RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-26 12:43:36 +01:00
Code Issues Projects Releases Wiki Activity

Move hash computation code into a separate class and file.

Browse Source 
No functional change intended.

llvm-svn: 188028
This commit is contained in:
Eric Christopher 2013-08-08 23:45:55 +00:00
parent 69e05b321a
commit 81863ce2b7
3 changed files with 191 additions and 99 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

136
lib/CodeGen/AsmPrinter/DIEHash.cpp Normal file
View File

@ -0,0 +1,136 @@
//===-- llvm/CodeGen/DIEHash.cpp - Dwarf Hashing Framework ----------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file contains support for DWARF4 hashing of DIEs.
//
//===----------------------------------------------------------------------===//
#define DEBUG_TYPE "dwarfdebug"
#include "DIE.h"
#include "DIEHash.h"
#include "DwarfCompileUnit.h"
#include "llvm/ADT/ArrayRef.h"
#include "llvm/ADT/StringRef.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/Dwarf.h"
#include "llvm/Support/Endian.h"
#include "llvm/Support/MD5.h"
#include "llvm/Support/raw_ostream.h"
using namespace llvm;
/// \brief Grabs the string in whichever attribute is passed in and returns
/// a reference to it.
static StringRef getDIEStringAttr(DIE *Die, uint16_t Attr) {
const SmallVectorImpl<DIEValue *> &Values = Die->getValues();
const DIEAbbrev &Abbrevs = Die->getAbbrev();
// Iterate through all the attributes until we find the one we're
// looking for, if we can't find it return an empty string.
for (size_t i = 0; i < Values.size(); ++i) {
if (Abbrevs.getData()[i].getAttribute() == Attr) {
DIEValue *V = Values[i];
assert(isa<DIEString>(V) && "String requested. Not a string.");
DIEString *S = cast<DIEString>(V);
return S->getString();
}
}
return StringRef("");
}
/// \brief Adds the string in \p Str to the hash. This also hashes
/// a trailing NULL with the string.
void DIEHash::addString(StringRef Str) {
DEBUG(dbgs() << "Adding string " << Str << " to hash.\n");
Hash.update(Str);
Hash.update(makeArrayRef((uint8_t)'\0'));
}
// FIXME: The LEB128 routines are copied and only slightly modified out of
// LEB128.h.
/// \brief Adds the unsigned in \p Value to the hash encoded as a ULEB128.
void DIEHash::addULEB128(uint64_t Value) {
DEBUG(dbgs() << "Adding ULEB128 " << Value << " to hash.\n");
do {
uint8_t Byte = Value & 0x7f;
Value >>= 7;
if (Value != 0)
Byte |= 0x80; // Mark this byte to show that more bytes will follow.
Hash.update(Byte);
} while (Value != 0);
}
/// \brief Including \p Parent adds the context of Parent to the hash..
void DIEHash::addParentContext(DIE *Parent) {
DEBUG(dbgs() << "Adding parent context to hash...\n");
// [7.27.2] For each surrounding type or namespace beginning with the
// outermost such construct...
SmallVector<DIE *, 1> Parents;
while (Parent->getTag() != dwarf::DW_TAG_compile_unit) {
Parents.push_back(Parent);
Parent = Parent->getParent();
}
// Reverse iterate over our list to go from the outermost construct to the
// innermost.
for (SmallVectorImpl<DIE *>::reverse_iterator I = Parents.rbegin(),
E = Parents.rend();
I != E; ++I) {
DIE *Die = *I;
// ... Append the letter "C" to the sequence...
addULEB128('C');
// ... Followed by the DWARF tag of the construct...
addULEB128(Die->getTag());
// ... Then the name, taken from the DW_AT_name attribute.
StringRef Name = getDIEStringAttr(Die, dwarf::DW_AT_name);
DEBUG(dbgs() << "... adding context: " << Name << "\n");
if (!Name.empty())
addString(Name);
}
}
/// This is based on the type signature computation given in section 7.27 of the
/// DWARF4 standard. It is the md5 hash of a flattened description of the DIE
/// with
/// the exception that we are hashing only the context and the name of the type.
uint64_t DIEHash::computeDIEODRSignature(DIE *Die) {
// Add the contexts to the hash. We won't be computing the ODR hash for
// function local types so it's safe to use the generic context hashing
// algorithm here.
// FIXME: If we figure out how to account for linkage in some way we could
// actually do this with a slight modification to the parent hash algorithm.
DIE *Parent = Die->getParent();
if (Parent)
addParentContext(Parent);
// Add the current DIE information.
// Add the DWARF tag of the DIE.
addULEB128(Die->getTag());
// Add the name of the type to the hash.
addString(getDIEStringAttr(Die, dwarf::DW_AT_name));
// Now get the result.
MD5::MD5Result Result;
Hash.final(Result);
// ... take the least significant 8 bytes and return those. Our MD5
// implementation always returns its results in little endian, swap bytes
// appropriately.
return *reinterpret_cast<support::ulittle64_t *>(Result + 8);
}

46
lib/CodeGen/AsmPrinter/DIEHash.h Normal file
View File

@ -0,0 +1,46 @@
//===-- llvm/CodeGen/DIEHash.h - Dwarf Hashing Framework -------*- C++ -*--===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file contains support for DWARF4 hashing of DIEs.
//
//===----------------------------------------------------------------------===//
#include "llvm/Support/MD5.h"
namespace llvm {
class CompileUnit;
/// \brief An object containing the capability of hashing and adding hash
/// attributes onto a DIE.
class DIEHash {
public:
/// \brief Initializes. The hash is default initialized.
DIEHash() {}
/// \brief Computes the ODR signature
uint64_t computeDIEODRSignature(DIE *Die);
// Helper routines to process parts of a DIE.
private:
/// \brief Adds the parent context of \param Die to the hash.
void addParentContext(DIE *Die);
// Routines that add DIEValues to the hash.
private:
/// \brief Encodes and adds \param Value to the hash as a ULEB128.
void addULEB128(uint64_t Value);
/// \brief Adds \param Str to the hash and includes a NULL byte.
void addString(StringRef Str);
private:
MD5 Hash;
};
}

108
lib/CodeGen/AsmPrinter/DwarfDebug.cpp
View File

@ -14,6 +14,7 @@
#define DEBUG_TYPE "dwarfdebug"
#include "DwarfDebug.h"
#include "DIE.h"
#include "DIEHash.h"
#include "DwarfAccelTable.h"
#include "DwarfCompileUnit.h"
#include "llvm/ADT/STLExtras.h"
@ -962,8 +963,7 @@ void DwarfDebug::collectDeadVariables() {
DeleteContainerSeconds(DeadFnScopeMap);
}
// Type Signature [7.27] computation code.
typedef ArrayRef<uint8_t> HashValue;
// Type Signature [7.27] and ODR Hash code.
/// \brief Grabs the string in whichever attribute is passed in and returns
/// a reference to it. Returns "" if the attribute doesn't exist.
@ -976,100 +976,6 @@ static StringRef getDIEStringAttr(DIE *Die, unsigned Attr) {
return StringRef("");
}
/// \brief Adds the string in \p Str to the hash in \p Hash. This also hashes
/// a trailing NULL with the string.
static void addStringToHash(MD5 &Hash, StringRef Str) {
DEBUG(dbgs() << "Adding string " << Str << " to hash.\n");
Hash.update(Str);
Hash.update(makeArrayRef((uint8_t)'\0'));
}
// FIXME: These are copied and only slightly modified out of LEB128.h.
/// \brief Adds the unsigned in \p N to the hash in \p Hash. This also encodes
/// the unsigned as a ULEB128.
static void addULEB128ToHash(MD5 &Hash, uint64_t Value) {
DEBUG(dbgs() << "Adding ULEB128 " << Value << " to hash.\n");
do {
uint8_t Byte = Value & 0x7f;
Value >>= 7;
if (Value != 0)
Byte |= 0x80; // Mark this byte to show that more bytes will follow.
Hash.update(Byte);
} while (Value != 0);
}
/// \brief Including \p Parent adds the context of Parent to \p Hash.
static void addParentContextToHash(MD5 &Hash, DIE *Parent) {
DEBUG(dbgs() << "Adding parent context to hash...\n");
// [7.27.2] For each surrounding type or namespace beginning with the
// outermost such construct...
SmallVector<DIE *, 1> Parents;
while (Parent->getTag() != dwarf::DW_TAG_compile_unit) {
Parents.push_back(Parent);
Parent = Parent->getParent();
}
// Reverse iterate over our list to go from the outermost construct to the
// innermost.
for (SmallVectorImpl<DIE *>::reverse_iterator I = Parents.rbegin(),
E = Parents.rend();
I != E; ++I) {
DIE *Die = *I;
// ... Append the letter "C" to the sequence...
addULEB128ToHash(Hash, 'C');
// ... Followed by the DWARF tag of the construct...
addULEB128ToHash(Hash, Die->getTag());
// ... Then the name, taken from the DW_AT_name attribute.
StringRef Name = getDIEStringAttr(Die, dwarf::DW_AT_name);
DEBUG(dbgs() << "... adding context: " << Name << "\n");
if (!Name.empty())
addStringToHash(Hash, Name);
}
}
/// This is based on the type signature computation given in section 7.27 of the
/// DWARF4 standard. It is the md5 hash of a flattened description of the DIE with
/// the exception that we are hashing only the context and the name of the type.
static void addDIEODRSignature(MD5 &Hash, CompileUnit *CU, DIE *Die) {
// Add the contexts to the hash. We won't be computing the ODR hash for
// function local types so it's safe to use the generic context hashing
// algorithm here.
// FIXME: If we figure out how to account for linkage in some way we could
// actually do this with a slight modification to the parent hash algorithm.
DIE *Parent = Die->getParent();
if (Parent)
addParentContextToHash(Hash, Parent);
// Add the current DIE information.
// Add the DWARF tag of the DIE.
addULEB128ToHash(Hash, Die->getTag());
// Add the name of the type to the hash.
addStringToHash(Hash, getDIEStringAttr(Die, dwarf::DW_AT_name));
// Now get the result.
MD5::MD5Result Result;
Hash.final(Result);
// ... take the least significant 8 bytes and store those as the attribute.
// Our MD5 implementation always returns its results in little endian, swap
// bytes appropriately.
uint64_t Signature = *reinterpret_cast<support::ulittle64_t *>(Result + 8);
// FIXME: This should be added onto the type unit, not the type, but this
// works as an intermediate stage.
CU->addUInt(Die, dwarf::DW_AT_GNU_odr_signature, dwarf::DW_FORM_data8,
Signature);
}
/// Return true if the current DIE is contained within an anonymous namespace.
static bool isContainedInAnonNamespace(DIE *Die) {
DIE *Parent = Die->getParent();
@ -1090,7 +996,7 @@ static bool shouldAddODRHash(CompileUnit *CU, DIE *Die) {
return CU->getLanguage() == dwarf::DW_LANG_C_plus_plus &&
getDIEStringAttr(Die, dwarf::DW_AT_name) != "" &&
!isContainedInAnonNamespace(Die);
}
}
void DwarfDebug::finalizeModuleInfo() {
// Collect info for variables that were optimized out.
@ -1111,12 +1017,16 @@ void DwarfDebug::finalizeModuleInfo() {
// out type.
// FIXME: Do type splitting.
for (unsigned i = 0, e = TypeUnits.size(); i != e; ++i) {
MD5 Hash;
DIE *Die = TypeUnits[i];
DIEHash Hash;
// If we've requested ODR hashes and it's applicable for an ODR hash then
// add the ODR signature now.
// FIXME: This should be added onto the type unit, not the type, but this
// works as an intermediate stage.
if (GenerateODRHash && shouldAddODRHash(CUMap.begin()->second, Die))
addDIEODRSignature(Hash, CUMap.begin()->second, Die);
CUMap.begin()->second->addUInt(Die, dwarf::DW_AT_GNU_odr_signature,
dwarf::DW_FORM_data8,
Hash.computeDIEODRSignature(Die));
}
// Compute DIE offsets and sizes.