intel/llvm
1
0
Fork 0
mirror of https://github.com/intel/llvm.git synced 2026-01-26 03:56:16 +08:00
Code Issues Packages Projects Releases Wiki Activity

COFF: Fix ICF correctness bug.

Browse Source 
When comparing two COMDAT sections, we need to take section values
and associative sections into account. This patch fixes that bug.
It fixes a crash bug of llvm-tblgen when linked with /opt:lldicf.

One thing I don't understand yet is that this logic seems to be
too strict. MSVC linker is able to create more compact executables
(which of course work correctly). With this ICF algorithm, LLD is
able to make executable smaller, but the outputs are larger than
MSVC's. There must be something I'm missing here.

llvm-svn: 240897
This commit is contained in:
Rui Ueyama
2015-06-28 01:30:54 +00:00
parent 45c4812851
commit 871847e32d
6 changed files with 98 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
lld/COFF/Chunks.cpp
View File

@@ -173,6 +173,13 @@ bool SectionChunk::equals(const SectionChunk *X) const {
if (getContents() != X->getContents())
return false;
// Compare associative sections
if (AssocChildren.size() != X->AssocChildren.size())
return false;
for (size_t I = 0, E = AssocChildren.size(); I != E; ++I)
if (AssocChildren[I]->Ptr != X->AssocChildren[I]->Ptr)
return false;
// Compare relocations
auto Eq = [&](const coff_relocation &R1, const coff_relocation &R2) {
if (R1.Type != R2.Type)
@@ -181,11 +188,13 @@ bool SectionChunk::equals(const SectionChunk *X) const {
return false;
SymbolBody *B1 = File->getSymbolBody(R1.SymbolTableIndex);
SymbolBody *B2 = X->File->getSymbolBody(R2.SymbolTableIndex);
if (B1 == B2)
return true;
auto *D1 = dyn_cast<DefinedRegular>(B1);
auto *D2 = dyn_cast<DefinedRegular>(B2);
if (D1 && D2 && D1->getChunk() == D2->getChunk())
return true;
return B1 == B2;
return (D1 && D2 &&
D1->getValue() == D2->getValue() &&
D1->getChunk() == D2->getChunk());
};
return std::equal(Relocs.begin(), Relocs.end(), X->Relocs.begin(), Eq);
}