringRef getInstrProfInitFuncName() { return "__llvm_profile_init"; }
/// Return the name of the hook variable defined in profile runtime library.
/// A reference to the variable causes the linker to link in the runtime
/// initialization module (which defines the hook variable).
inline StringRef getInstrProfRuntimeHookVarName() {
/// Return the name of the compiler generated function that references the
/// runtime hook variable. The function is a weak global.
inline StringRef getInstrProfRuntimeHookVarUseFuncName() {
return "__llvm_profile_runtime_user";
/// Return the marker used to separate PGO names during serialization.
inline StringRef getInstrProfNameSeparator() { return "\01"; }
/// Return the modified name for function \c F suitable to be
/// used the key for profile lookup. Variable \c InLTO indicates if this
/// is called in LTO optimization passes.
std::string getPGOFuncName(const Function &F, bool InLTO = false,
uint64_t Version = INSTR_PROF_INDEX_VERSION);
/// Return the modified name for a function suitable to be
/// used the key for profile lookup. The function's original
/// name is \c RawFuncName and has linkage of type \c Linkage.
/// The function is defined in module \c FileName.
std::string getPGOFuncName(StringRef RawFuncName,
GlobalValue::LinkageTypes Linkage,
StringRef FileName,
uint64_t Version = INSTR_PROF_INDEX_VERSION);
/// Return the name of the global variable used to store a function
/// name in PGO instrumentation. \c FuncName is the name of the function
/// returned by the \c getPGOFuncName call.
std::string getPGOFuncNameVarName(StringRef FuncName,
GlobalValue::LinkageTypes Linkage);
/// Create and return the global variable for function name used in PGO
/// instrumentation. \c FuncName is the name of the function returned
/// by \c getPGOFuncName call.
GlobalVariable *createPGOFuncNameVar(Function &F, StringRef PGOFuncName);
/// Create and return the global variable for function name used in PGO
/// instrumentation. /// \c FuncName is the name of the function
/// returned by \c getPGOFuncName call, \c M is the owning module,
/// and \c Linkage is the linkage of the instrumented function.
GlobalVariable *createPGOFuncNameVar(Module &M,
GlobalValue::LinkageTypes Linkage,
StringRef PGOFuncName);
/// Return the initializer in string of the PGO name var \c NameVar.
StringRef getPGOFuncNameVarInitializer(GlobalVariable *NameVar);
/// Given a PGO function name, remove the filename prefix and return
/// the original (static) function name.
StringRef getFuncNameWithoutPrefix(StringRef PGOFuncName,
StringRef FileName = "<unknown>");
/// Given a vector of strings (function PGO names) \c NameStrs, the
/// method generates a combined string \c Result thatis ready to be
/// serialized. The \c Result string is comprised of three fields:
/// The first field is the legnth of the uncompressed strings, and the
/// the second field is the length of the zlib-compressed string.
/// Both fields are encoded in ULEB128. If \c doCompress is false, the
/// third field is the uncompressed strings; otherwise it is the
/// compressed string. When the string compression is off, the
/// second field will have value zero.
Error collectPGOFuncNameStrings(ArrayRef<std::string> NameStrs,
bool doCompression, std::string &Result);
/// Produce \c Result string with the same format described above. The input
/// is vector of PGO function name variables that are referenced.
Error collectPGOFuncNameStrings(ArrayRef<GlobalVariable *> NameVars,
std::string &Result, bool doCompression = true);
/// \c NameStrings is a string composed of one of more sub-strings encoded in
/// the format described above. The substrings are separated by 0 or more zero
/// bytes. This method decodes the string and populates the \c Symtab.
Error readPGOFuncNameStrings(StringRef NameStrings, InstrProfSymtab &Symtab);
/// Check if INSTR_PROF_RAW_VERSION_VAR is defined. This global is only being
/// set in IR PGO compilation.
bool isIRPGOFlagSet(const Module *M);
/// Check if we can safely rename this Comdat function. Instances of the same
/// comdat function may have different control flows thus can not share the
/// same counter variable.
bool canRenameComdatFunc(const Function &F, bool CheckAddressTaken = false);
enum InstrProfValueKind : uint32_t {
#define VALUE_PROF_KIND(Enumerator, Value) Enumerator = Value,
#include "llvm/ProfileData/"
/// Get the value profile data for value site \p SiteIdx from \p InstrProfR
/// and annotate the instruction \p Inst with the value profile meta data.
/// Annotate up to \p MaxMDCount (default 3) number of records per value site.
void annotateValueSite(Module &M, Instruction &Inst,
const InstrProfRecord &InstrProfR,
InstrProfValueKind ValueKind, uint32_t SiteIndx,
uint32_t MaxMDCount = 3);
/// Same as the above interface but using an ArrayRef, as well as \p Sum.
void annotateValueSite(Module &M, Instruction &Inst,
ArrayRef<InstrProfValueData> VDs, uint64_t Sum,
InstrProfValueKind ValueKind, uint32_t MaxMDCount);
/// Extract the value profile data from \p Inst which is annotated with
/// value profile meta data. Return false if there is no value data annotated,
/// otherwise return true.
bool getValueProfDataFromInst(const Instruction &Inst,
InstrProfValueKind ValueKind,
uint32_t MaxNumValueData,
InstrProfValueData ValueData[],
uint32_t &ActualNumValueData, uint64_t &TotalC);
inline StringRef getPGOFuncNameMetadataName() { return "PGOFuncName"; }
/// Return the PGOFuncName meta data associated with a function.
MDNode *getPGOFuncNameMetadata(const Function &F);
/// Create the PGOFuncName meta data if PGOFuncName is different from
/// function's raw name. This should only apply to internal linkage functions
/// declared by users only.
void createPGOFuncNameMetadata(Function &F, StringRef PGOFuncName);
/// Check if we can use Comdat for profile variables. This will eliminate
/// the duplicated profile variables for Comdat functions.
bool needsComdatForCounter(const Function &F, const Module &M);
const std::error_category &instrprof_category();
enum class instrprof_error {
success = 0,
inline std::error_code make_error_code(instrprof_error E) {
return std::error_code(static_cast<int>(E), instrprof_category());
class InstrProfError : public ErrorInfo<InstrProfError> {
InstrProfError(instrprof_error Err) : Err(Err) {
assert(Err != instrprof_error::success && "Not an error");
std::string message() const override;
void log(raw_ostream &OS) const override { OS << message(); }
std::error_code convertToErrorCode() const override {
return make_error_code(Err);
instrprof_error get() const { return Err; }
/// Consume an Error and return the raw enum value contained within it. The
/// Error must either be a success value, or contain a single InstrProfError.
static instrprof_error take(Error E) {
auto Err = instrprof_error::success;
handleAllErrors(std::move(E), [&Err](const InstrProfError &IPE) {
assert(Err == instrprof_error::success && "Multiple errors encountered");
Err = IPE.get();
return Err;
static char ID;
instrprof_error Err;
class SoftInstrProfErrors {
/// Count the number of soft instrprof_errors encountered and keep track of
/// the first such error for reporting purposes.
/// The first soft error encountered.
instrprof_error FirstError = instrprof_error::success;
/// The number of hash mismatches.
unsigned NumHashMismatches = 0;
/// The number of count mismatches.
unsigned NumCountMismatches = 0;
/// The number of counter overflows.
unsigned NumCounterOverflows = 0;
/// The number of value site count mismatches.
unsigned NumValueSiteCountMismatches = 0;
SoftInstrProfErrors() = default;
~SoftInstrProfErrors() {
assert(FirstError == instrprof_error::success &&
"Unchecked soft error encountered");
/// Track a soft error (\p IE) and increment its associated counter.
void addError(instrprof_error IE);
/// Get the number of hash mismatches.
unsigned getNumHashMismatches() const { return NumHashMismatches; }
/// Get the number of count mismatches.
unsigned getNumCountMismatches() const { return NumCountMismatches; }
/// Get the number of counter overflows.
unsigned getNumCounterOverflows() const { return NumCounterOverflows; }
/// Get the number of value site count mismatches.
unsigned getNumValueSiteCountMismatches() const {
return NumValueSiteCountMismatches;
/// Return the first encountered error and reset FirstError to a success
/// value.
Error takeError() {
if (FirstError == instrprof_error::success)
return Error::success();
auto E = make_error<InstrProfError>(FirstError);
FirstError = instrprof_error::success;
return E;
namespace object {
class SectionRef;
} // end namespace object
namespace IndexedInstrProf {
uint64_t ComputeHash(StringRef K);
} // end namespace IndexedInstrProf
/// A symbol table used for function PGO name look-up with keys
/// (such as pointers, md5hash values) to the function. A function's
/// PGO name or name's md5hash are used in retrieving the profile
/// data of the function. See \c getPGOFuncName() method for details
/// on how PGO name is formed.
class InstrProfSymtab {
using AddrHashMap = std::vector<std::pair<uint64_t, uint64_t>>;
StringRef Data;
uint64_t Address = 0;
// Unique name strings.
StringSet<> NameTab;
// A map from MD5 keys to function name strings.
std::vector<std::pair<uint64_t, StringRef>> MD5NameMap;
// A map from MD5 keys to function define. We only populate this map
// when build the Symtab from a Module.
std::vector<std::pair<uint64_t, Function *>> MD5FuncMap;
// A map from function runtime address to function name MD5 hash.
// This map is only populated and used by raw instr profile reader.
AddrHashMap AddrToMD5Map;
bool Sorted = false;
static StringRef getExternalSymbol() {
return "** External Symbol **";
// If the symtab is created by a series of calls to \c addFuncName, \c
// finalizeSymtab needs to be called before looking up function names.
// This is required because the underlying map is a vector (for space
// efficiency) which needs to be sorted.
inline void finalizeSymtab();
InstrProfSymtab() = default;
/// Create InstrProfSymtab from an object file section which
/// contains function PGO names. When section may contain raw
/// string data or string data in compressed form. This method
/// only initialize the symtab with reference to the data and
/// the section base address. The decompression will be delayed
/// until before it is used. See also \c create(StringRef) method.
Error create(object::SectionRef &Section);
/// This interface is used by reader of CoverageMapping test
/// format.
inline Error create(StringRef D, uint64_t BaseAddr);
/// \c NameStrings is a string composed of one of more sub-strings
/// encoded in the format described in \c collectPGOFuncNameStrings.
/// This method is a wrapper to \c readPGOFuncNameStrings method.
inline Error create(StringRef NameStrings);
/// A wrapper interface to populate the PGO symtab with functions
/// decls from module \c M. This interface is used by transformation
/// passes such as indirect function call promotion. Variable \c InLTO
/// indicates if this is called from LTO optimization passes.
Error create(Module &M, bool InLTO = false);
/// Create InstrProfSymtab from a set of names iteratable from
/// \p IterRange. This interface is used by IndexedProfReader.
template <typename NameIterRange> Error create(const NameIterRange &IterRange);
/// Update the symtab by adding \p FuncName to the table. This interface
/// is used by the raw and text profile readers.
Error addFuncName(StringRef FuncName) {
if (FuncName.empty())
return make_error<InstrProfError>(instrprof_error::malformed);
auto Ins = NameTab.insert(FuncName);
if (Ins.second) {
IndexedInstrProf::ComputeHash(FuncName), Ins.first->getKey()));
Sorted = false;
return Error::success();
/// Map a function address to its name's MD5 hash. This interface
/// is only used by the raw profiler reader.
void mapAddress(uint64_t Addr, uint64_t MD5Val) {
AddrToMD5Map.push_back(std::make_pair(Addr, MD5Val));
/// Return a function's hash, or 0, if the function isn't in this SymTab.
uint64_t getFunctionHashFromAddress(uint64_t Address);
/// Return function's PGO name from the function name's symbol
/// address in the object file. If an error occurs, return
/// an empty string.
StringRef getFuncName(uint64_t FuncNameAddress, size_t NameSize);
/// Return function's PGO name from the name's md5 hash value.
/// If not found, return an empty string.
inline StringRef getFuncName(uint64_t FuncMD5Hash);
/// Just like getFuncName, except that it will return a non-empty StringRef
/// if the function is external to this symbol table. All such cases
/// will be represented using the same StringRef value.
inline StringRef getFuncNameOrExternalSymbol(uint64_t FuncMD5Hash);
/// True if Symbol is the value used to represent external symbols.
static bool isExternalSymbol(const StringRef &Symbol) {
return Symbol == InstrProfSymtab::getExternalSymbol();
/// Return function from the name's md5 hash. Return nullptr if not found.
inline Function *getFunction(uint64_t FuncMD5Hash);
/// Return the function's original assembly name by stripping off
/// the prefix attached (to symbols with priviate linkage). For
/// global functions, it returns the same string as getFuncName.
inline StringRef getOrigFuncName(uint64_t FuncMD5Hash);
/// Return the name section data.
inline StringRef getNameData() const { return Data; }
Error InstrProfSymtab::create(StringRef D, uint64_t BaseAddr) {
Data = D;
Address = BaseAddr;
return Error::success();
Error InstrProfSymtab::create(StringRef NameStrings) {
return readPGOFuncNameStrings(NameStrings, *this);
template <typename NameIterRange>
Error InstrProfSymtab::create(const NameIterRange &IterRange) {
for (auto Name : IterRange)
if (Error E = addFuncName(Name))
return E;
return Error::success();
void InstrProfSymtab::finalizeSymtab() {
if (Sorted)
llvm::sort(MD5NameMap, less_first());
llvm::sort(MD5FuncMap, less_first());
llvm::sort(AddrToMD5Map, less_first());
AddrToMD5Map.erase(std::unique(AddrToMD5Map.begin(), AddrToMD5Map.end()),
Sorted = true;
StringRef InstrProfSymtab::getFuncNameOrExternalSymbol(uint64_t FuncMD5Hash) {
StringRef ret = getFuncName(FuncMD5Hash);
if (ret.empty())
return InstrProfSymtab::getExternalSymbol();
return ret;
StringRef InstrProfSymtab::getFuncName(uint64_t FuncMD5Hash) {
auto Result =
std::lower_bound(MD5NameMap.begin(), MD5NameMap.end(), FuncMD5Hash,
[](const std::pair<uint64_t, std::string> &LHS,
uint64_t RHS) { return LHS.first < RHS; });
if (Result != MD5NameMap.end() && Result->first == FuncMD5Hash)
return Result->second;
return StringRef();
Function* InstrProfSymtab::getFunction(uint64_t FuncMD5Hash) {
auto Result =
std::lower_bound(MD5FuncMap.begin(), MD5FuncMap.end(), FuncMD5Hash,
[](const std::pair<uint64_t, Function*> &LHS,
uint64_t RHS) { return LHS.first < RHS; });
if (Result != MD5FuncMap.end() && Result->first == FuncMD5Hash)
return Result->second;
return nullptr;
// See also getPGOFuncName implementation. These two need to be
// matched.
StringRef InstrProfSymtab::getOrigFuncName(uint64_t FuncMD5Hash) {
StringRef PGOName = getFuncName(FuncMD5Hash);
size_t S = PGOName.find_first_of(':');
if (S == StringRef::npos)
return PGOName;
return PGOName.drop_front(S + 1);
struct InstrProfValueSiteRecord {
/// Value profiling data pairs at a given value site.
std::list<InstrProfValueData> ValueData;
InstrProfValueSiteRecord() { ValueData.clear(); }
template <class InputIterator>
InstrProfValueSiteRecord(InputIterator F, InputIterator L)
: ValueData(F, L) {}
/// Sort ValueData ascending by Value
void sortByTargetValues() {
[](const InstrProfValueData &left, const InstrProfValueData &right) {
return left.Value < right.Value;
/// Sort ValueData Descending by Count
inline void sortByCount();
/// Merge data from another InstrProfValueSiteRecord
/// Optionally scale merged counts by \p Weight.
void merge(InstrProfValueSiteRecord &Input, uint64_t Weight,
function_ref<void(instrprof_error)> Warn);
/// Scale up value profile data counts.
void scale(uint64_t Weight, function_ref<void(instrprof_error)> Warn);
/// Profiling information for a single function.
struct InstrProfRecord {
std::vector<uint64_t> Counts;
InstrProfRecord() = default;
InstrProfRecord(std::vector<uint64_t> Counts) : Counts(std::move(Counts)) {}
InstrProfRecord(InstrProfRecord &&) = default;
InstrProfRecord(const InstrProfRecord &RHS)
: Counts(RHS.Counts),
? llvm::make_unique<ValueProfData>(*RHS.ValueData)
: nullptr) {}
InstrProfRecord &operator=(InstrProfRecord &&) = default;
InstrProfRecord &operator=(const InstrProfRecord &RHS) {
Counts = RHS.Counts;
if (!RHS.ValueData) {
ValueData = nullptr;
return *this;
if (!ValueData)
ValueData = llvm::make_unique<ValueProfData>(*RHS.ValueData);
*ValueData = *RHS.ValueData;
return *this;
/// Return the number of value profile kinds with non-zero number
/// of profile sites.
inline uint32_t getNumValueKinds() const;
/// Return the number of instrumented sites for ValueKind.
inline uint32_t getNumValueSites(uint32_t ValueKind) const;
/// Return the total number of ValueData for ValueKind.
inline uint32_t getNumValueData(uint32_t ValueKind) const;
/// Return the number of value data collected for ValueKind at profiling
/// site: Site.
inline uint32_t getNumValueDataForSite(uint32_t ValueKind,
uint32_t Site) const;
/// Return the array of profiled values at \p Site. If \p TotalC
/// is not null, the total count of all target values at this site
/// will be stored in \c *TotalC.
inline std::unique_ptr<InstrProfValueData[]>
getValueForSite(uint32_t ValueKind, uint32_t Site,
uint64_t *TotalC = nullptr) const;
/// Get the target value/counts of kind \p ValueKind collected at site
/// \p Site and store the result in array \p Dest. Return the total
/// counts of all target values at this site.
inline uint64_t getValueForSite(InstrProfValueData Dest[], uint32_t ValueKind,
uint32_t Site) const;
/// Reserve space for NumValueSites sites.
inline void reserveSites(uint32_t ValueKind, uint32_t NumValueSites);
/// Add ValueData for ValueKind at value Site.
void addValueData(uint32_t ValueKind, uint32_t Site,
InstrProfValueData *VData, uint32_t N,
InstrProfSymtab *SymTab);
/// Merge the counts in \p Other into this one.
/// Optionally scale merged counts by \p Weight.
void merge(InstrProfRecord &Other, uint64_t Weight,
function_ref<void(instrprof_error)> Warn);
/// Scale up profile counts (including value profile data) by
/// \p Weight.
void scale(uint64_t Weight, function_ref<void(instrprof_error)> Warn);
/// Sort value profile data (per site) by count.
void sortValueData() {
for (uint32_t Kind = IPVK_First; Kind <= IPVK_Last; ++Kind)
for (auto &SR : getValueSitesForKind(Kind))
/// Clear value data entries and edge counters.
void Clear() {
/// Clear value data entries
void clearValueData() { ValueData = nullptr; }
struct ValueProfData {
std::vector<InstrProfValueSiteRecord> IndirectCallSites;
std::vector<InstrProfValueSiteRecord> MemOPSizes;
std::unique_ptr<ValueProfData> ValueData;
getValueSitesForKind(uint32_t ValueKind) {
// Cast to /add/ const (should be an implicit_cast, ideally, if that's ever
// implemented in LLVM) to call the const overload of this function, then
// cast away the constness from the result.
auto AR = const_cast<const InstrProfRecord *>(this)->getValueSitesForKind(
return makeMutableArrayRef(
const_cast<InstrProfValueSiteRecord *>(, AR.size());
getValueSitesForKind(uint32_t ValueKind) const {
if (!ValueData)
return None;
switch (ValueKind) {
case IPVK_IndirectCallTarget:
return ValueData->IndirectCallSites;
case IPVK_MemOPSize:
return ValueData->MemOPSizes;
llvm_unreachable("Unknown value kind!");
std::vector<InstrProfValueSiteRecord> &
getOrCreateValueSitesForKind(uint32_t ValueKind) {
if (!ValueData)
ValueData = llvm::make_unique<ValueProfData>();
switch (ValueKind) {
case IPVK_IndirectCallTarget:
return ValueData->IndirectCallSites;
case IPVK_MemOPSize:
return ValueData->MemOPSizes;
llvm_unreachable("Unknown value kind!");
// Map indirect call target name hash to name string.
uint64_t remapValue(uint64_t Value, uint32_t ValueKind,
InstrProfSymtab *SymTab);
// Merge Value Profile data from Src record to this record for ValueKind.
// Scale merged value counts by \p Weight.
void mergeValueProfData(uint32_t ValkeKind, InstrProfRecord &Src,
uint64_t Weight,
function_ref<void(instrprof_error)> Warn);
// Scale up value profile data count.
void scaleValueProfData(uint32_t ValueKind, uint64_t Weight,
function_ref<void(instrprof_error)> Warn);
struct NamedInstrProfRecord : InstrProfRecord {
StringRef Name;
uint64_t Hash;
// We reserve this bit as the flag for context sensitive profile record.
static const int CS_FLAG_IN_FUNC_HASH = 60;
NamedInstrProfRecord() = default;
NamedInstrProfRecord(StringRef Name, uint64_t Hash,
std::vector<uint64_t> Counts)
: InstrProfRecord(std::move(Counts)), Name(Name), Hash(Hash) {}
static bool hasCSFlagInHash(uint64_t FuncHash) {
return ((FuncHash >> CS_FLAG_IN_FUNC_HASH) & 1);
static void setCSFlagInHash(uint64_t &FuncHash) {
FuncHash |= ((uint64_t)1 << CS_FLAG_IN_FUNC_HASH);
uint32_t InstrProfRecord::getNumValueKinds() const {
uint32_t NumValueKinds = 0;
for (uint32_t Kind = IPVK_First; Kind <= IPVK_Last; ++Kind)
NumValueKinds += !(getValueSitesForKind(Kind).empty());
return NumValueKinds;
uint32_t InstrProfRecord::getNumValueData(uint32_t ValueKind) const {
uint32_t N = 0;
for (auto &SR : getValueSitesForKind(ValueKind))
N += SR.ValueData.size();
return N;
uint32_t InstrProfRecord::getNumValueSites(uint32_t ValueKind) const {
return getValueSitesForKind(ValueKind).size();
uint32_t InstrProfRecord::getNumValueDataForSite(uint32_t ValueKind,
uint32_t Site) const {
return getValueSitesForKind(ValueKind)[Site].ValueData.size();
InstrProfRecord::getValueForSite(uint32_t ValueKind, uint32_t Site,
uint64_t *TotalC) const {
uint64_t Dummy;
uint64_t &TotalCount = (TotalC == nullptr ? Dummy : *TotalC);
uint32_t N = getNumValueDataForSite(ValueKind, Site);
if (N == 0) {
TotalCount = 0;
return std::unique_ptr<InstrProfValueData[]>(nullptr);
auto VD = llvm::make_unique<InstrProfValueData[]>(N);
TotalCount = getValueForSite(VD.get(), ValueKind, Site);
return VD;
uint64_t InstrProfRecord::getValueForSite(InstrProfValueData Dest[],
uint32_t ValueKind,
uint32_t Site) const {
uint32_t I = 0;
uint64_t TotalCount = 0;
for (auto V : getValueSitesForKind(ValueKind)[Site].ValueData) {
Dest[I].Value = V.Value;
Dest[I].Count = V.Count;
TotalCount = SaturatingAdd(TotalCount, V.Count);
return TotalCount;
void InstrProfRecord::reserveSites(uint32_t ValueKind, uint32_t NumValueSites) {
if (!NumValueSites)
inline support::endianness getHostEndianness() {
return sys::IsLittleEndianHost ? support::little : support::big;
// Include definitions for value profile data
#include "llvm/ProfileData/"
void InstrProfValueSiteRecord::sortByCount() {
[](const InstrProfValueData &left, const InstrProfValueData &right) {
return left.Count > right.Count;
// Now truncate
if (ValueData.size() > max_s)
namespace IndexedInstrProf {
enum class HashT : uint32_t {
Last = MD5
inline uint64_t ComputeHash(HashT Type, StringRef K) {
switch (Type) {
case HashT::MD5:
return MD5Hash(K);
llvm_unreachable("Unhandled hash type");
const uint64_t Magic = 0x8169666f72706cff; // "\xfflprofi\x81"
enum ProfVersion {
// Version 1 is the first version. In this version, the value of
// a key/value pair can only include profile data of a single function.
// Due to this restriction, the number of block counters for a given
// function is not recorded but derived from the length of the value.
Version1 = 1,
// The version 2 format supports recording profile data of multiple
// functions which share the same key in one value field. To support this,
// the number block counters is recorded as an uint64_t field right after the
// function structural hash.
Version2 = 2,
// Version 3 supports value profile data. The value profile data is expected
// to follow the block counter profile data.
Version3 = 3,
// In this version, profile summary data \c IndexedInstrProf::Summary is
// stored after the profile header.
Version4 = 4,
// In this version, the frontend PGO stable hash algorithm defaults to V2.
Version5 = 5,
// The current version is 5.
const uint64_t Version = ProfVersion::CurrentVersion;
const HashT HashType = HashT::MD5;
inline uint64_t ComputeHash(StringRef K) { return ComputeHash(HashType, K); }
// This structure defines the file header of the LLVM profile
// data file in indexed-format.
struct Header {
uint64_t Magic;
uint64_t Version;
uint64_t Unused; // Becomes unused since version 4
uint64_t HashType;
uint64_t HashOffset;
// Profile summary data recorded in the profile data file in indexed
// format. It is introduced in version 4. The summary data follows
// right after the profile file header.
struct Summary {
struct Entry {
uint64_t Cutoff; ///< The required percentile of total execution count.
MinBlockCount; ///< The minimum execution count for this percentile.
uint64_t NumBlocks; ///< Number of blocks >= the minumum execution count.
// The field kind enumerator to assigned value mapping should remain
// unchanged when a new kind is added or an old kind gets deleted in
// the future.
enum SummaryFieldKind {
/// The total number of functions instrumented.
TotalNumFunctions = 0,
/// Total number of instrumented blocks/edges.
TotalNumBlocks = 1,
/// The maximal execution count among all functions.
/// This field does not exist for profile data from IR based
/// instrumentation.
MaxFunctionCount = 2,
/// Max block count of the program.
MaxBlockCount = 3,
/// Max internal block count of the program (excluding entry blocks).
MaxInternalBlockCount = 4,
/// The sum of all instrumented block counts.
TotalBlockCount = 5,
NumKinds = TotalBlockCount + 1
// The number of summmary fields following the summary header.
uint64_t NumSummaryFields;
// The number of Cutoff Entries (Summary::Entry) following summary fields.
uint64_t NumCutoffEntries;
Summary() = delete;
Summary(uint32_t Size) { memset(this, 0, Size); }
void operator delete(void *ptr) { ::operator delete(ptr); }
static uint32_t getSize(uint32_t NumSumFields, uint32_t NumCutoffEntries) {
return sizeof(Summary) + NumCutoffEntries * sizeof(Entry) +
NumSumFields * sizeof(uint64_t);
const uint64_t *getSummaryDataBase() const {
return reinterpret_cast<const uint64_t *>(this + 1);
uint64_t *getSummaryDataBase() {
return reinterpret_cast<uint64_t *>(this + 1);
const Entry *getCutoffEntryBase() const {
return reinterpret_cast<const Entry *>(
Entry *getCutoffEntryBase() {
return reinterpret_cast<Entry *>(&getSummaryDataBase()[NumSummaryFields]);
uint64_t get(SummaryFieldKind K) const {
return getSummaryDataBase()[K];
void set(SummaryFieldKind K, uint64_t V) {
getSummaryDataBase()[K] = V;
const Entry &getEntry(uint32_t I) const { return getCutoffEntryBase()[I]; }
void setEntry(uint32_t I, const ProfileSummaryEntry &E) {
Entry &ER = getCutoffEntryBase()[I];
ER.Cutoff = E.Cutoff;
ER.MinBlockCount = E.MinCount;
ER.NumBlocks = E.NumCounts;
inline std::unique_ptr<Summary> allocSummary(uint32_t TotalSize) {
return std::unique_ptr<Summary>(new (::operator new(TotalSize))
} // end namespace IndexedInstrProf
namespace RawInstrProf {
// Version 1: First version
// Version 2: Added value profile data section. Per-function control data
// struct has more fields to describe value profile information.
// Version 3: Compressed name section support. Function PGO name reference
// from control data struct is changed from raw pointer to Name's MD5 value.
// Version 4: ValueDataBegin and ValueDataSizes fields are removed from the
// raw header.
// Version 5: Bit 60 of FuncHash is reserved for the flag for the context
// sensitive records.
const uint64_t Version = INSTR_PROF_RAW_VERSION;
template <class IntPtrT> inline uint64_t getMagic();
template <> inline uint64_t getMagic<uint64_t>() {
template <> inline uint64_t getMagic<uint32_t>() {
// Per-function profile data header/control structure.
// The definition should match the structure defined in
// compiler-rt/lib/profile/InstrProfiling.h.
// It should also match the synthesized type in
// Transforms/Instrumentation/InstrProfiling.cpp:getOrCreateRegionCounters.
template <class IntPtrT> struct alignas(8) ProfileData {
#define INSTR_PROF_DATA(Type, LLVMType, Name, Init) Type Name;
#include "llvm/ProfileData/"
// File header structure of the LLVM profile data in raw format.
// The definition should match the header referenced in
// compiler-rt/lib/profile/InstrProfilingFile.c and
// InstrProfilingBuffer.c.
struct Header {
#define INSTR_PROF_RAW_HEADER(Type, Name, Init) const Type Name;
#include "llvm/ProfileData/"
} // end namespace RawInstrProf
// Parse MemOP Size range option.
void getMemOPSizeRangeFromOption(StringRef Str, int64_t &RangeStart,
int64_t &RangeLast);
// Create a COMDAT variable INSTR_PROF_RAW_VERSION_VAR to make the runtime
// aware this is an ir_level profile so it can set the version flag.
void createIRLevelProfileFlagVar(Module &M, bool IsCS);
// Create the variable for the profile file name.
void createProfileFileNameVar(Module &M, StringRef InstrProfileOutput);
} // end namespace llvm