[lld] [lld] Merge equivalent symbols found during ICF (PR #134342)
Pranav Kant via llvm-commits
llvm-commits at lists.llvm.org
Thu Apr 3 21:06:32 PDT 2025
https://github.com/pranavk updated https://github.com/llvm/llvm-project/pull/134342
>From 07701ce112b5a6dcd5f5284d2d811762c9abce93 Mon Sep 17 00:00:00 2001
From: Pranav Kant <prka at google.com>
Date: Fri, 4 Apr 2025 03:44:16 +0000
Subject: [PATCH 1/2] [lld] Merge equivalent symbols found during ICF
---
lld/ELF/ICF.cpp | 51 ++++++++++++++++++-
lld/ELF/SymbolTable.cpp | 7 +++
lld/ELF/SymbolTable.h | 1 +
lld/test/ELF/aarch64-got-merging-icf.s | 68 ++++++++++++++++++++++++++
lld/test/ELF/icf-preemptible.s | 3 ++
5 files changed, 129 insertions(+), 1 deletion(-)
create mode 100644 lld/test/ELF/aarch64-got-merging-icf.s
diff --git a/lld/ELF/ICF.cpp b/lld/ELF/ICF.cpp
index 1cdcf6be9d8a9..467487e10f310 100644
--- a/lld/ELF/ICF.cpp
+++ b/lld/ELF/ICF.cpp
@@ -333,6 +333,28 @@ bool ICF<ELFT>::equalsConstant(const InputSection *a, const InputSection *b) {
: constantEq(a, ra.relas, b, rb.relas);
}
+template <class RelTy>
+static SmallVector<Symbol *> getReloc(const InputSection *sec,
+ Relocs<RelTy> relocs) {
+ SmallVector<Symbol *> syms;
+ for (auto ri = relocs.begin(), re = relocs.end(); ri != re; ++ri) {
+ Symbol &sym = sec->file->getRelocTargetSym(*ri);
+ syms.push_back(&sym);
+ }
+ return syms;
+}
+
+template <class ELFT>
+static SmallVector<Symbol *> getRelocTargetSyms(const InputSection *sec) {
+ const RelsOrRelas<ELFT> rel = sec->template relsOrRelas<ELFT>();
+ if (rel.areRelocsCrel())
+ return getReloc(sec, rel.crels);
+ if (rel.areRelocsRel())
+ return getReloc(sec, rel.rels);
+
+ return getReloc(sec, rel.relas);
+}
+
// Compare two lists of relocations. Returns true if all pairs of
// relocations point to the same section in terms of ICF.
template <class ELFT>
@@ -535,14 +557,35 @@ template <class ELFT> void ICF<ELFT>::run() {
auto print = [&ctx = ctx]() -> ELFSyncStream {
return {ctx, ctx.arg.printIcfSections ? DiagLevel::Msg : DiagLevel::None};
};
+
+ DenseMap<Symbol *, Symbol *> symbolMap;
// Merge sections by the equivalence class.
+ // Merge symbols identified as equivalent during ICF
forEachClassRange(0, sections.size(), [&](size_t begin, size_t end) {
if (end - begin == 1)
return;
print() << "selected section " << sections[begin];
+ SmallVector<Symbol *> syms = getRelocTargetSyms<ELFT>(sections[begin]);
for (size_t i = begin + 1; i < end; ++i) {
print() << " removing identical section " << sections[i];
sections[begin]->replace(sections[i]);
+ SmallVector<Symbol *> replacedSyms =
+ getRelocTargetSyms<ELFT>(sections[i]);
+ assert(syms.size() == replacedSyms.size() &&
+ "Should have same number of syms!");
+ for (size_t i = 0; i < syms.size(); i++) {
+ if (syms[i] == replacedSyms[i] || !syms[i]->isGlobal() ||
+ !replacedSyms[i]->isGlobal())
+ continue;
+ auto [it, inserted] =
+ symbolMap.insert(std::make_pair(replacedSyms[i], syms[i]));
+ print() << " selected symbol: " << syms[i]->getName().data()
+ << "; replaced symbol: " << replacedSyms[i]->getName().data();
+ if (!inserted) {
+ print() << " replacement already exists: "
+ << it->getSecond()->getName().data();
+ }
+ }
// At this point we know sections merged are fully identical and hence
// we want to remove duplicate implicit dependencies such as link order
@@ -561,11 +604,17 @@ template <class ELFT> void ICF<ELFT>::run() {
d->folded = true;
}
};
- for (Symbol *sym : ctx.symtab->getSymbols())
+ for (Symbol *sym : ctx.symtab->getSymbols()) {
fold(sym);
+ if (Symbol *s = symbolMap.lookup(sym))
+ ctx.symtab->redirect(sym, s);
+ }
parallelForEach(ctx.objectFiles, [&](ELFFileBase *file) {
for (Symbol *sym : file->getLocalSymbols())
fold(sym);
+ for (Symbol *&sym : file->getMutableGlobalSymbols())
+ if (Symbol *s = symbolMap.lookup(sym))
+ sym = s;
});
// InputSectionDescription::sections is populated by processSectionCommands().
diff --git a/lld/ELF/SymbolTable.cpp b/lld/ELF/SymbolTable.cpp
index b8a70d4e898fc..2199f159692b0 100644
--- a/lld/ELF/SymbolTable.cpp
+++ b/lld/ELF/SymbolTable.cpp
@@ -29,6 +29,13 @@ using namespace llvm::ELF;
using namespace lld;
using namespace lld::elf;
+void SymbolTable::redirect(Symbol *from, Symbol *to) {
+ int &fromIdx = symMap[CachedHashStringRef(from->getName())];
+ const int toIdx = symMap[CachedHashStringRef(to->getName())];
+
+ fromIdx = toIdx;
+}
+
void SymbolTable::wrap(Symbol *sym, Symbol *real, Symbol *wrap) {
// Redirect __real_foo to the original foo and foo to the original __wrap_foo.
int &idx1 = symMap[CachedHashStringRef(sym->getName())];
diff --git a/lld/ELF/SymbolTable.h b/lld/ELF/SymbolTable.h
index d6443742f7baa..e3a39bac85f97 100644
--- a/lld/ELF/SymbolTable.h
+++ b/lld/ELF/SymbolTable.h
@@ -41,6 +41,7 @@ class SymbolTable {
SymbolTable(Ctx &ctx) : ctx(ctx) {}
ArrayRef<Symbol *> getSymbols() const { return symVector; }
+ void redirect(Symbol *from, Symbol *to);
void wrap(Symbol *sym, Symbol *real, Symbol *wrap);
Symbol *insert(StringRef name);
diff --git a/lld/test/ELF/aarch64-got-merging-icf.s b/lld/test/ELF/aarch64-got-merging-icf.s
new file mode 100644
index 0000000000000..9f359cbf3a0a9
--- /dev/null
+++ b/lld/test/ELF/aarch64-got-merging-icf.s
@@ -0,0 +1,68 @@
+// REQUIRES: aarch64
+
+# RUN: llvm-mc -filetype=obj -triple=aarch64 %s -o %t
+# RUN: ld.lld %t -o %t2 --icf=all
+# RUN: llvm-objdump --section-headers %t2 | FileCheck %s --check-prefix=EXE
+
+# RUN: ld.lld -shared %t -o %t3 --icf=all
+# RUN: llvm-objdump --section-headers %t3 | FileCheck %s --check-prefix=DSO
+
+## All .rodata.* sections should merge into a single GOT entry
+# EXE: {{.*}}.got 00000008{{.*}}
+
+## When symbols are preemptible in DSO mode, GOT entries wouldn't be merged
+# DSO: {{.*}}.got 00000020{{.*}}
+
+.addrsig
+
+callee:
+ret
+
+.section .rodata.dummy1,"a", at progbits
+sym1:
+.long 111
+.long 122
+.byte 123
+
+.section .rodata.dummy2,"a", at progbits
+sym2:
+.long 111
+.long 122
+sym3:
+.byte 123
+
+.macro f, index
+
+.section .text.f1_\index,"ax", at progbits
+f1_\index:
+adrp x0, :got:g\index
+mov x1, #\index
+b f2_\index
+
+.section .text.f2_\index,"ax", at progbits
+f2_\index:
+ldr x0, [x0, :got_lo12:g\index]
+b callee
+
+.globl g\index
+.section .rodata.g\index,"a", at progbits
+g_\index:
+.long 111
+.long 122
+
+g\index:
+.byte 123
+
+.section .text._start,"ax", at progbits
+bl f1_\index
+
+.endm
+
+.section .text._start,"ax", at progbits
+.globl _start
+_start:
+
+f 0
+f 1
+f 2
+f 3
diff --git a/lld/test/ELF/icf-preemptible.s b/lld/test/ELF/icf-preemptible.s
index 4bd1eca438b19..f79cf73b911ba 100644
--- a/lld/test/ELF/icf-preemptible.s
+++ b/lld/test/ELF/icf-preemptible.s
@@ -11,12 +11,15 @@
# EXE-NOT: {{.}}
# EXE: selected section {{.*}}:(.text.g1)
# EXE-NEXT: removing identical section {{.*}}:(.text.g2)
+# EXE-NEXT: selected symbol: f1; replaced symbol: f2
# EXE-NEXT: removing identical section {{.*}}:(.text.g3)
# EXE-NEXT: selected section {{.*}}:(.text.f1)
# EXE-NEXT: removing identical section {{.*}}:(.text.f2)
# EXE-NEXT: selected section {{.*}}:(.text.h1)
# EXE-NEXT: removing identical section {{.*}}:(.text.h2)
+# EXE-NEXT: selected symbol: g1; replaced symbol: g2
# EXE-NEXT: removing identical section {{.*}}:(.text.h3)
+# EXE-NEXT: selected symbol: g1; replaced symbol: g3
# EXE-NOT: {{.}}
## Definitions are preemptible in a DSO. Only leaf functions can be folded.
>From 7e2cfb4ab12391378199bfb3fb2bbb433f8e59cf Mon Sep 17 00:00:00 2001
From: Pranav Kant <prka at google.com>
Date: Fri, 4 Apr 2025 04:06:17 +0000
Subject: [PATCH 2/2] add comments to test case
---
lld/test/ELF/aarch64-got-merging-icf.s | 3 +++
1 file changed, 3 insertions(+)
diff --git a/lld/test/ELF/aarch64-got-merging-icf.s b/lld/test/ELF/aarch64-got-merging-icf.s
index 9f359cbf3a0a9..c5717106d582a 100644
--- a/lld/test/ELF/aarch64-got-merging-icf.s
+++ b/lld/test/ELF/aarch64-got-merging-icf.s
@@ -33,17 +33,20 @@ sym3:
.macro f, index
+# (Kept unique) first instruction of the GOT code sequence
.section .text.f1_\index,"ax", at progbits
f1_\index:
adrp x0, :got:g\index
mov x1, #\index
b f2_\index
+# Folded, second instruction of the GOT code sequence
.section .text.f2_\index,"ax", at progbits
f2_\index:
ldr x0, [x0, :got_lo12:g\index]
b callee
+# Folded
.globl g\index
.section .rodata.g\index,"a", at progbits
g_\index:
More information about the llvm-commits
mailing list