[llvm] [ThinLTO][WPD] Suppress WPD on a class if any superclass is visible in a shared library for index-based WPD (PR #131721)
via llvm-commits
llvm-commits at lists.llvm.org
Mon Mar 17 20:58:42 PDT 2025
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-lto
Author: Mingming Liu (mingmingl-llvm)
<details>
<summary>Changes</summary>
Before this patch, whole program devirtualization is suppressed on a class if any superclass is visible to regular object files, by recording the class GUID in `VisibleToRegularObjSymbols`.
This patch suppresses whole program devirtualization on a class if any superclass is visible (or escape to shared library) for index-based WPD.
Implementation summaries:
1. In llvm/lib/LTO/LTO.cpp, `IsVisibleToRegularObj` is updated to look at the global resolution's `ExportDynamic` bit.
2. In llvm/tools/llvm-lto2/llvm-lto2.cpp,
- three command line options are added so `llvm-lto2` can override `Conf.HasWholeProgramVisibility`, `Conf.ValidateAllVtablesHaveTypeInfos` and `Conf.AllVtablesHaveTypeInfos`.
- `-r` parses `d` as `ExportDynamic`
The test case is reduced from a small C++ program (main.cc, lib.cc/h pasted below in [1]). To reproduce the program failure without this patch, compile lib.cc into a shared library, and provide it to a ThinLTO build of main.cc (commands are pasted in [2]).
[1]
* lib.h
```
#include <cstdio>
class Derived {
public:
void dispatch();
virtual void print();
virtual void sum();
};
void Derived::dispatch() {
static_cast<Derived*>(this)->print();
static_cast<Derived*>(this)->sum();
}
void Derived::sum() {
printf("Derived::sum\n");
}
__attribute__((noinline)) void* create(int i);
__attribute__((noinline)) void* getPtr(int i);
```
* lib.cc
```
#include "lib.h"
#include <cstdio>
#include <iostream>
class Derived2 : public Derived {
public:
void print() override {
printf("DerivedSharedLib\n");
}
void sum() override {
printf("DerivedSharedLib::sum\n");
}
};
void Derived::print() {
printf("Derived\n");
}
__attribute__((noinline)) void* create(int i) {
if (i & 1)
return new Derived2();
return new Derived();
}
```
* main.cc
```
cat main.cc
#include "lib.h"
class DerivedN : public Derived {
public:
};
__attribute__((noinline)) void* getPtr(int x) {
return new DerivedN();
}
int main() {
Derived*b = static_cast<Derived*>(create(201));
b->dispatch();
delete b;
Derived* a = static_cast<Derived*>(getPtr(202));
a->dispatch();
delete a;
return 0;
}
```
[2]
```
# compile lib.o in a shared library.
$ ./bin/clang++ -O2 -fPIC -c lib.cc -o lib.o
$ ./bin/clang++ -shared -o libdata.so lib.o
# Provide the shared library in `-ldata`
$ ./bin/clang++ -v -g -ldata --save-temps -fno-discard-value-names -Wl,-mllvm,-print-before=wholeprogramdevirt -Wl,-mllvm,-wholeprogramdevirt-check=trap -Rpass=wholeprogramdevirt -Wl,--lto-whole-program-visibility -Wl,--lto-validate-all-vtables-have-type-infos -mllvm -disable-icp=true -Wl,-mllvm,-disable-icp=false -flto=thin -fwhole-program-vtables -fno-split-lto-unit -fuse-ld=lld main.cc -L . -o main >/tmp/wholeprogramdevirt.ir 2>&1
# Run the program hits a segmentation fault with `-Wl,-mllvm,-wholeprogramdevirt-check=trap`
$ LD_LIBRARY_PATH=. ./main
DerivedSharedLib
Trace/breakpoint trap (core dumped)
```
---
Full diff: https://github.com/llvm/llvm-project/pull/131721.diff
3 Files Affected:
- (modified) llvm/lib/LTO/LTO.cpp (+1-1)
- (added) llvm/test/ThinLTO/X86/nodevirt_exort_dynamic.ll (+111)
- (modified) llvm/tools/llvm-lto2/llvm-lto2.cpp (+21)
``````````diff
diff --git a/llvm/lib/LTO/LTO.cpp b/llvm/lib/LTO/LTO.cpp
index e895a46b8cd77..6ec2dc381a932 100644
--- a/llvm/lib/LTO/LTO.cpp
+++ b/llvm/lib/LTO/LTO.cpp
@@ -1905,7 +1905,7 @@ Error LTO::runThinLTO(AddStreamFn AddStream, FileCache Cache,
auto IsVisibleToRegularObj = [&](StringRef name) {
auto It = GlobalResolutions->find(name);
return (It == GlobalResolutions->end() ||
- It->second.VisibleOutsideSummary);
+ It->second.VisibleOutsideSummary || It->second.ExportDynamic);
};
getVisibleToRegularObjVtableGUIDs(ThinLTO.CombinedIndex,
diff --git a/llvm/test/ThinLTO/X86/nodevirt_exort_dynamic.ll b/llvm/test/ThinLTO/X86/nodevirt_exort_dynamic.ll
new file mode 100644
index 0000000000000..ea33291a75161
--- /dev/null
+++ b/llvm/test/ThinLTO/X86/nodevirt_exort_dynamic.ll
@@ -0,0 +1,111 @@
+; RUN: rm -rf %t && mkdir %t && cd %t
+
+; Generate unsplit module with summary for ThinLTO index-based WPD.
+; RUN: opt -thinlto-bc -o summary.o %s
+
+; RUN: llvm-dis -o - summary.o
+
+;; TODO: Implement the fix for WPD in regular or hybrid LTO, and add test coverage.
+
+; Index based WPD
+; For `_ZTI7Derived`, the 'llvm-lto2' resolution arguments specifies `VisibleOutsideSummary` as false
+; and `ExportDynamic` as false. The callsite inside @_ZN4Base8dispatchEv
+; got devirtualized.
+; RUN: llvm-lto2 run summary.o -save-temps -pass-remarks=. \
+; RUN: -o tmp \
+; RUN: --whole-program-visibility-enabled-in-lto=true \
+; RUN: --validate-all-vtables-have-type-infos=true \
+; RUN: --all-vtables-have-type-infos=true \
+; RUN: -r=summary.o,__cxa_pure_virtual, \
+; RUN: -r=summary.o,_ZN8DerivedNC2Ev,x \
+; RUN: -r=summary.o,_ZN4Base8dispatchEv,px \
+; RUN: -r=summary.o,_ZN7DerivedC2Ev, \
+; RUN: -r=summary.o,_ZN8DerivedN5printEv,px \
+; RUN: -r=summary.o,_ZTS4Base, \
+; RUN: -r=summary.o,_ZTV8DerivedN,p \
+; RUN: -r=summary.o,_ZTI8DerivedN,p \
+; RUN: -r=summary.o,_ZTI4Base, \
+; RUN: -r=summary.o,_ZTS8DerivedN,p \
+; RUN: -r=summary.o,_ZTI7Derived, \
+; RUN: -r=summary.o,_ZTV4Base 2>&1 | FileCheck --allow-empty %s --check-prefix=REMARK
+
+; REMARK: single-impl: devirtualized a call to _ZN8DerivedN5printEv
+
+; Index based WPD
+; For `_ZTI7Derived`, the 'llvm-lto2' resolution arguments specifies `VisibleOutsideSummary` as false
+; and `ExportDynamic` as true. The callsite inside @_ZN4Base8dispatchEv won't
+; get devirtualized.
+; RUN: llvm-lto2 run summary.o -save-temps -pass-remarks=. \
+; RUN: -o tmp \
+; RUN: --whole-program-visibility-enabled-in-lto=true \
+; RUN: --validate-all-vtables-have-type-infos=true \
+; RUN: --all-vtables-have-type-infos=true \
+; RUN: -r=summary.o,__cxa_pure_virtual, \
+; RUN: -r=summary.o,_ZN8DerivedNC2Ev,x \
+; RUN: -r=summary.o,_ZN4Base8dispatchEv,px \
+; RUN: -r=summary.o,_ZN7DerivedC2Ev, \
+; RUN: -r=summary.o,_ZN8DerivedN5printEv,px \
+; RUN: -r=summary.o,_ZTS4Base, \
+; RUN: -r=summary.o,_ZTV8DerivedN,p \
+; RUN: -r=summary.o,_ZTI8DerivedN,p \
+; RUN: -r=summary.o,_ZTI4Base, \
+; RUN: -r=summary.o,_ZTS8DerivedN,p \
+; RUN: -r=summary.o,_ZTI7Derived,d \
+; RUN: -r=summary.o,_ZTV4Base 2>&1 | FileCheck %s --allow-empty --implicit-check-not='single-impl: devirtualized a call to'
+
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+ at _ZTV8DerivedN = linkonce_odr hidden constant { [3 x ptr] } { [3 x ptr] [ptr null, ptr @_ZTI8DerivedN, ptr @_ZN8DerivedN5printEv] }, !type !0, !type !1, !type !2, !type !3, !type !4, !type !5, !vcall_visibility !6
+ at _ZTI8DerivedN = linkonce_odr hidden constant { ptr, ptr, ptr } { ptr null, ptr @_ZTS8DerivedN, ptr @_ZTI7Derived }
+ at _ZTS8DerivedN = linkonce_odr hidden constant [10 x i8] c"8DerivedN\00", align 1
+ at _ZTI7Derived = external constant ptr
+ at _ZTV4Base = linkonce_odr hidden constant { [3 x ptr] } { [3 x ptr] [ptr null, ptr @_ZTI4Base, ptr @__cxa_pure_virtual] }, !type !0, !type !1, !vcall_visibility !6
+ at _ZTI4Base = linkonce_odr hidden constant { ptr, ptr } { ptr null, ptr @_ZTS4Base }
+ at _ZTS4Base = linkonce_odr hidden constant [6 x i8] c"4Base\00", align 1
+
+ at llvm.used = appending global [1 x ptr] [ptr @_ZN8DerivedNC2Ev], section "llvm.metadata"
+
+define hidden void @_ZN4Base8dispatchEv(ptr %this) {
+entry:
+ %this.addr = alloca ptr
+ store ptr %this, ptr %this.addr
+ %this1 = load ptr, ptr %this.addr
+ %vtable = load ptr, ptr %this1
+ %0 = call i1 @llvm.type.test(ptr %vtable, metadata !"_ZTS7Derived")
+ call void @llvm.assume(i1 %0)
+ %vfn = getelementptr inbounds ptr, ptr %vtable, i64 0
+ %1 = load ptr, ptr %vfn
+ call void %1(ptr %this1)
+ ret void
+}
+
+define linkonce_odr hidden void @_ZN8DerivedNC2Ev(ptr %this) #0 {
+entry:
+ %this.addr = alloca ptr
+ store ptr %this, ptr %this.addr
+ %this1 = load ptr, ptr %this.addr
+ call void @_ZN7DerivedC2Ev(ptr %this1)
+ store ptr getelementptr inbounds inrange(-16, 8) ({ [3 x ptr] }, ptr @_ZTV8DerivedN, i32 0, i32 0, i32 2), ptr %this1
+ ret void
+}
+
+define linkonce_odr hidden void @_ZN8DerivedN5printEv(ptr %this) #0 {
+entry:
+ ret void
+}
+
+attributes #0 = { noinline optnone }
+
+declare void @__cxa_pure_virtual()
+declare i1 @llvm.type.test(ptr, metadata)
+declare void @llvm.assume(i1)
+declare void @_ZN7DerivedC2Ev(ptr)
+
+!0 = !{i64 16, !"_ZTS4Base"}
+!1 = !{i64 16, !"_ZTSM4BaseFvvE.virtual"}
+!2 = !{i64 16, !"_ZTS7Derived"}
+!3 = !{i64 16, !"_ZTSM7DerivedFvvE.virtual"}
+!4 = !{i64 16, !"_ZTS8DerivedN"}
+!5 = !{i64 16, !"_ZTSM8DerivedNFvvE.virtual"}
+!6 = !{i64 0}
diff --git a/llvm/tools/llvm-lto2/llvm-lto2.cpp b/llvm/tools/llvm-lto2/llvm-lto2.cpp
index dbb7f2f3028aa..0317a610a25a3 100644
--- a/llvm/tools/llvm-lto2/llvm-lto2.cpp
+++ b/llvm/tools/llvm-lto2/llvm-lto2.cpp
@@ -187,6 +187,18 @@ static cl::opt<bool> EnableFreestanding(
cl::desc("Enable Freestanding (disable builtins / TLI) during LTO"),
cl::Hidden);
+static cl::opt<bool> WholeProgramVisibilityEnabledInLTO(
+ "whole-program-visibility-enabled-in-lto",
+ cl::desc("Enable whole program visibility during LTO"), cl::Hidden);
+
+static cl::opt<bool> ValidateAllVtablesHaveTypeInfos(
+ "validate-all-vtables-have-type-infos",
+ cl::desc("Validate that all vtables have type infos in LTO"), cl::Hidden);
+
+static cl::opt<bool>
+ AllVtablesHaveTypeInfos("all-vtables-have-type-infos", cl::Hidden,
+ cl::desc("All vtables have type infos"));
+
extern cl::opt<cl::boolOrDefault> LoadBitcodeIntoNewDbgInfoFormat;
extern cl::opt<cl::boolOrDefault> PreserveInputDbgFormat;
@@ -257,6 +269,8 @@ static int run(int argc, char **argv) {
Res.VisibleToRegularObj = true;
else if (C == 'r')
Res.LinkerRedefined = true;
+ else if (C == 'd')
+ Res.ExportDynamic = true;
else {
llvm::errs() << "invalid character " << C << " in resolution: " << R
<< '\n';
@@ -332,6 +346,13 @@ static int run(int argc, char **argv) {
Conf.PTO.LoopVectorization = Conf.OptLevel > 1;
Conf.PTO.SLPVectorization = Conf.OptLevel > 1;
+ if (WholeProgramVisibilityEnabledInLTO.getNumOccurrences() > 0)
+ Conf.HasWholeProgramVisibility = WholeProgramVisibilityEnabledInLTO;
+ if (ValidateAllVtablesHaveTypeInfos.getNumOccurrences() > 0)
+ Conf.ValidateAllVtablesHaveTypeInfos = ValidateAllVtablesHaveTypeInfos;
+ if (AllVtablesHaveTypeInfos.getNumOccurrences() > 0)
+ Conf.AllVtablesHaveTypeInfos = AllVtablesHaveTypeInfos;
+
ThinBackend Backend;
if (ThinLTODistributedIndexes)
Backend = createWriteIndexesThinBackend(llvm::hardware_concurrency(Threads),
``````````
</details>
https://github.com/llvm/llvm-project/pull/131721
More information about the llvm-commits
mailing list