Merge pull request llvm#306 from AMD-Lightning-Internal/amd/merge/upstream_merge_20250125084352

merge main into amd-staging

Author: ronlieb

.github/workflows/libc-fullbuild-tests.yml

@@ -11,12 +11,19 @@ on:
 
 jobs:
   build:
-    runs-on: ubuntu-24.04
+    runs-on: ${{ matrix.os }}
     strategy:
       fail-fast: false
       matrix:
         include:
-          - c_compiler: clang
+          - os: ubuntu-24.04
+            ccache-variant: sccache
+            c_compiler: clang
+            cpp_compiler: clang++
+          # TODO: remove ccache logic when https://github.com/hendrikmuhs/ccache-action/issues/279 is resolved.
+          - os: ubuntu-24.04-arm
+            ccache-variant: ccache
+            c_compiler: clang
             cpp_compiler: clang++
           # TODO: add back gcc build when it is fixed
           # - c_compiler: gcc
@@ -35,7 +42,7 @@ jobs:
       with:
         max-size: 1G
         key: libc_fullbuild_${{ matrix.c_compiler }}
-        variant: sccache
+        variant: ${{ matrix.ccache-variant }}
 
     # Notice:
     # - MPFR is required by some of the mathlib tests.
@@ -62,8 +69,8 @@ jobs:
         -DCMAKE_CXX_COMPILER=${{ matrix.cpp_compiler }}
         -DCMAKE_C_COMPILER=${{ matrix.c_compiler }}
         -DCMAKE_BUILD_TYPE=MinSizeRel
-        -DCMAKE_C_COMPILER_LAUNCHER=sccache
-        -DCMAKE_CXX_COMPILER_LAUNCHER=sccache
+        -DCMAKE_C_COMPILER_LAUNCHER=${{ matrix.ccache-variant }}
+        -DCMAKE_CXX_COMPILER_LAUNCHER=${{ matrix.ccache-variant }}
         -DCMAKE_INSTALL_PREFIX=${{ steps.strings.outputs.build-install-dir }}
         -DLLVM_ENABLE_RUNTIMES="libc;compiler-rt"
         -DLLVM_LIBC_FULL_BUILD=ON

.github/workflows/libc-overlay-tests.yml

@@ -19,14 +19,28 @@ jobs:
         include:
           # TODO: add linux gcc when it is fixed
           - os: ubuntu-24.04
+            ccache-variant: sccache
+            compiler:
+              c_compiler: clang
+              cpp_compiler: clang++
+          # TODO: remove ccache logic when https://github.com/hendrikmuhs/ccache-action/issues/279 is resolved.
+          - os: ubuntu-24.04-arm
+            ccache-variant: ccache
             compiler:
               c_compiler: clang
               cpp_compiler: clang++
           - os: windows-2022
+            ccache-variant: sccache
+            compiler:
+              c_compiler: clang-cl
+              cpp_compiler: clang-cl
+          - os: windows-2025
+            ccache-variant: sccache
             compiler:
               c_compiler: clang-cl
               cpp_compiler: clang-cl
           - os: macos-14
+            ccache-variant: sccache
             compiler:
               c_compiler: clang
               cpp_compiler: clang++
@@ -46,7 +60,7 @@ jobs:
       with:
         max-size: 1G
         key: libc_overlay_build_${{ matrix.os }}_${{ matrix.compiler.c_compiler }}
-        variant: sccache
+        variant: ${{ matrix.ccache-variant }}
 
     # MPFR is required by some of the mathlib tests.
     - name: Prepare dependencies (Ubuntu)
@@ -82,8 +96,8 @@ jobs:
         -DCMAKE_CXX_COMPILER=${{ matrix.compiler.cpp_compiler }}
         -DCMAKE_C_COMPILER=${{ matrix.compiler.c_compiler }}
         -DCMAKE_BUILD_TYPE=MinSizeRel
-        -DCMAKE_C_COMPILER_LAUNCHER=sccache
-        -DCMAKE_CXX_COMPILER_LAUNCHER=sccache
+        -DCMAKE_C_COMPILER_LAUNCHER=${{ matrix.ccache-variant }}
+        -DCMAKE_CXX_COMPILER_LAUNCHER=${{ matrix.ccache-variant }}
         -DCMAKE_POLICY_DEFAULT_CMP0141=NEW
         -DCMAKE_MSVC_DEBUG_INFORMATION_FORMAT=Embedded
         -DLLVM_ENABLE_RUNTIMES=libc

clang-tools-extra/clang-tidy/ClangTidyProfiling.cpp

@@ -36,23 +36,25 @@ ClangTidyProfiling::StorageParams::StorageParams(llvm::StringRef ProfilePrefix,
                       .str();
 }
 
-void ClangTidyProfiling::printUserFriendlyTable(llvm::raw_ostream &OS) {
-  TG->print(OS);
+void ClangTidyProfiling::printUserFriendlyTable(llvm::raw_ostream &OS,
+                                                llvm::TimerGroup &TG) {
+  TG.print(OS);
   OS.flush();
 }
 
-void ClangTidyProfiling::printAsJSON(llvm::raw_ostream &OS) {
+void ClangTidyProfiling::printAsJSON(llvm::raw_ostream &OS,
+                                     llvm::TimerGroup &TG) {
   OS << "{\n";
   OS << R"("file": ")" << Storage->SourceFilename << "\",\n";
   OS << R"("timestamp": ")" << Storage->Timestamp << "\",\n";
   OS << "\"profile\": {\n";
-  TG->printJSONValues(OS, "");
+  TG.printJSONValues(OS, "");
   OS << "\n}\n";
   OS << "}\n";
   OS.flush();
 }
 
-void ClangTidyProfiling::storeProfileData() {
+void ClangTidyProfiling::storeProfileData(llvm::TimerGroup &TG) {
   assert(Storage && "We should have a filename.");
 
   llvm::SmallString<256> OutputDirectory(Storage->StoreFilename);
@@ -71,19 +73,18 @@ void ClangTidyProfiling::storeProfileData() {
     return;
   }
 
-  printAsJSON(OS);
+  printAsJSON(OS, TG);
 }
 
 ClangTidyProfiling::ClangTidyProfiling(std::optional<StorageParams> Storage)
     : Storage(std::move(Storage)) {}
 
 ClangTidyProfiling::~ClangTidyProfiling() {
-  TG.emplace("clang-tidy", "clang-tidy checks profiling", Records);
-
+  llvm::TimerGroup TG{"clang-tidy", "clang-tidy checks profiling", Records};
   if (!Storage)
-    printUserFriendlyTable(llvm::errs());
+    printUserFriendlyTable(llvm::errs(), TG);
   else
-    storeProfileData();
+    storeProfileData(TG);
 }
 
 } // namespace clang::tidy

clang-tools-extra/clang-tidy/ClangTidyProfiling.h

@@ -34,14 +34,11 @@ class ClangTidyProfiling {
   };
 
 private:
-  std::optional<llvm::TimerGroup> TG;
-
   std::optional<StorageParams> Storage;
 
-  void printUserFriendlyTable(llvm::raw_ostream &OS);
-  void printAsJSON(llvm::raw_ostream &OS);
-
-  void storeProfileData();
+  void printUserFriendlyTable(llvm::raw_ostream &OS, llvm::TimerGroup &TG);
+  void printAsJSON(llvm::raw_ostream &OS, llvm::TimerGroup &TG);
+  void storeProfileData(llvm::TimerGroup &TG);
 
 public:
   llvm::StringMap<llvm::TimeRecord> Records;

clang-tools-extra/clang-tidy/misc/UnusedUsingDeclsCheck.cpp

@@ -51,6 +51,10 @@ UnusedUsingDeclsCheck::UnusedUsingDeclsCheck(StringRef Name,
       HeaderFileExtensions(Context->getHeaderFileExtensions()) {}
 
 void UnusedUsingDeclsCheck::registerMatchers(MatchFinder *Finder) {
+  // We don't emit warnings on unused-using-decls from headers, so bail out if
+  // the main file is a header.
+  if (utils::isFileExtension(getCurrentMainFile(), HeaderFileExtensions))
+    return;
   Finder->addMatcher(usingDecl(isExpansionInMainFile()).bind("using"), this);
   auto DeclMatcher = hasDeclaration(namedDecl().bind("used"));
   Finder->addMatcher(loc(templateSpecializationType(DeclMatcher)), this);
@@ -83,12 +87,6 @@ void UnusedUsingDeclsCheck::registerMatchers(MatchFinder *Finder) {
 void UnusedUsingDeclsCheck::check(const MatchFinder::MatchResult &Result) {
   if (Result.Context->getDiagnostics().hasUncompilableErrorOccurred())
     return;
-  // We don't emit warnings on unused-using-decls from headers, so bail out if
-  // the main file is a header.
-  if (auto MainFile = Result.SourceManager->getFileEntryRefForID(
-          Result.SourceManager->getMainFileID());
-      utils::isFileExtension(MainFile->getName(), HeaderFileExtensions))
-    return;
 
   if (const auto *Using = Result.Nodes.getNodeAs<UsingDecl>("using")) {
     // Ignores using-declarations defined in macros.

clang/include/clang/Basic/DiagnosticDriverKinds.td

@@ -458,6 +458,10 @@ def warn_drv_deprecated_arg_ofast : Warning<
   "argument '-Ofast' is deprecated; use '-O3 -ffast-math' for the same behavior,"
   " or '-O3' to enable only conforming optimizations">,
   InGroup<DeprecatedOFast>;
+def warn_drv_deprecated_arg_ofast_for_flang : Warning<
+  "argument '-Ofast' is deprecated; use '-O3 -ffast-math -fstack-arrays' for the same behavior,"
+  " or '-O3 -fstack-arrays' to enable only conforming optimizations">,
+  InGroup<DeprecatedOFast>;
 def warn_drv_deprecated_custom : Warning<
   "argument '%0' is deprecated, %1">, InGroup<Deprecated>;
 def warn_drv_assuming_mfloat_abi_is : Warning<

clang/include/clang/Driver/Options.td

@@ -933,10 +933,12 @@ def O : Joined<["-"], "O">, Group<O_Group>,
 def O_flag : Flag<["-"], "O">, Visibility<[ClangOption, CC1Option, FC1Option]>,
   Alias<O>, AliasArgs<["1"]>;
 def Ofast : Joined<["-"], "Ofast">, Group<O_Group>,
-  Visibility<[ClangOption, CC1Option, FlangOption]>,
-  HelpTextForVariants<[ClangOption, CC1Option],
-                      "Deprecated; use '-O3 -ffast-math' for the same behavior,"
-                      " or '-O3' to enable only conforming optimizations">;
+  Visibility<[ClangOption, CC1Option, FlangOption, FC1Option]>,
+  HelpTextForVariants<[FlangOption, FC1Option],
+    "Deprecated; use '-O3 -ffast-math -fstack-arrays' for the same behavior,"
+    " or '-O3 -fstack-arrays' to enable only conforming optimizations">,
+  HelpText<"Deprecated; use '-O3 -ffast-math' for the same behavior,"
+           " or '-O3' to enable only conforming optimizations">;
 def P : Flag<["-"], "P">,
   Visibility<[ClangOption, CC1Option, FlangOption, FC1Option]>,
   Group<Preprocessor_Group>,

clang/include/clang/Sema/SemaInternal.h

@@ -72,7 +72,7 @@ inline std::pair<unsigned, unsigned> getDepthAndIndex(const NamedDecl *ND) {
 /// Retrieve the depth and index of an unexpanded parameter pack.
 inline std::pair<unsigned, unsigned>
 getDepthAndIndex(UnexpandedParameterPack UPP) {
-  if (const auto *TTP = UPP.first.dyn_cast<const TemplateTypeParmType *>())
+  if (const auto *TTP = dyn_cast<const TemplateTypeParmType *>(UPP.first))
     return std::make_pair(TTP->getDepth(), TTP->getIndex());
 
   return getDepthAndIndex(cast<NamedDecl *>(UPP.first));

clang/lib/AST/ByteCode/Descriptor.cpp

@@ -428,17 +428,17 @@ QualType Descriptor::getElemQualType() const {
 }
 
 SourceLocation Descriptor::getLocation() const {
-  if (auto *D = Source.dyn_cast<const Decl *>())
+  if (auto *D = dyn_cast<const Decl *>(Source))
     return D->getLocation();
-  if (auto *E = Source.dyn_cast<const Expr *>())
+  if (auto *E = dyn_cast<const Expr *>(Source))
     return E->getExprLoc();
   llvm_unreachable("Invalid descriptor type");
 }
 
 SourceInfo Descriptor::getLoc() const {
-  if (const auto *D = Source.dyn_cast<const Decl *>())
+  if (const auto *D = dyn_cast<const Decl *>(Source))
     return SourceInfo(D);
-  if (const auto *E = Source.dyn_cast<const Expr *>())
+  if (const auto *E = dyn_cast<const Expr *>(Source))
     return SourceInfo(E);
   llvm_unreachable("Invalid descriptor type");
 }

clang/lib/AST/ByteCode/Interp.cpp

@@ -873,13 +873,17 @@ bool CheckNewDeleteForms(InterpState &S, CodePtr OpPC,
 
 bool CheckDeleteSource(InterpState &S, CodePtr OpPC, const Expr *Source,
                        const Pointer &Ptr) {
-  // The two sources we currently allow are new expressions and
-  // __builtin_operator_new calls.
+  // Regular new type(...) call.
   if (isa_and_nonnull<CXXNewExpr>(Source))
     return true;
-  if (const CallExpr *CE = dyn_cast_if_present<CallExpr>(Source);
+  // operator new.
+  if (const auto *CE = dyn_cast_if_present<CallExpr>(Source);
       CE && CE->getBuiltinCallee() == Builtin::BI__builtin_operator_new)
     return true;
+  // std::allocator.allocate() call
+  if (const auto *MCE = dyn_cast_if_present<CXXMemberCallExpr>(Source);
+      MCE && MCE->getMethodDecl()->getIdentifier()->isStr("allocate"))
+    return true;
 
   // Whatever this is, we didn't heap allocate it.
   const SourceInfo &Loc = S.Current->getSource(OpPC);
@@ -1489,7 +1493,8 @@ bool CheckNewTypeMismatch(InterpState &S, CodePtr OpPC, const Expr *E,
   const auto *NewExpr = cast<CXXNewExpr>(E);
   QualType StorageType = Ptr.getType();
 
-  if (isa_and_nonnull<CXXNewExpr>(Ptr.getFieldDesc()->asExpr()) &&
+  if ((isa_and_nonnull<CXXNewExpr>(Ptr.getFieldDesc()->asExpr()) ||
+       isa_and_nonnull<CXXMemberCallExpr>(Ptr.getFieldDesc()->asExpr())) &&
       StorageType->isPointerType()) {
     // FIXME: Are there other cases where this is a problem?
     StorageType = StorageType->getPointeeType();

clang/lib/AST/ByteCode/InterpBuiltin.cpp

@@ -1584,6 +1584,7 @@ static bool interp__builtin_operator_new(InterpState &S, CodePtr OpPC,
   // Walk up the call stack to find the appropriate caller and get the
   // element type from it.
   QualType ElemType;
+  const CallExpr *NewCall = nullptr;
 
   for (const InterpFrame *F = Frame; F; F = F->Caller) {
     const Function *Func = F->getFunction();
@@ -1606,6 +1607,7 @@ static bool interp__builtin_operator_new(InterpState &S, CodePtr OpPC,
     if (CTSD->isInStdNamespace() && ClassII && ClassII->isStr("allocator") &&
         TAL.size() >= 1 && TAL[0].getKind() == TemplateArgument::Type) {
       ElemType = TAL[0].getAsType();
+      NewCall = cast<CallExpr>(F->Caller->getExpr(F->getRetPC()));
       break;
     }
   }
@@ -1616,6 +1618,7 @@ static bool interp__builtin_operator_new(InterpState &S, CodePtr OpPC,
                        : diag::note_constexpr_new);
     return false;
   }
+  assert(NewCall);
 
   if (ElemType->isIncompleteType() || ElemType->isFunctionType()) {
     S.FFDiag(Call, diag::note_constexpr_new_not_complete_object_type)
@@ -1654,7 +1657,7 @@ static bool interp__builtin_operator_new(InterpState &S, CodePtr OpPC,
   if (ElemT) {
     if (NumElems.ule(1)) {
       const Descriptor *Desc =
-          S.P.createDescriptor(Call, *ElemT, Descriptor::InlineDescMD,
+          S.P.createDescriptor(NewCall, *ElemT, Descriptor::InlineDescMD,
                                /*IsConst=*/false, /*IsTemporary=*/false,
                                /*IsMutable=*/false);
       Block *B = Allocator.allocate(Desc, S.getContext().getEvalID(),
@@ -1667,7 +1670,7 @@ static bool interp__builtin_operator_new(InterpState &S, CodePtr OpPC,
     assert(NumElems.ugt(1));
 
     Block *B =
-        Allocator.allocate(Call, *ElemT, NumElems.getZExtValue(),
+        Allocator.allocate(NewCall, *ElemT, NumElems.getZExtValue(),
                            S.Ctx.getEvalID(), DynamicAllocator::Form::Operator);
     assert(B);
     S.Stk.push<Pointer>(B);

clang/lib/CodeGen/CGCoroutine.cpp

@@ -626,7 +626,7 @@ struct CallCoroDelete final : public EHScopeStack::Cleanup {
 
     // Get back to the block we were originally and move coro.free there.
     auto *InsertPt = SaveInsertBlock->getTerminator();
-    CoroFree->moveBefore(InsertPt);
+    CoroFree->moveBefore(InsertPt->getIterator());
     CGF.Builder.SetInsertPoint(InsertPt);
 
     // Add if (auto *mem = coro.free) Deallocate;

clang/lib/CodeGen/CGException.cpp

@@ -1251,11 +1251,12 @@ void CodeGenFunction::ExitCXXTryStmt(const CXXTryStmt &S, bool IsFnTryBlock) {
   llvm::BasicBlock *WasmCatchStartBlock = nullptr;
   if (EHPersonality::get(*this).isWasmPersonality()) {
     auto *CatchSwitch =
-        cast<llvm::CatchSwitchInst>(DispatchBlock->getFirstNonPHI());
+        cast<llvm::CatchSwitchInst>(DispatchBlock->getFirstNonPHIIt());
     WasmCatchStartBlock = CatchSwitch->hasUnwindDest()
                               ? CatchSwitch->getSuccessor(1)
                               : CatchSwitch->getSuccessor(0);
-    auto *CPI = cast<llvm::CatchPadInst>(WasmCatchStartBlock->getFirstNonPHI());
+    auto *CPI =
+        cast<llvm::CatchPadInst>(WasmCatchStartBlock->getFirstNonPHIIt());
     CurrentFuncletPad = CPI;
   }
 
@@ -1858,7 +1859,7 @@ Address CodeGenFunction::recoverAddrOfEscapedLocal(CodeGenFunction &ParentCGF,
            "expected alloca or localrecover in parent LocalDeclMap");
     RecoverCall = cast<llvm::CallInst>(ParentRecover->clone());
     RecoverCall->setArgOperand(1, ParentFP);
-    RecoverCall->insertBefore(AllocaInsertPt);
+    RecoverCall->insertBefore(AllocaInsertPt->getIterator());
   }
 
   // Bitcast the variable, rename it, and insert it in the local decl map.
@@ -2252,7 +2253,7 @@ void CodeGenFunction::ExitSEHTryStmt(const SEHTryStmt &S) {
   // __except blocks don't get outlined into funclets, so immediately do a
   // catchret.
   llvm::CatchPadInst *CPI =
-      cast<llvm::CatchPadInst>(CatchPadBB->getFirstNonPHI());
+      cast<llvm::CatchPadInst>(CatchPadBB->getFirstNonPHIIt());
   llvm::BasicBlock *ExceptBB = createBasicBlock("__except");
   Builder.CreateCatchRet(CPI, ExceptBB);
   EmitBlock(ExceptBB);

clang/lib/CodeGen/CGOpenMPRuntime.cpp

@@ -1355,7 +1355,7 @@ void CGOpenMPRuntime::setLocThreadIdInsertPt(CodeGenFunction &CGF,
                                                  CGF.Builder.GetInsertBlock());
   } else {
     Elem.ServiceInsertPt = new llvm::BitCastInst(Undef, CGF.Int32Ty, "svcpt");
-    Elem.ServiceInsertPt->insertAfter(CGF.AllocaInsertPt);
+    Elem.ServiceInsertPt->insertAfter(CGF.AllocaInsertPt->getIterator());
   }
 }
 

clang/lib/CodeGen/MicrosoftCXXABI.cpp

@@ -918,7 +918,7 @@ void MicrosoftCXXABI::emitBeginCatch(CodeGenFunction &CGF,
   VarDecl *CatchParam = S->getExceptionDecl();
   llvm::BasicBlock *CatchPadBB = CGF.Builder.GetInsertBlock();
   llvm::CatchPadInst *CPI =
-      cast<llvm::CatchPadInst>(CatchPadBB->getFirstNonPHI());
+      cast<llvm::CatchPadInst>(CatchPadBB->getFirstNonPHIIt());
   CGF.CurrentFuncletPad = CPI;
 
   // If this is a catch-all or the catch parameter is unnamed, we don't need to

clang/lib/Driver/ToolChains/Flang.cpp

@@ -937,6 +937,7 @@ void Flang::ConstructJob(Compilation &C, const JobAction &JA,
       D.Diag(diag::warn_O4_is_O3);
     } else if (A->getOption().matches(options::OPT_Ofast)) {
       CmdArgs.push_back("-O3");
+      D.Diag(diag::warn_drv_deprecated_arg_ofast_for_flang);
     } else {
       A->render(Args, CmdArgs);
     }