replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Revert r70075 and r70078, which reorganized the PCH on-disk hash table 

for identifiers to separate "interesting" from "uninteresting"
identifiers. However, to cope with compiler invocations where the
predefines buffers mismatch, we need to be able to search the complete
identifier table. Cocoa.h.pch is now about 500k larger that it used to
be :(

llvm-svn: 70320
This commit is contained in:
Douglas Gregor 2009-04-28 20:01:51 +00:00
parent 3d948164f7
commit ab4df58193
2 changed files with 12 additions and 76 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

49
clang/lib/Frontend/PCHReader.cpp
View File

@ -1943,47 +1943,14 @@ IdentifierInfo *PCHReader::DecodeIdentifierInfo(unsigned ID) {
uint32_t Offset = IdentifierOffsets[ID - 1];
const char *Str = IdentifierTableData + Offset;
// If there is an identifier lookup table, but the offset of this
// string is after the identifier table itself, then we know that
// this string is not in the on-disk hash table. Therefore,
// disable lookup into the hash table when looking for this
// identifier.
PCHIdentifierLookupTable *IdTable
= (PCHIdentifierLookupTable *)IdentifierLookupTable;
if (!IdTable ||
Offset >= uint32_t(IdTable->getBuckets() - IdTable->getBase())) {
// Turn off lookup into the on-disk hash table. We know that
// this identifier is not there.
if (IdTable)
PP.getIdentifierTable().setExternalIdentifierLookup(0);
// All of the strings in the PCH file are preceded by a 16-bit
// length. Extract that 16-bit length to avoid having to execute
// strlen().
const char *StrLenPtr = Str - 2;
unsigned StrLen = (((unsigned) StrLenPtr[0])
| (((unsigned) StrLenPtr[1]) << 8)) - 1;
IdentifiersLoaded[ID - 1]=&PP.getIdentifierTable().get(Str, Str + StrLen);
// Turn on lookup into the on-disk hash table, if we have an
// on-disk hash table.
if (IdTable)
PP.getIdentifierTable().setExternalIdentifierLookup(this);
} else {
// The identifier is a key in our on-disk hash table. Since we
// know where the hash table entry starts, just read in this
// (key, value) pair.
PCHIdentifierLookupTrait Trait(const_cast<PCHReader &>(*this));
const unsigned char *Pos = (const unsigned char *)Str - 4;
std::pair<unsigned, unsigned> KeyDataLengths
= Trait.ReadKeyDataLength(Pos);
PCHIdentifierLookupTrait::internal_key_type InternalKey
= Trait.ReadKey(Pos, KeyDataLengths.first);
Pos = (const unsigned char *)Str + KeyDataLengths.first;
IdentifiersLoaded[ID - 1] = Trait.ReadData(InternalKey, Pos,
KeyDataLengths.second);
}
// All of the strings in the PCH file are preceded by a 16-bit
// length. Extract that 16-bit length to avoid having to execute
// strlen().
const char *StrLenPtr = Str - 2;
unsigned StrLen = (((unsigned) StrLenPtr[0])
| (((unsigned) StrLenPtr[1]) << 8)) - 1;
IdentifiersLoaded[ID - 1]
= &PP.getIdentifierTable().get(Str, Str + StrLen);
}
return IdentifiersLoaded[ID - 1];

39
clang/lib/Frontend/PCHWriter.cpp
View File

@ -1318,12 +1318,10 @@ public:
DEnd = IdentifierResolver::end();
D != DEnd; ++D)
DataLen += sizeof(pch::DeclID);
// We emit the key length after the data length so that the
// "uninteresting" identifiers following the identifier hash table
// structure will have the same (key length, key characters)
// layout as the keys in the hash table. This also matches the
// format for identifiers in pretokenized headers.
clang::io::Emit16(Out, DataLen);
// We emit the key length after the data length so that every
// string is preceded by a 16-bit length. This matches the PTH
// format for storing identifiers.
clang::io::Emit16(Out, KeyLen);
return std::make_pair(KeyLen, DataLen);
}
@ -1384,33 +1382,12 @@ void PCHWriter::WriteIdentifierTable(Preprocessor &PP) {
{
OnDiskChainedHashTableGenerator<PCHIdentifierTableTrait> Generator;
llvm::SmallVector<const IdentifierInfo *, 32> UninterestingIdentifiers;
// Create the on-disk hash table representation.
for (llvm::DenseMap<const IdentifierInfo *, pch::IdentID>::iterator
ID = IdentifierIDs.begin(), IDEnd = IdentifierIDs.end();
ID != IDEnd; ++ID) {
assert(ID->first && "NULL identifier in identifier table");
// Classify each identifier as either "interesting" or "not
// interesting". Interesting identifiers are those that have
// additional information that needs to be read from the PCH
// file, e.g., a built-in ID, declaration chain, or macro
// definition. These identifiers are placed into the hash table
// so that they can be found when looked up in the user program.
// All other identifiers are "uninteresting", which means that
// the IdentifierInfo built by default has all of the
// information we care about. Such identifiers are placed after
// the hash table.
const IdentifierInfo *II = ID->first;
if (II->isPoisoned() ||
II->isExtensionToken() ||
II->hasMacroDefinition() ||
II->getObjCOrBuiltinID() ||
II->getFETokenInfo<void>())
Generator.insert(ID->first, ID->second);
else
UninterestingIdentifiers.push_back(II);
Generator.insert(ID->first, ID->second);
}
// Create the on-disk hash table in a buffer.
@ -1422,14 +1399,6 @@ void PCHWriter::WriteIdentifierTable(Preprocessor &PP) {
// Make sure that no bucket is at offset 0
clang::io::Emit32(Out, 0);
BucketOffset = Generator.Emit(Out, Trait);
for (unsigned I = 0, N = UninterestingIdentifiers.size(); I != N; ++I) {
const IdentifierInfo *II = UninterestingIdentifiers[I];
unsigned N = II->getLength() + 1;
clang::io::Emit16(Out, N);
SetIdentifierOffset(II, Out.tell());
Out.write(II->getName(), N);
}
}
// Create a blob abbreviation