caio/clang-p2996
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Create synthetic symbol names on demand to improve memory consumption and startup times.

Browse Source 
This fix was created after profiling the target creation of a large C/C++/ObjC application that contained almost 4,000,000 redacted symbol names. The symbol table parsing code was creating names for each of these synthetic symbols and adding them to the name indexes. The code was also adding the object file basename to the end of the symbol name which doesn't allow symbols from different shared libraries to share the names in the constant string pool.

Prior to this fix this was creating 180MB of "___lldb_unnamed_symbol" symbol names and was taking a long time to generate each name, add them to the string pool and then add each of these names to the name index.

This patch fixes the issue by:
- not adding a name to synthetic symbols at creation time, and allows name to be dynamically generated when accessed
- doesn't add synthetic symbol names to the name indexes, but catches this special case as name lookup time. Users won't typically set breakpoints or lookup these synthetic names, but support was added to do the lookup in case it does happen
- removes the object file baseanme from the generated names to allow the names to be shared in the constant string pool

Prior to this fix the startup times for a large application was:
12.5 seconds (cold file caches)
8.5 seconds (warm file caches)

After this fix:
9.7 seconds (cold file caches)
5.7 seconds (warm file caches)

The names of the symbols are auto generated by appending the symbol's UserID to the end of the "___lldb_unnamed_symbol" string and is only done when the name is requested from a synthetic symbol if it has no name.

Differential Revision: https://reviews.llvm.org/D105160
This commit is contained in:
Greg Clayton
2021-06-29 13:12:36 -07:00
parent 98b9fc9b93
commit c8164d0276
10 changed files with 148 additions and 70 deletions
Download Patch File Download Diff File Expand all files Collapse all files

72
lldb/source/Plugins/ObjectFile/ELF/ObjectFileELF.cpp
View File

@@ -1880,7 +1880,7 @@ void ObjectFileELF::CreateSections(SectionList &unified_section_list) {
unified_section_list.AddSection(symtab_section_sp);
}
}
}
}
}
std::shared_ptr<ObjectFileELF> ObjectFileELF::GetGnuDebugDataObjectFile() {
@@ -2813,31 +2813,37 @@ Symtab *ObjectFileELF::GetSymtab() {
if (is_valid_entry_point && !m_symtab_up->FindSymbolContainingFileAddress(
entry_point_file_addr)) {
uint64_t symbol_id = m_symtab_up->GetNumSymbols();
Symbol symbol(symbol_id,
GetNextSyntheticSymbolName().GetCString(), // Symbol name.
eSymbolTypeCode, // Type of this symbol.
true, // Is this globally visible?
false, // Is this symbol debug info?
false, // Is this symbol a trampoline?
true, // Is this symbol artificial?
entry_point_addr.GetSection(), // Section where this
// symbol is defined.
0, // Offset in section or symbol value.
0, // Size.
false, // Size is valid.
false, // Contains linker annotations?
0); // Symbol flags.
m_symtab_up->AddSymbol(symbol);
// Don't set the name for any synthetic symbols, the Symbol
// object will generate one if needed when the name is accessed
// via accessors.
SectionSP section_sp = entry_point_addr.GetSection();
Symbol symbol(
/*symID=*/symbol_id,
/*name=*/llvm::StringRef(), // Name will be auto generated.
/*type=*/eSymbolTypeCode,
/*external=*/true,
/*is_debug=*/false,
/*is_trampoline=*/false,
/*is_artificial=*/true,
/*section_sp=*/section_sp,
/*offset=*/0,
/*size=*/0, // FDE can span multiple symbols so don't use its size.
/*size_is_valid=*/false,
/*contains_linker_annotations=*/false,
/*flags=*/0);
// When the entry point is arm thumb we need to explicitly set its
// class address to reflect that. This is important because expression
// evaluation relies on correctly setting a breakpoint at this
// address.
if (arch.GetMachine() == llvm::Triple::arm &&
(entry_point_file_addr & 1))
(entry_point_file_addr & 1)) {
symbol.GetAddressRef().SetOffset(entry_point_addr.GetOffset() ^ 1);
m_address_class_map[entry_point_file_addr ^ 1] =
AddressClass::eCodeAlternateISA;
else
} else {
m_address_class_map[entry_point_file_addr] = AddressClass::eCode;
}
m_symtab_up->AddSymbol(symbol);
}
}
@@ -2917,22 +2923,24 @@ void ObjectFileELF::ParseUnwindSymbols(Symtab *symbol_table,
section_list->FindSectionContainingFileAddress(file_addr);
if (section_sp) {
addr_t offset = file_addr - section_sp->GetFileAddress();
const char *symbol_name = GetNextSyntheticSymbolName().GetCString();
uint64_t symbol_id = ++last_symbol_id;
// Don't set the name for any synthetic symbols, the Symbol
// object will generate one if needed when the name is accessed
// via accessors.
Symbol eh_symbol(
symbol_id, // Symbol table index.
symbol_name, // Symbol name.
eSymbolTypeCode, // Type of this symbol.
true, // Is this globally visible?
false, // Is this symbol debug info?
false, // Is this symbol a trampoline?
true, // Is this symbol artificial?
section_sp, // Section in which this symbol is defined or null.
offset, // Offset in section or symbol value.
0, // Size: Don't specify the size as an FDE can
false, // Size is valid: cover multiple symbols.
false, // Contains linker annotations?
0); // Symbol flags.
/*symID=*/symbol_id,
/*name=*/llvm::StringRef(), // Name will be auto generated.
/*type=*/eSymbolTypeCode,
/*external=*/true,
/*is_debug=*/false,
/*is_trampoline=*/false,
/*is_artificial=*/true,
/*section_sp=*/section_sp,
/*offset=*/offset,
/*size=*/0, // FDE can span multiple symbols so don't use its size.
/*size_is_valid=*/false,
/*contains_linker_annotations=*/false,
/*flags=*/0);
new_symbols.push_back(eh_symbol);
}
}