[PowerPC][AIX] Support #pragma comment copyright for AIX#178184
Open
tonykuttai wants to merge 4 commits into
Open
[PowerPC][AIX] Support #pragma comment copyright for AIX#178184tonykuttai wants to merge 4 commits into
tonykuttai wants to merge 4 commits into
Conversation
tonykuttai
force-pushed
the
tvarghese/pragma-copyright-comment-update
branch
from
750cee4 to
527b870
Compare
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
tonykuttai
force-pushed
the
tvarghese/pragma-copyright-comment-update
branch
2 times, most recently
from
552eba6 to
bcb075b
Compare
🐧 Linux x64 Test Results
✅ The build succeeded and all tests passed. |
🪟 Windows x64 Test Results
✅ The build succeeded and all tests passed. |
llvmbot
added
backend:AArch64
backend:Hexagon
backend:PowerPC
clang:frontend
Member
|
@llvm/pr-subscribers-lto @llvm/pr-subscribers-llvm-transforms Author: Tony Varghese (tonykuttai) Changes
This change builds on top of XCOFF associated metadata and was reviewed at [PowerPC][AIX] Support #pragma comment copyright for AIX Patch is 29.81 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/178184.diff 22 Files Affected:
diff --git a/clang/include/clang/Basic/DiagnosticParseKinds.td b/clang/include/clang/Basic/DiagnosticParseKinds.td
index 457d3644de35a..ede6a90a31cc6 100644
--- a/clang/include/clang/Basic/DiagnosticParseKinds.td
+++ b/clang/include/clang/Basic/DiagnosticParseKinds.td
@@ -1358,6 +1358,10 @@ def err_pragma_comment_unknown_kind : Error<"unknown kind of pragma comment">;
// PS4 recognizes only #pragma comment(lib)
def warn_pragma_comment_ignored : Warning<"'#pragma comment %0' ignored">,
InGroup<IgnoredPragmas>;
+def warn_pragma_comment_once
+ : Warning<"'#pragma comment %0' "
+ "can be specified only once per translation unit - ignored">,
+ InGroup<IgnoredPragmas>;
// - #pragma detect_mismatch
def err_pragma_detect_mismatch_malformed : Error<
"pragma detect_mismatch is malformed; it requires two comma-separated "
diff --git a/clang/include/clang/Basic/PragmaKinds.h b/clang/include/clang/Basic/PragmaKinds.h
index 42f049f7323d2..52ca58855d460 100644
--- a/clang/include/clang/Basic/PragmaKinds.h
+++ b/clang/include/clang/Basic/PragmaKinds.h
@@ -17,7 +17,8 @@ enum PragmaMSCommentKind {
PCK_Lib, // #pragma comment(lib, ...)
PCK_Compiler, // #pragma comment(compiler, ...)
PCK_ExeStr, // #pragma comment(exestr, ...)
- PCK_User // #pragma comment(user, ...)
+ PCK_User, // #pragma comment(user, ...)
+ PCK_Copyright // #pragma comment(copyright, ...)
};
enum PragmaMSStructKind {
diff --git a/clang/lib/AST/TextNodeDumper.cpp b/clang/lib/AST/TextNodeDumper.cpp
index 7bc0404db1bee..c58c33f95c193 100644
--- a/clang/lib/AST/TextNodeDumper.cpp
+++ b/clang/lib/AST/TextNodeDumper.cpp
@@ -2526,6 +2526,9 @@ void TextNodeDumper::VisitPragmaCommentDecl(const PragmaCommentDecl *D) {
case PCK_User:
OS << "user";
break;
+ case PCK_Copyright:
+ OS << "copyright";
+ break;
}
StringRef Arg = D->getArg();
if (!Arg.empty())
diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index d50c9605a30b3..e63adc98a95aa 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -1588,6 +1588,12 @@ void CodeGenModule::Release() {
EmitBackendOptionsMetadata(getCodeGenOpts());
+ // Emit copyright metadata
+ if (LoadTimeComment) {
+ auto *NMD = getModule().getOrInsertNamedMetadata("comment_string.loadtime");
+ NMD->addOperand(LoadTimeComment);
+ }
+
// If there is device offloading code embed it in the host now.
EmbedObject(&getModule(), CodeGenOpts, *getFileSystem(), getDiags());
@@ -3481,6 +3487,31 @@ void CodeGenModule::AddDependentLib(StringRef Lib) {
LinkerOptionsMetadata.push_back(llvm::MDNode::get(C, MDOpts));
}
+/// Process AIX copyright pragma and create LLVM metadata.
+/// #pragma comment(copyright, "string") embed copyright
+/// information into the object file's loader section.
+///
+/// Example: #pragma comment(copyright, "Copyright IBM Corp. 2024")
+///
+/// This should only be called once per translation unit.
+void CodeGenModule::ProcessPragmaComment(PragmaMSCommentKind Kind,
+ StringRef Comment) {
+ // Ensure we are only processing Copyright Pragmas
+ assert(Kind == PCK_Copyright &&
+ "Unexpected pragma comment kind, ProcessPragmaComment should only be "
+ "called for PCK_Copyright");
+
+ // Only one copyright pragma allowed per translation unit
+ if (LoadTimeComment) {
+ return;
+ }
+
+ // Create llvm metadata with the comment string
+ auto &C = getLLVMContext();
+ llvm::Metadata *Ops[] = {llvm::MDString::get(C, Comment)};
+ LoadTimeComment = llvm::MDNode::get(C, Ops);
+}
+
/// Add link options implied by the given module, including modules
/// it depends on, using a postorder walk.
static void addLinkOptionsPostorder(CodeGenModule &CGM, Module *Mod,
@@ -7683,6 +7714,12 @@ void CodeGenModule::EmitTopLevelDecl(Decl *D) {
case PCK_Lib:
AddDependentLib(PCD->getArg());
break;
+ case PCK_Copyright:
+ // Skip pragmas deserialized from modules/PCHs. Process the pragma comment
+ // only if it originated in this TU and the target OS is AIX.
+ if (!PCD->isFromASTFile() && getTriple().isOSAIX())
+ ProcessPragmaComment(PCD->getCommentKind(), PCD->getArg());
+ break;
case PCK_Compiler:
case PCK_ExeStr:
case PCK_User:
diff --git a/clang/lib/CodeGen/CodeGenModule.h b/clang/lib/CodeGen/CodeGenModule.h
index cc18e21d45759..21a75958dcc9d 100644
--- a/clang/lib/CodeGen/CodeGenModule.h
+++ b/clang/lib/CodeGen/CodeGenModule.h
@@ -587,6 +587,9 @@ class CodeGenModule : public CodeGenTypeCache {
/// A vector of metadata strings for dependent libraries for ELF.
SmallVector<llvm::MDNode *, 16> ELFDependentLibraries;
+ /// Metadata for copyright pragma comment (if present).
+ llvm::MDNode *LoadTimeComment = nullptr;
+
/// @name Cache for Objective-C runtime types
/// @{
@@ -1490,6 +1493,8 @@ class CodeGenModule : public CodeGenTypeCache {
/// Appends a dependent lib to the appropriate metadata value.
void AddDependentLib(StringRef Lib);
+ /// Process pragma comment
+ void ProcessPragmaComment(PragmaMSCommentKind Kind, StringRef Comment);
llvm::GlobalVariable::LinkageTypes getFunctionLinkage(GlobalDecl GD);
diff --git a/clang/lib/Parse/ParsePragma.cpp b/clang/lib/Parse/ParsePragma.cpp
index def2817c930b2..5865319470511 100644
--- a/clang/lib/Parse/ParsePragma.cpp
+++ b/clang/lib/Parse/ParsePragma.cpp
@@ -237,6 +237,7 @@ struct PragmaCommentHandler : public PragmaHandler {
private:
Sema &Actions;
+ bool SeenCopyrightInTU = false; // TU-scoped
};
struct PragmaDetectMismatchHandler : public PragmaHandler {
@@ -480,7 +481,8 @@ void Parser::initializePragmaHandlers() {
PP.AddPragmaHandler(OpenACCHandler.get());
if (getLangOpts().MicrosoftExt ||
- getTargetInfo().getTriple().isOSBinFormatELF()) {
+ getTargetInfo().getTriple().isOSBinFormatELF() ||
+ getTargetInfo().getTriple().isOSAIX()) {
MSCommentHandler = std::make_unique<PragmaCommentHandler>(Actions);
PP.AddPragmaHandler(MSCommentHandler.get());
}
@@ -607,7 +609,8 @@ void Parser::resetPragmaHandlers() {
OpenACCHandler.reset();
if (getLangOpts().MicrosoftExt ||
- getTargetInfo().getTriple().isOSBinFormatELF()) {
+ getTargetInfo().getTriple().isOSBinFormatELF() ||
+ getTargetInfo().getTriple().isOSAIX()) {
PP.RemovePragmaHandler(MSCommentHandler.get());
MSCommentHandler.reset();
}
@@ -3287,13 +3290,21 @@ void PragmaCommentHandler::HandlePragma(Preprocessor &PP,
// Verify that this is one of the 5 explicitly listed options.
IdentifierInfo *II = Tok.getIdentifierInfo();
PragmaMSCommentKind Kind =
- llvm::StringSwitch<PragmaMSCommentKind>(II->getName())
- .Case("linker", PCK_Linker)
- .Case("lib", PCK_Lib)
- .Case("compiler", PCK_Compiler)
- .Case("exestr", PCK_ExeStr)
- .Case("user", PCK_User)
- .Default(PCK_Unknown);
+ llvm::StringSwitch<PragmaMSCommentKind>(II->getName())
+ .Case("linker", PCK_Linker)
+ .Case("lib", PCK_Lib)
+ .Case("compiler", PCK_Compiler)
+ .Case("exestr", PCK_ExeStr)
+ .Case("user", PCK_User)
+ .Case("copyright", PCK_Copyright)
+ .Default(PCK_Unknown);
+
+ // Restrict copyright to AIX targets only. This could be applied for z/OS
+ // and extended with other IBM pragma comment kinds.
+ if (!PP.getTargetInfo().getTriple().isOSAIX() && Kind == PCK_Copyright) {
+ Kind = PCK_Unknown;
+ }
+
if (Kind == PCK_Unknown) {
PP.Diag(Tok.getLocation(), diag::err_pragma_comment_unknown_kind);
return;
@@ -3305,6 +3316,18 @@ void PragmaCommentHandler::HandlePragma(Preprocessor &PP,
return;
}
+ // On AIX, pragma comment copyright can each appear only once in a TU.
+ if (Kind == PCK_Copyright) {
+ assert(PP.getTargetInfo().getTriple().isOSAIX() &&
+ "Pragma Comment Copyright is supported only on AIX");
+ if (SeenCopyrightInTU) {
+ PP.Diag(Tok.getLocation(), diag::warn_pragma_comment_once)
+ << II->getName();
+ return;
+ }
+ SeenCopyrightInTU = true;
+ }
+
// Read the optional string if present.
PP.Lex(Tok);
std::string ArgumentString;
@@ -3331,6 +3354,10 @@ void PragmaCommentHandler::HandlePragma(Preprocessor &PP,
return;
}
+ // Accept and ignore well-formed copyright with empty string.
+ if (Kind == PCK_Copyright && ArgumentString.empty())
+ return;
+
// If the pragma is lexically sound, notify any interested PPCallbacks.
if (PP.getPPCallbacks())
PP.getPPCallbacks()->PragmaComment(CommentLoc, II, ArgumentString);
diff --git a/clang/test/CodeGen/PowerPC/pragma-comment-copyright-aix-modules.cpp b/clang/test/CodeGen/PowerPC/pragma-comment-copyright-aix-modules.cpp
new file mode 100644
index 0000000000000..9020df6e737d6
--- /dev/null
+++ b/clang/test/CodeGen/PowerPC/pragma-comment-copyright-aix-modules.cpp
@@ -0,0 +1,28 @@
+// RUN: split-file %s %t
+
+// Build the module interface to a PCM
+// RUN: %clang_cc1 -std=c++20 -triple powerpc-ibm-aix \
+// RUN: -emit-module-interface %t/copymod.cppm -o %t/copymod.pcm
+
+// Verify that module interface emits copyright global when compiled to IR
+// RUN: %clang_cc1 -std=c++20 -triple powerpc-ibm-aix -emit-llvm %t/copymod.cppm -o - \
+// RUN: | FileCheck %s --check-prefix=CHECK-MOD
+// CHECK-MOD: @__loadtime_comment_str = internal unnamed_addr constant [10 x i8] c"module me\00", section "__loadtime_comment"
+// CHECK-MOD: @llvm.used = appending global {{.*}} @__loadtime_comment_str
+
+// Compile an importing TU that uses the prebuilt module and verify that it
+// does NOT re-emit the module's copyright global.
+// RUN: %clang_cc1 -std=c++20 -triple powerpc-ibm-aix \
+// RUN: -fprebuilt-module-path=%t -emit-llvm %t/importmod.cc -o - \
+// RUN: | FileCheck %s --check-prefix=CHECK-IMPORT
+// CHECK-IMPORT-NOT: @__loadtime_comment_str
+// CHECK-IMPORT-NOT: c"module me\00"
+
+//--- copymod.cppm
+export module copymod;
+#pragma comment(copyright, "module me")
+export inline void f() {}
+
+//--- importmod.cc
+import copymod;
+void g() { f(); }
diff --git a/clang/test/CodeGen/PowerPC/pragma-comment-copyright-aix.c b/clang/test/CodeGen/PowerPC/pragma-comment-copyright-aix.c
new file mode 100644
index 0000000000000..98d2db416496e
--- /dev/null
+++ b/clang/test/CodeGen/PowerPC/pragma-comment-copyright-aix.c
@@ -0,0 +1,34 @@
+// RUN: %clang_cc1 %s -triple powerpc-ibm-aix -O0 -disable-llvm-passes -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 %s -triple powerpc64-ibm-aix -O0 -disable-llvm-passes -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 %s -triple powerpc-ibm-aix -verify
+// RUN: %clang_cc1 %s -triple powerpc64-ibm-aix -verify
+// RUN: %clang_cc1 %s -DTEST_EMPTY_COPYRIGHT -triple powerpc-ibm-aix -verify
+
+// RUN: %clang_cc1 %s -x c++ -triple powerpc-ibm-aix -O0 -disable-llvm-passes -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 %s -x c++ -triple powerpc64-ibm-aix -O0 -disable-llvm-passes -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 %s -x c++ -triple powerpc-ibm-aix -verify
+// RUN: %clang_cc1 %s -x c++ -triple powerpc64-ibm-aix -verify
+// RUN: %clang_cc1 %s -x c++ -DTEST_EMPTY_COPYRIGHT -triple powerpc-ibm-aix -verify
+
+#ifndef TEST_EMPTY_COPYRIGHT
+// Test basic pragma comment types
+#pragma comment(copyright, "@(#) Copyright")
+
+// Test duplicate copyright - should warn and ignore
+#pragma comment(copyright, "Duplicate Copyright") // expected-warning {{'#pragma comment copyright' can be specified only once per translation unit - ignored}}
+
+int main() { return 0; }
+
+// Check that both metadata sections are present
+// CHECK: !comment_string.loadtime = !{![[copyright:[0-9]+]]}
+
+// Check individual metadata content
+// CHECK: ![[copyright]] = !{!"@(#) Copyright"}
+
+#else
+// Test empty copyright string - valid with no warning
+#pragma comment(copyright, "") // expected-no-diagnostics
+
+int main() { return 0; }
+
+#endif
diff --git a/clang/test/CodeGen/lto-newpm-pipeline.c b/clang/test/CodeGen/lto-newpm-pipeline.c
index ea9784a76f923..b9466b9558b06 100644
--- a/clang/test/CodeGen/lto-newpm-pipeline.c
+++ b/clang/test/CodeGen/lto-newpm-pipeline.c
@@ -27,6 +27,7 @@
// CHECK-FULL-O0: Running pass: VerifierPass
// CHECK-FULL-O0-NEXT: Running analysis: VerifierAnalysis
+// CHECK-FULL-O0-NEXT: Running pass: LowerCommentStringPass
// CHECK-FULL-O0-NEXT: Running analysis: InnerAnalysisManagerProxy
// CHECK-FULL-O0-NEXT: Running pass: EntryExitInstrumenterPass
// CHECK-FULL-O0-NEXT: Running pass: AlwaysInlinerPass
@@ -41,6 +42,7 @@
// CHECK-THIN-O0: Running pass: VerifierPass
// CHECK-THIN-O0-NEXT: Running analysis: VerifierAnalysis
+// CHECK-THIN-O0-NEXT: Running pass: LowerCommentStringPass
// CHECK-THIN-O0-NEXT: Running analysis: InnerAnalysisManagerProxy
// CHECK-THIN-O0-NEXT: Running pass: EntryExitInstrumenterPass
// CHECK-THIN-O0-NEXT: Running pass: AlwaysInlinerPass
diff --git a/llvm/include/llvm/Transforms/Utils/LowerCommentStringPass.h b/llvm/include/llvm/Transforms/Utils/LowerCommentStringPass.h
new file mode 100644
index 0000000000000..4c6109e1176d3
--- /dev/null
+++ b/llvm/include/llvm/Transforms/Utils/LowerCommentStringPass.h
@@ -0,0 +1,24 @@
+//===-- LowerCommentStringPass.h - Lower Comment string metadata --===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===---------------------------------------------------------------------===//
+
+#ifndef LLVM_TRANSFORMS_UTILS_LOWERCOMMENTSTRINGPASS_H
+#define LLVM_TRANSFORMS_UTILS_LOWERCOMMENTSTRINGPASS_H
+
+#include "llvm/IR/PassManager.h"
+
+namespace llvm {
+class LowerCommentStringPass : public PassInfoMixin<LowerCommentStringPass> {
+public:
+ PreservedAnalyses run(Module &M, ModuleAnalysisManager &AM);
+
+ static bool isRequired() { return true; }
+};
+
+} // namespace llvm
+
+#endif // LLVM_TRANSFORMS_UTILS_LOWERCOMMENTSTRINGPASS_H
diff --git a/llvm/lib/Passes/PassBuilder.cpp b/llvm/lib/Passes/PassBuilder.cpp
index 45955426d66a0..5c014644a6150 100644
--- a/llvm/lib/Passes/PassBuilder.cpp
+++ b/llvm/lib/Passes/PassBuilder.cpp
@@ -361,6 +361,7 @@
#include "llvm/Transforms/Utils/LibCallsShrinkWrap.h"
#include "llvm/Transforms/Utils/LoopSimplify.h"
#include "llvm/Transforms/Utils/LoopVersioning.h"
+#include "llvm/Transforms/Utils/LowerCommentStringPass.h"
#include "llvm/Transforms/Utils/LowerGlobalDtors.h"
#include "llvm/Transforms/Utils/LowerIFunc.h"
#include "llvm/Transforms/Utils/LowerInvoke.h"
diff --git a/llvm/lib/Passes/PassBuilderPipelines.cpp b/llvm/lib/Passes/PassBuilderPipelines.cpp
index 1584d30875570..e48833528c85a 100644
--- a/llvm/lib/Passes/PassBuilderPipelines.cpp
+++ b/llvm/lib/Passes/PassBuilderPipelines.cpp
@@ -142,6 +142,7 @@
#include "llvm/Transforms/Utils/ExtraPassManager.h"
#include "llvm/Transforms/Utils/InjectTLIMappings.h"
#include "llvm/Transforms/Utils/LibCallsShrinkWrap.h"
+#include "llvm/Transforms/Utils/LowerCommentStringPass.h"
#include "llvm/Transforms/Utils/Mem2Reg.h"
#include "llvm/Transforms/Utils/MoveAutoInit.h"
#include "llvm/Transforms/Utils/NameAnonGlobals.h"
@@ -1470,6 +1471,9 @@ PassBuilder::buildModuleOptimizationPipeline(OptimizationLevel Level,
const bool LTOPreLink = isLTOPreLink(LTOPhase);
ModulePassManager MPM;
+ // Process copyright metadata early, before any optimizations
+ MPM.addPass(LowerCommentStringPass());
+
// Run partial inlining pass to partially inline functions that have
// large bodies.
if (RunPartialInlining)
@@ -2318,6 +2322,9 @@ PassBuilder::buildO0DefaultPipeline(OptimizationLevel Level,
ModulePassManager MPM;
+ // Process copyright metadata at O0 before any other transformations
+ MPM.addPass(LowerCommentStringPass());
+
// Perform pseudo probe instrumentation in O0 mode. This is for the
// consistency between different build modes. For example, a LTO build can be
// mixed with an O0 prelink and an O2 postlink. Loading a sample profile in
diff --git a/llvm/lib/Passes/PassRegistry.def b/llvm/lib/Passes/PassRegistry.def
index 2cfb5b2592601..05cc8775b8d71 100644
--- a/llvm/lib/Passes/PassRegistry.def
+++ b/llvm/lib/Passes/PassRegistry.def
@@ -62,6 +62,7 @@ MODULE_PASS("check-debugify", NewPMCheckDebugifyPass())
MODULE_PASS("constmerge", ConstantMergePass())
MODULE_PASS("coro-cleanup", CoroCleanupPass())
MODULE_PASS("coro-early", CoroEarlyPass())
+MODULE_PASS("lower-comment-string", LowerCommentStringPass())
MODULE_PASS("cross-dso-cfi", CrossDSOCFIPass())
MODULE_PASS("ctx-instr-gen",
PGOInstrumentationGen(PGOInstrumentationType::CTXPROF))
diff --git a/llvm/lib/Transforms/Utils/CMakeLists.txt b/llvm/lib/Transforms/Utils/CMakeLists.txt
index 2b5f5cf344e60..28f86c5e96010 100644
--- a/llvm/lib/Transforms/Utils/CMakeLists.txt
+++ b/llvm/lib/Transforms/Utils/CMakeLists.txt
@@ -53,6 +53,7 @@ add_llvm_component_library(LLVMTransformUtils
LoopUtils.cpp
LoopVersioning.cpp
LowerAtomic.cpp
+ LowerCommentStringPass.cpp
LowerGlobalDtors.cpp
LowerIFunc.cpp
LowerInvoke.cpp
diff --git a/llvm/lib/Transforms/Utils/LowerCommentStringPass.cpp b/llvm/lib/Transforms/Utils/LowerCommentStringPass.cpp
new file mode 100644
index 0000000000000..6deef2f75e0a3
--- /dev/null
+++ b/llvm/lib/Transforms/Utils/LowerCommentStringPass.cpp
@@ -0,0 +1,148 @@
+//===-- LowerCommentStringPass.cpp - Lower Comment string metadata -------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===---------------------------------------------------------------------===//
+//
+// LowerCommentStringPass pass lowers module-level comment string metadata
+// emitted by Clang:
+//
+// !comment_string.loadtime = !{!"Copyright ..."}
+//
+// into concrete, translation-unit–local globals.
+// This Pass is enabled only for AIX.
+// For each module (translation unit), the pass performs the following:
+//
+// 1. Creates a null-terminated, internal constant string global
+// (`__loadtime_comment_str`) containing the copyright text in
+// `__loadtime_comment` section.
+//
+// 2. Marks the string in `llvm.used` so it cannot be dropped by
+// optimization or LTO.
+//
+// 3. Attaches `!implicit.ref` metadata referencing the string to every
+// defined function in the module. The PowerPC AIX backend recognizes
+// this metadata and emits a `.ref` directive from the function to the
+// string, creating a concrete relocation that prevents the linker from
+// discarding it (as long as the referencing symbol is kept).
+//
+// Input IR:
+// !comment_string.loadtime = !{!"Copyright"}
+// Output IR:
+// @__loadtime_comment_str = internal constant [N x i8] c"Copyright\00",
+// section "__loadtime_comment"
+// @llvm.used = appending global [1 x ptr] [ptr @__loadtime_comment_str]
+//
+// define i32 @func() !implicit.ref !5 { ... }
+// !5 = !{ptr @__loadtime_comment_str}
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/Transforms/Utils/LowerCommentStringPass.h"
+
+#include "llvm/ADT/SmallVector.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/IR/Attributes.h"
+#include "llvm/IR/Constants.h"
+#include "llvm/IR/DerivedTypes.h"
+#include "llvm/IR/Function.h"
+#include "llvm/IR/GlobalValue.h"
+#include "llvm/IR/GlobalVariable.h"
+#include "llvm/IR/MDBuilder.h"
+#include "llvm/IR/Metadata.h"
+#include "llvm/IR/Module.h"
+#include "llvm/IR/Type.h"
+#include "llvm/IR/Value.h"
+#include "llvm/Passes/PassBuilder.h"
+#include "llvm/Support/CommandLine.h"
+#include "llvm/Support/Debug.h"
+#include "llvm/TargetParser/Triple.h"
+#include "llvm/Transforms/Utils/ModuleUtils.h"
+
+#define DEBUG_TYPE "lower-comment-string"
+
+using namespace llvm;
+
+static cl::opt<bool>
+ DisableCopyrightMetadata("disable-lower-comment-string", cl::ReallyHidden,
+ cl::desc("Disable LowerCommentString pass."),
+ cl::init(false));
+
+static bool isAIXTriple(const Module &M) {
+ return Triple(M.getTargetTriple()).isOSAIX();
+}
+
+P...
[truncated]
|
Member
|
@llvm/pr-subscribers-backend-aarch64 Author: Tony Varghese (tonykuttai) Changes
This change builds on top of XCOFF associated metadata and was reviewed at [PowerPC][AIX] Support #pragma comment copyright for AIX Patch is 29.81 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/178184.diff 22 Files Affected:
diff --git a/clang/include/clang/Basic/DiagnosticParseKinds.td b/clang/include/clang/Basic/DiagnosticParseKinds.td
index 457d3644de35a..ede6a90a31cc6 100644
--- a/clang/include/clang/Basic/DiagnosticParseKinds.td
+++ b/clang/include/clang/Basic/DiagnosticParseKinds.td
@@ -1358,6 +1358,10 @@ def err_pragma_comment_unknown_kind : Error<"unknown kind of pragma comment">;
// PS4 recognizes only #pragma comment(lib)
def warn_pragma_comment_ignored : Warning<"'#pragma comment %0' ignored">,
InGroup<IgnoredPragmas>;
+def warn_pragma_comment_once
+ : Warning<"'#pragma comment %0' "
+ "can be specified only once per translation unit - ignored">,
+ InGroup<IgnoredPragmas>;
// - #pragma detect_mismatch
def err_pragma_detect_mismatch_malformed : Error<
"pragma detect_mismatch is malformed; it requires two comma-separated "
diff --git a/clang/include/clang/Basic/PragmaKinds.h b/clang/include/clang/Basic/PragmaKinds.h
index 42f049f7323d2..52ca58855d460 100644
--- a/clang/include/clang/Basic/PragmaKinds.h
+++ b/clang/include/clang/Basic/PragmaKinds.h
@@ -17,7 +17,8 @@ enum PragmaMSCommentKind {
PCK_Lib, // #pragma comment(lib, ...)
PCK_Compiler, // #pragma comment(compiler, ...)
PCK_ExeStr, // #pragma comment(exestr, ...)
- PCK_User // #pragma comment(user, ...)
+ PCK_User, // #pragma comment(user, ...)
+ PCK_Copyright // #pragma comment(copyright, ...)
};
enum PragmaMSStructKind {
diff --git a/clang/lib/AST/TextNodeDumper.cpp b/clang/lib/AST/TextNodeDumper.cpp
index 7bc0404db1bee..c58c33f95c193 100644
--- a/clang/lib/AST/TextNodeDumper.cpp
+++ b/clang/lib/AST/TextNodeDumper.cpp
@@ -2526,6 +2526,9 @@ void TextNodeDumper::VisitPragmaCommentDecl(const PragmaCommentDecl *D) {
case PCK_User:
OS << "user";
break;
+ case PCK_Copyright:
+ OS << "copyright";
+ break;
}
StringRef Arg = D->getArg();
if (!Arg.empty())
diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index d50c9605a30b3..e63adc98a95aa 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -1588,6 +1588,12 @@ void CodeGenModule::Release() {
EmitBackendOptionsMetadata(getCodeGenOpts());
+ // Emit copyright metadata
+ if (LoadTimeComment) {
+ auto *NMD = getModule().getOrInsertNamedMetadata("comment_string.loadtime");
+ NMD->addOperand(LoadTimeComment);
+ }
+
// If there is device offloading code embed it in the host now.
EmbedObject(&getModule(), CodeGenOpts, *getFileSystem(), getDiags());
@@ -3481,6 +3487,31 @@ void CodeGenModule::AddDependentLib(StringRef Lib) {
LinkerOptionsMetadata.push_back(llvm::MDNode::get(C, MDOpts));
}
+/// Process AIX copyright pragma and create LLVM metadata.
+/// #pragma comment(copyright, "string") embed copyright
+/// information into the object file's loader section.
+///
+/// Example: #pragma comment(copyright, "Copyright IBM Corp. 2024")
+///
+/// This should only be called once per translation unit.
+void CodeGenModule::ProcessPragmaComment(PragmaMSCommentKind Kind,
+ StringRef Comment) {
+ // Ensure we are only processing Copyright Pragmas
+ assert(Kind == PCK_Copyright &&
+ "Unexpected pragma comment kind, ProcessPragmaComment should only be "
+ "called for PCK_Copyright");
+
+ // Only one copyright pragma allowed per translation unit
+ if (LoadTimeComment) {
+ return;
+ }
+
+ // Create llvm metadata with the comment string
+ auto &C = getLLVMContext();
+ llvm::Metadata *Ops[] = {llvm::MDString::get(C, Comment)};
+ LoadTimeComment = llvm::MDNode::get(C, Ops);
+}
+
/// Add link options implied by the given module, including modules
/// it depends on, using a postorder walk.
static void addLinkOptionsPostorder(CodeGenModule &CGM, Module *Mod,
@@ -7683,6 +7714,12 @@ void CodeGenModule::EmitTopLevelDecl(Decl *D) {
case PCK_Lib:
AddDependentLib(PCD->getArg());
break;
+ case PCK_Copyright:
+ // Skip pragmas deserialized from modules/PCHs. Process the pragma comment
+ // only if it originated in this TU and the target OS is AIX.
+ if (!PCD->isFromASTFile() && getTriple().isOSAIX())
+ ProcessPragmaComment(PCD->getCommentKind(), PCD->getArg());
+ break;
case PCK_Compiler:
case PCK_ExeStr:
case PCK_User:
diff --git a/clang/lib/CodeGen/CodeGenModule.h b/clang/lib/CodeGen/CodeGenModule.h
index cc18e21d45759..21a75958dcc9d 100644
--- a/clang/lib/CodeGen/CodeGenModule.h
+++ b/clang/lib/CodeGen/CodeGenModule.h
@@ -587,6 +587,9 @@ class CodeGenModule : public CodeGenTypeCache {
/// A vector of metadata strings for dependent libraries for ELF.
SmallVector<llvm::MDNode *, 16> ELFDependentLibraries;
+ /// Metadata for copyright pragma comment (if present).
+ llvm::MDNode *LoadTimeComment = nullptr;
+
/// @name Cache for Objective-C runtime types
/// @{
@@ -1490,6 +1493,8 @@ class CodeGenModule : public CodeGenTypeCache {
/// Appends a dependent lib to the appropriate metadata value.
void AddDependentLib(StringRef Lib);
+ /// Process pragma comment
+ void ProcessPragmaComment(PragmaMSCommentKind Kind, StringRef Comment);
llvm::GlobalVariable::LinkageTypes getFunctionLinkage(GlobalDecl GD);
diff --git a/clang/lib/Parse/ParsePragma.cpp b/clang/lib/Parse/ParsePragma.cpp
index def2817c930b2..5865319470511 100644
--- a/clang/lib/Parse/ParsePragma.cpp
+++ b/clang/lib/Parse/ParsePragma.cpp
@@ -237,6 +237,7 @@ struct PragmaCommentHandler : public PragmaHandler {
private:
Sema &Actions;
+ bool SeenCopyrightInTU = false; // TU-scoped
};
struct PragmaDetectMismatchHandler : public PragmaHandler {
@@ -480,7 +481,8 @@ void Parser::initializePragmaHandlers() {
PP.AddPragmaHandler(OpenACCHandler.get());
if (getLangOpts().MicrosoftExt ||
- getTargetInfo().getTriple().isOSBinFormatELF()) {
+ getTargetInfo().getTriple().isOSBinFormatELF() ||
+ getTargetInfo().getTriple().isOSAIX()) {
MSCommentHandler = std::make_unique<PragmaCommentHandler>(Actions);
PP.AddPragmaHandler(MSCommentHandler.get());
}
@@ -607,7 +609,8 @@ void Parser::resetPragmaHandlers() {
OpenACCHandler.reset();
if (getLangOpts().MicrosoftExt ||
- getTargetInfo().getTriple().isOSBinFormatELF()) {
+ getTargetInfo().getTriple().isOSBinFormatELF() ||
+ getTargetInfo().getTriple().isOSAIX()) {
PP.RemovePragmaHandler(MSCommentHandler.get());
MSCommentHandler.reset();
}
@@ -3287,13 +3290,21 @@ void PragmaCommentHandler::HandlePragma(Preprocessor &PP,
// Verify that this is one of the 5 explicitly listed options.
IdentifierInfo *II = Tok.getIdentifierInfo();
PragmaMSCommentKind Kind =
- llvm::StringSwitch<PragmaMSCommentKind>(II->getName())
- .Case("linker", PCK_Linker)
- .Case("lib", PCK_Lib)
- .Case("compiler", PCK_Compiler)
- .Case("exestr", PCK_ExeStr)
- .Case("user", PCK_User)
- .Default(PCK_Unknown);
+ llvm::StringSwitch<PragmaMSCommentKind>(II->getName())
+ .Case("linker", PCK_Linker)
+ .Case("lib", PCK_Lib)
+ .Case("compiler", PCK_Compiler)
+ .Case("exestr", PCK_ExeStr)
+ .Case("user", PCK_User)
+ .Case("copyright", PCK_Copyright)
+ .Default(PCK_Unknown);
+
+ // Restrict copyright to AIX targets only. This could be applied for z/OS
+ // and extended with other IBM pragma comment kinds.
+ if (!PP.getTargetInfo().getTriple().isOSAIX() && Kind == PCK_Copyright) {
+ Kind = PCK_Unknown;
+ }
+
if (Kind == PCK_Unknown) {
PP.Diag(Tok.getLocation(), diag::err_pragma_comment_unknown_kind);
return;
@@ -3305,6 +3316,18 @@ void PragmaCommentHandler::HandlePragma(Preprocessor &PP,
return;
}
+ // On AIX, pragma comment copyright can each appear only once in a TU.
+ if (Kind == PCK_Copyright) {
+ assert(PP.getTargetInfo().getTriple().isOSAIX() &&
+ "Pragma Comment Copyright is supported only on AIX");
+ if (SeenCopyrightInTU) {
+ PP.Diag(Tok.getLocation(), diag::warn_pragma_comment_once)
+ << II->getName();
+ return;
+ }
+ SeenCopyrightInTU = true;
+ }
+
// Read the optional string if present.
PP.Lex(Tok);
std::string ArgumentString;
@@ -3331,6 +3354,10 @@ void PragmaCommentHandler::HandlePragma(Preprocessor &PP,
return;
}
+ // Accept and ignore well-formed copyright with empty string.
+ if (Kind == PCK_Copyright && ArgumentString.empty())
+ return;
+
// If the pragma is lexically sound, notify any interested PPCallbacks.
if (PP.getPPCallbacks())
PP.getPPCallbacks()->PragmaComment(CommentLoc, II, ArgumentString);
diff --git a/clang/test/CodeGen/PowerPC/pragma-comment-copyright-aix-modules.cpp b/clang/test/CodeGen/PowerPC/pragma-comment-copyright-aix-modules.cpp
new file mode 100644
index 0000000000000..9020df6e737d6
--- /dev/null
+++ b/clang/test/CodeGen/PowerPC/pragma-comment-copyright-aix-modules.cpp
@@ -0,0 +1,28 @@
+// RUN: split-file %s %t
+
+// Build the module interface to a PCM
+// RUN: %clang_cc1 -std=c++20 -triple powerpc-ibm-aix \
+// RUN: -emit-module-interface %t/copymod.cppm -o %t/copymod.pcm
+
+// Verify that module interface emits copyright global when compiled to IR
+// RUN: %clang_cc1 -std=c++20 -triple powerpc-ibm-aix -emit-llvm %t/copymod.cppm -o - \
+// RUN: | FileCheck %s --check-prefix=CHECK-MOD
+// CHECK-MOD: @__loadtime_comment_str = internal unnamed_addr constant [10 x i8] c"module me\00", section "__loadtime_comment"
+// CHECK-MOD: @llvm.used = appending global {{.*}} @__loadtime_comment_str
+
+// Compile an importing TU that uses the prebuilt module and verify that it
+// does NOT re-emit the module's copyright global.
+// RUN: %clang_cc1 -std=c++20 -triple powerpc-ibm-aix \
+// RUN: -fprebuilt-module-path=%t -emit-llvm %t/importmod.cc -o - \
+// RUN: | FileCheck %s --check-prefix=CHECK-IMPORT
+// CHECK-IMPORT-NOT: @__loadtime_comment_str
+// CHECK-IMPORT-NOT: c"module me\00"
+
+//--- copymod.cppm
+export module copymod;
+#pragma comment(copyright, "module me")
+export inline void f() {}
+
+//--- importmod.cc
+import copymod;
+void g() { f(); }
diff --git a/clang/test/CodeGen/PowerPC/pragma-comment-copyright-aix.c b/clang/test/CodeGen/PowerPC/pragma-comment-copyright-aix.c
new file mode 100644
index 0000000000000..98d2db416496e
--- /dev/null
+++ b/clang/test/CodeGen/PowerPC/pragma-comment-copyright-aix.c
@@ -0,0 +1,34 @@
+// RUN: %clang_cc1 %s -triple powerpc-ibm-aix -O0 -disable-llvm-passes -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 %s -triple powerpc64-ibm-aix -O0 -disable-llvm-passes -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 %s -triple powerpc-ibm-aix -verify
+// RUN: %clang_cc1 %s -triple powerpc64-ibm-aix -verify
+// RUN: %clang_cc1 %s -DTEST_EMPTY_COPYRIGHT -triple powerpc-ibm-aix -verify
+
+// RUN: %clang_cc1 %s -x c++ -triple powerpc-ibm-aix -O0 -disable-llvm-passes -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 %s -x c++ -triple powerpc64-ibm-aix -O0 -disable-llvm-passes -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 %s -x c++ -triple powerpc-ibm-aix -verify
+// RUN: %clang_cc1 %s -x c++ -triple powerpc64-ibm-aix -verify
+// RUN: %clang_cc1 %s -x c++ -DTEST_EMPTY_COPYRIGHT -triple powerpc-ibm-aix -verify
+
+#ifndef TEST_EMPTY_COPYRIGHT
+// Test basic pragma comment types
+#pragma comment(copyright, "@(#) Copyright")
+
+// Test duplicate copyright - should warn and ignore
+#pragma comment(copyright, "Duplicate Copyright") // expected-warning {{'#pragma comment copyright' can be specified only once per translation unit - ignored}}
+
+int main() { return 0; }
+
+// Check that both metadata sections are present
+// CHECK: !comment_string.loadtime = !{![[copyright:[0-9]+]]}
+
+// Check individual metadata content
+// CHECK: ![[copyright]] = !{!"@(#) Copyright"}
+
+#else
+// Test empty copyright string - valid with no warning
+#pragma comment(copyright, "") // expected-no-diagnostics
+
+int main() { return 0; }
+
+#endif
diff --git a/clang/test/CodeGen/lto-newpm-pipeline.c b/clang/test/CodeGen/lto-newpm-pipeline.c
index ea9784a76f923..b9466b9558b06 100644
--- a/clang/test/CodeGen/lto-newpm-pipeline.c
+++ b/clang/test/CodeGen/lto-newpm-pipeline.c
@@ -27,6 +27,7 @@
// CHECK-FULL-O0: Running pass: VerifierPass
// CHECK-FULL-O0-NEXT: Running analysis: VerifierAnalysis
+// CHECK-FULL-O0-NEXT: Running pass: LowerCommentStringPass
// CHECK-FULL-O0-NEXT: Running analysis: InnerAnalysisManagerProxy
// CHECK-FULL-O0-NEXT: Running pass: EntryExitInstrumenterPass
// CHECK-FULL-O0-NEXT: Running pass: AlwaysInlinerPass
@@ -41,6 +42,7 @@
// CHECK-THIN-O0: Running pass: VerifierPass
// CHECK-THIN-O0-NEXT: Running analysis: VerifierAnalysis
+// CHECK-THIN-O0-NEXT: Running pass: LowerCommentStringPass
// CHECK-THIN-O0-NEXT: Running analysis: InnerAnalysisManagerProxy
// CHECK-THIN-O0-NEXT: Running pass: EntryExitInstrumenterPass
// CHECK-THIN-O0-NEXT: Running pass: AlwaysInlinerPass
diff --git a/llvm/include/llvm/Transforms/Utils/LowerCommentStringPass.h b/llvm/include/llvm/Transforms/Utils/LowerCommentStringPass.h
new file mode 100644
index 0000000000000..4c6109e1176d3
--- /dev/null
+++ b/llvm/include/llvm/Transforms/Utils/LowerCommentStringPass.h
@@ -0,0 +1,24 @@
+//===-- LowerCommentStringPass.h - Lower Comment string metadata --===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===---------------------------------------------------------------------===//
+
+#ifndef LLVM_TRANSFORMS_UTILS_LOWERCOMMENTSTRINGPASS_H
+#define LLVM_TRANSFORMS_UTILS_LOWERCOMMENTSTRINGPASS_H
+
+#include "llvm/IR/PassManager.h"
+
+namespace llvm {
+class LowerCommentStringPass : public PassInfoMixin<LowerCommentStringPass> {
+public:
+ PreservedAnalyses run(Module &M, ModuleAnalysisManager &AM);
+
+ static bool isRequired() { return true; }
+};
+
+} // namespace llvm
+
+#endif // LLVM_TRANSFORMS_UTILS_LOWERCOMMENTSTRINGPASS_H
diff --git a/llvm/lib/Passes/PassBuilder.cpp b/llvm/lib/Passes/PassBuilder.cpp
index 45955426d66a0..5c014644a6150 100644
--- a/llvm/lib/Passes/PassBuilder.cpp
+++ b/llvm/lib/Passes/PassBuilder.cpp
@@ -361,6 +361,7 @@
#include "llvm/Transforms/Utils/LibCallsShrinkWrap.h"
#include "llvm/Transforms/Utils/LoopSimplify.h"
#include "llvm/Transforms/Utils/LoopVersioning.h"
+#include "llvm/Transforms/Utils/LowerCommentStringPass.h"
#include "llvm/Transforms/Utils/LowerGlobalDtors.h"
#include "llvm/Transforms/Utils/LowerIFunc.h"
#include "llvm/Transforms/Utils/LowerInvoke.h"
diff --git a/llvm/lib/Passes/PassBuilderPipelines.cpp b/llvm/lib/Passes/PassBuilderPipelines.cpp
index 1584d30875570..e48833528c85a 100644
--- a/llvm/lib/Passes/PassBuilderPipelines.cpp
+++ b/llvm/lib/Passes/PassBuilderPipelines.cpp
@@ -142,6 +142,7 @@
#include "llvm/Transforms/Utils/ExtraPassManager.h"
#include "llvm/Transforms/Utils/InjectTLIMappings.h"
#include "llvm/Transforms/Utils/LibCallsShrinkWrap.h"
+#include "llvm/Transforms/Utils/LowerCommentStringPass.h"
#include "llvm/Transforms/Utils/Mem2Reg.h"
#include "llvm/Transforms/Utils/MoveAutoInit.h"
#include "llvm/Transforms/Utils/NameAnonGlobals.h"
@@ -1470,6 +1471,9 @@ PassBuilder::buildModuleOptimizationPipeline(OptimizationLevel Level,
const bool LTOPreLink = isLTOPreLink(LTOPhase);
ModulePassManager MPM;
+ // Process copyright metadata early, before any optimizations
+ MPM.addPass(LowerCommentStringPass());
+
// Run partial inlining pass to partially inline functions that have
// large bodies.
if (RunPartialInlining)
@@ -2318,6 +2322,9 @@ PassBuilder::buildO0DefaultPipeline(OptimizationLevel Level,
ModulePassManager MPM;
+ // Process copyright metadata at O0 before any other transformations
+ MPM.addPass(LowerCommentStringPass());
+
// Perform pseudo probe instrumentation in O0 mode. This is for the
// consistency between different build modes. For example, a LTO build can be
// mixed with an O0 prelink and an O2 postlink. Loading a sample profile in
diff --git a/llvm/lib/Passes/PassRegistry.def b/llvm/lib/Passes/PassRegistry.def
index 2cfb5b2592601..05cc8775b8d71 100644
--- a/llvm/lib/Passes/PassRegistry.def
+++ b/llvm/lib/Passes/PassRegistry.def
@@ -62,6 +62,7 @@ MODULE_PASS("check-debugify", NewPMCheckDebugifyPass())
MODULE_PASS("constmerge", ConstantMergePass())
MODULE_PASS("coro-cleanup", CoroCleanupPass())
MODULE_PASS("coro-early", CoroEarlyPass())
+MODULE_PASS("lower-comment-string", LowerCommentStringPass())
MODULE_PASS("cross-dso-cfi", CrossDSOCFIPass())
MODULE_PASS("ctx-instr-gen",
PGOInstrumentationGen(PGOInstrumentationType::CTXPROF))
diff --git a/llvm/lib/Transforms/Utils/CMakeLists.txt b/llvm/lib/Transforms/Utils/CMakeLists.txt
index 2b5f5cf344e60..28f86c5e96010 100644
--- a/llvm/lib/Transforms/Utils/CMakeLists.txt
+++ b/llvm/lib/Transforms/Utils/CMakeLists.txt
@@ -53,6 +53,7 @@ add_llvm_component_library(LLVMTransformUtils
LoopUtils.cpp
LoopVersioning.cpp
LowerAtomic.cpp
+ LowerCommentStringPass.cpp
LowerGlobalDtors.cpp
LowerIFunc.cpp
LowerInvoke.cpp
diff --git a/llvm/lib/Transforms/Utils/LowerCommentStringPass.cpp b/llvm/lib/Transforms/Utils/LowerCommentStringPass.cpp
new file mode 100644
index 0000000000000..6deef2f75e0a3
--- /dev/null
+++ b/llvm/lib/Transforms/Utils/LowerCommentStringPass.cpp
@@ -0,0 +1,148 @@
+//===-- LowerCommentStringPass.cpp - Lower Comment string metadata -------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===---------------------------------------------------------------------===//
+//
+// LowerCommentStringPass pass lowers module-level comment string metadata
+// emitted by Clang:
+//
+// !comment_string.loadtime = !{!"Copyright ..."}
+//
+// into concrete, translation-unit–local globals.
+// This Pass is enabled only for AIX.
+// For each module (translation unit), the pass performs the following:
+//
+// 1. Creates a null-terminated, internal constant string global
+// (`__loadtime_comment_str`) containing the copyright text in
+// `__loadtime_comment` section.
+//
+// 2. Marks the string in `llvm.used` so it cannot be dropped by
+// optimization or LTO.
+//
+// 3. Attaches `!implicit.ref` metadata referencing the string to every
+// defined function in the module. The PowerPC AIX backend recognizes
+// this metadata and emits a `.ref` directive from the function to the
+// string, creating a concrete relocation that prevents the linker from
+// discarding it (as long as the referencing symbol is kept).
+//
+// Input IR:
+// !comment_string.loadtime = !{!"Copyright"}
+// Output IR:
+// @__loadtime_comment_str = internal constant [N x i8] c"Copyright\00",
+// section "__loadtime_comment"
+// @llvm.used = appending global [1 x ptr] [ptr @__loadtime_comment_str]
+//
+// define i32 @func() !implicit.ref !5 { ... }
+// !5 = !{ptr @__loadtime_comment_str}
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/Transforms/Utils/LowerCommentStringPass.h"
+
+#include "llvm/ADT/SmallVector.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/IR/Attributes.h"
+#include "llvm/IR/Constants.h"
+#include "llvm/IR/DerivedTypes.h"
+#include "llvm/IR/Function.h"
+#include "llvm/IR/GlobalValue.h"
+#include "llvm/IR/GlobalVariable.h"
+#include "llvm/IR/MDBuilder.h"
+#include "llvm/IR/Metadata.h"
+#include "llvm/IR/Module.h"
+#include "llvm/IR/Type.h"
+#include "llvm/IR/Value.h"
+#include "llvm/Passes/PassBuilder.h"
+#include "llvm/Support/CommandLine.h"
+#include "llvm/Support/Debug.h"
+#include "llvm/TargetParser/Triple.h"
+#include "llvm/Transforms/Utils/ModuleUtils.h"
+
+#define DEBUG_TYPE "lower-comment-string"
+
+using namespace llvm;
+
+static cl::opt<bool>
+ DisableCopyrightMetadata("disable-lower-comment-string", cl::ReallyHidden,
+ cl::desc("Disable LowerCommentString pass."),
+ cl::init(false));
+
+static bool isAIXTriple(const Module &M) {
+ return Triple(M.getTargetTriple()).isOSAIX();
+}
+
+P...
[truncated]
|
Contributor
Author
|
The following tests had to be modified as we are keeping the |
perry-ca
reviewed
Feb 5, 2026
tonykuttai
force-pushed
the
tvarghese/pragma-copyright-comment-update
branch
from
2b44311 to
2abc275
Compare
perry-ca
reviewed
Feb 6, 2026
Member
hubert-reinterpretcast
left a comment
hubert-reinterpretcast
left a comment
There was a problem hiding this comment.
Reviewed Clang changes (excluding tests).
tonykuttai
force-pushed
the
tvarghese/pragma-copyright-comment-update
branch
from
2c769c5 to
06546f7
Compare
Member
hubert-reinterpretcast
left a comment
hubert-reinterpretcast
left a comment
There was a problem hiding this comment.
Next round of review of Clang changes (excluding tests).
tonykuttai
force-pushed
the
tvarghese/pragma-copyright-comment-update
branch
2 times, most recently
from
08540ef to
0c36839
Compare
Contributor
Author
|
Addressed the review comments. |
w2yehia
reviewed
Apr 15, 2026
| @@ -1472,6 +1473,9 @@ PassBuilder::buildModuleOptimizationPipeline(OptimizationLevel Level, | |||
| const bool LTOPreLink = isLTOPreLink(LTOPhase); | |||
| ModulePassManager MPM; | |||
|
|
|||
| // Process copyright metadata early, before any optimizations | |||
Contributor
There was a problem hiding this comment.
why are we processing the metadata early rather than late (but not as late as link-time LTO)?
Contributor
Author
There was a problem hiding this comment.
The design rationale was to get an IR pass to replace the metadata with an IR definition and create the necessary .refs in the IR from the definitions in the IR. Thus by keeping the pass early per TU, the metadata only survives from the point of the front-end codegen until that IR pass. By the time LTO runs, the metadata is already consumed and concrete globals + .ref directives are already in the object file.
Contributor
There was a problem hiding this comment.
I did say "not as late as link-time LTO".
I meant late in the IR-IR pipeline during prelink (-flto -c) LTO or regular nonLTO (-c) compilation.
Looking at the source, at the end of buildModuleOptimizationPipeline might be what I meant.
The advantage of doing it late is that any compiler generated functions will also have a .ref to the copyright variables.
Contributor
Author
There was a problem hiding this comment.
You are right. I have moved the pass to the end of the pipelines in buildO0DefaultPipeline, buildModuleOptimizationPipeline and buildThinLTOPreLinkDefaultPipeline.
w2yehia
approved these changes
Apr 15, 2026
tonykuttai
force-pushed
the
tvarghese/pragma-copyright-comment-update
branch
2 times, most recently
from
29019ca to
865add9
Compare
Member
hubert-reinterpretcast
left a comment
hubert-reinterpretcast
left a comment
There was a problem hiding this comment.
I have limited my review to the front end (and front-end tests) only.
tonykuttai
force-pushed
the
tvarghese/pragma-copyright-comment-update
branch
from
83acf57 to
7d468a5
Compare
tonykuttai
force-pushed
the
tvarghese/pragma-copyright-comment-update
branch
from
f010482 to
9c2b2fe
Compare
Member
|
@tonykuttai, for future reference, it is easier for reviewers to do incremental reviews if force pushes are avoided on the PR branch. Note also that for 9c2b2fe, the commit message for the squashed commit is out-of-date due to changes made during the PR review process. |
mandlebug
reviewed
May 8, 2026
| // global and attach !implicit.ref to all defined functions. Running late | ||
| // ensures compiler-generated functions (e.g. from inlining, coroutines) | ||
| // are also anchored to the copyright string via .ref directives. | ||
| MPM.addPass(LowerCommentStringPass()); |
Member
There was a problem hiding this comment.
Is this pass being added to the opt pipeline so that we can get early mapping from functions to the string with the implicit.ref? Since its AIX only I would have expected it to get put into the target specific pipeline we build for llc instead of getting run for all targets.
Contributor
Author
There was a problem hiding this comment.
Primary motivation was to convert !comment_string.loadtime metadata to LLVM IR earlier in the pipeline. Also it needs to run in both pre-link and post-link LTO phases, which is simpler in opt. The pass itself is bailed out early for non-aix targets. Having it in the general pipeline makes future extension much easier. update isSupportedTarget() when adding z/OS or other targets that need similar load-time comment functionality.
hubert-reinterpretcast
approved these changes
May 9, 2026
Member
There was a problem hiding this comment.
LGTM (for the front-end) with specific suggestions to address code-style issues.
@tonykuttai, the general take-away from this round of review is that code comments should provide value to the understanding of the code where it appears and that said value needs to be balanced against the drafting, review, and maintenance burden of having the comment.
Especially for comments of a "design overview" nature, the information can instead be added to PR descriptions or separate documentation artifacts.
Edit: One thing that I did not state explicitly is that, among other issues, the comments were too AIX-specific.
tonykuttai
force-pushed
the
tvarghese/pragma-copyright-comment-update
branch
from
9116b1e to
7464d43
Compare
llvmorg-github-actions
Bot
added
the
LTO
- Emit !aix.copyright.comment from Clang for the pragma. - Lower it in LLVM to a TU-local string + llvm.used + !implicit.ref. - Add module-import and backend relocation tests.
Co-authored-by: Hubert Tong <hubert.reinterpretcast@gmail.com>
…imeComment() to the end of the class.
tonykuttai
force-pushed
the
tvarghese/pragma-copyright-comment-update
branch
from
5b07f43 to
c089860
Compare
This was referenced May 23, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
7 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This change supports pragma comment copyright.
Signature:
#pragma comment (copyright, "token_sequence")commentdirective can appear only once in a translation unit.Implementation is done through:
!comment_string.loadtimefrom Clang for the pragma.llvm.compiler.used+!implicit.refbyLowerCommentStringPass!implicit.refand emits.refdirectives into the object fileThis change builds on top of XCOFF associated metadata and was reviewed at [PowerPC][AIX] Support #pragma comment copyright for AIX
here is the documentation for
#pragma commentdirectives : https://www.ibm.com/docs/en/xl-c-and-cpp-aix/16.1.0?topic=descriptions-pragma-comment