merge main into amd-staging (llvm#1728)

Author: ranapratap55

bolt/lib/Target/RISCV/RISCVMCPlusBuilder.cpp

@@ -555,9 +555,9 @@ class RISCVMCPlusBuilder : public MCPlusBuilder {
                .addReg(RegCnt);
   }
 
-  InstructionListType createCmpJE(MCPhysReg RegNo, MCPhysReg RegTmp,
-                                  const MCSymbol *Target,
-                                  MCContext *Ctx) const {
+  InstructionListType createRegCmpJE(MCPhysReg RegNo, MCPhysReg RegTmp,
+                                     const MCSymbol *Target,
+                                     MCContext *Ctx) const {
     InstructionListType Insts;
     Insts.emplace_back(
         MCInstBuilder(RISCV::SUB).addReg(RegTmp).addReg(RegNo).addReg(RegNo));
@@ -718,7 +718,7 @@ class RISCVMCPlusBuilder : public MCPlusBuilder {
     Insts.emplace_back();
     loadReg(Insts.back(), RISCV::X10, RISCV::X10, 0);
     InstructionListType cmpJmp =
-        createCmpJE(RISCV::X10, RISCV::X11, IndCallHandler, Ctx);
+        createRegCmpJE(RISCV::X10, RISCV::X11, IndCallHandler, Ctx);
     Insts.insert(Insts.end(), cmpJmp.begin(), cmpJmp.end());
     Insts.emplace_back();
     createStackPointerIncrement(Insts.back(), 16);
@@ -777,14 +777,13 @@ class RISCVMCPlusBuilder : public MCPlusBuilder {
     return createGetter(Ctx, "__bolt_instr_num_funcs");
   }
 
-  void convertIndirectCallToLoad(MCInst &Inst, MCPhysReg Reg,
-                                 MCPhysReg ZeroReg) const {
+  void convertIndirectCallToLoad(MCInst &Inst, MCPhysReg Reg) override {
     bool IsTailCall = isTailCall(Inst);
     if (IsTailCall)
       removeAnnotation(Inst, MCPlus::MCAnnotation::kTailCall);
     Inst.setOpcode(RISCV::ADD);
     Inst.insert(Inst.begin(), MCOperand::createReg(Reg));
-    Inst.insert(Inst.begin() + 1, MCOperand::createReg(ZeroReg));
+    Inst.insert(Inst.begin() + 1, MCOperand::createReg(RISCV::X0));
     return;
   }
 
@@ -845,7 +844,7 @@ class RISCVMCPlusBuilder : public MCPlusBuilder {
     InstructionListType Insts;
     spillRegs(Insts, {RISCV::X10, RISCV::X11});
     Insts.emplace_back(CallInst);
-    convertIndirectCallToLoad(Insts.back(), RISCV::X10, RISCV::X0);
+    convertIndirectCallToLoad(Insts.back(), RISCV::X10);
     InstructionListType LoadImm = createLoadImmediate(RISCV::X11, CallSiteID);
     Insts.insert(Insts.end(), LoadImm.begin(), LoadImm.end());
     spillRegs(Insts, {RISCV::X10, RISCV::X11});

clang/cmake/caches/Fuchsia-stage2.cmake

@@ -459,6 +459,7 @@ set(LLVM_TOOLCHAIN_TOOLS
   llvm-libtool-darwin
   llvm-lipo
   llvm-ml
+  llvm-mt
   llvm-nm
   llvm-objcopy
   llvm-objdump
@@ -480,10 +481,6 @@ set(LLVM_TOOLCHAIN_TOOLS
   scan-build-py
   CACHE STRING "")
 
-if (LLVM_ENABLE_LIBXML2)
-  list(APPEND LLVM_TOOLCHAIN_TOOLS llvm-mt)
-endif()
-
 set(LLVM_Toolchain_DISTRIBUTION_COMPONENTS
   bolt
   clang

clang/docs/LibASTImporter.rst

@@ -468,7 +468,7 @@ Note, there may be several different ASTImporter objects which import into the s
         cxxRecordDecl(hasName("Y"), isDefinition()), ToUnit);
     ToYDef->dump();
     // An error is set for "ToYDef" in the shared state.
-    Optional<ASTImportError> OptErr =
+    std::optional<ASTImportError> OptErr =
         ImporterState->getImportDeclErrorIfAny(ToYDef);
     assert(OptErr);
 

clang/include/clang/AST/RecordLayout.h

@@ -75,8 +75,9 @@ class ASTRecordLayout {
   // performance or backwards compatibility preserving (e.g. AIX-ABI).
   CharUnits PreferredAlignment;
 
-  // UnadjustedAlignment - Maximum of the alignments of the record members in
-  // characters.
+  // UnadjustedAlignment - Alignment of record in characters before alignment
+  // adjustments. Maximum of the alignments of the record members and base
+  // classes in characters.
   CharUnits UnadjustedAlignment;
 
   /// RequiredAlignment - The required alignment of the object.  In the MS-ABI
@@ -186,7 +187,7 @@ class ASTRecordLayout {
   CharUnits getPreferredAlignment() const { return PreferredAlignment; }
 
   /// getUnadjustedAlignment - Get the record alignment in characters, before
-  /// alignment adjustement.
+  /// alignment adjustment.
   CharUnits getUnadjustedAlignment() const { return UnadjustedAlignment; }
 
   /// getSize - Get the record size in characters.

clang/include/clang/Driver/Options.td

@@ -8049,9 +8049,13 @@ def code_completion_with_fixits : Flag<["-"], "code-completion-with-fixits">,
 def skip_function_bodies : Flag<["-"], "skip-function-bodies">,
   HelpText<"Skip function bodies when possible">,
   MarshallingInfoFlag<FrontendOpts<"SkipFunctionBodies">>;
-def disable_free : Flag<["-"], "disable-free">,
-  HelpText<"Disable freeing of memory on exit">,
-  MarshallingInfoFlag<FrontendOpts<"DisableFree">>;
+defm disable_free : BoolOption<"",
+  "disable-free",
+  FrontendOpts<"DisableFree">,
+  DefaultFalse,
+  PosFlag<SetTrue, [], [ClangOption], "Disable">,
+  NegFlag<SetFalse, [], [ClangOption], "Don't disable">,
+  BothFlags<[], [ClangOption], " freeing of memory on exit">>;
 defm clear_ast_before_backend : BoolOption<"",
   "clear-ast-before-backend",
   CodeGenOpts<"ClearASTBeforeBackend">,

clang/lib/AST/ItaniumMangle.cpp

@@ -1405,16 +1405,6 @@ void CXXNameMangler::mangleUnresolvedPrefix(NestedNameSpecifier *qualifier,
     //   - a template template parameter with arguments
     // In all of these cases, we should have no prefix.
     if (NestedNameSpecifier *Prefix = qualifier->getPrefix()) {
-      if (const auto *DTST =
-              dyn_cast<DependentTemplateSpecializationType>(type)) {
-        Out << "srN";
-        TemplateName Template = getASTContext().getDependentTemplateName(
-            {Prefix, DTST->getDependentTemplateName().getName(),
-             /*HasTemplateKeyword=*/true});
-        mangleTemplatePrefix(Template);
-        mangleTemplateArgs(Template, DTST->template_arguments());
-        break;
-      }
       mangleUnresolvedPrefix(Prefix,
                              /*recursive=*/true);
     } else {
@@ -2618,7 +2608,8 @@ bool CXXNameMangler::mangleUnresolvedTypeOrSimpleId(QualType Ty,
         cast<DependentTemplateSpecializationType>(Ty);
     TemplateName Template = getASTContext().getDependentTemplateName(
         DTST->getDependentTemplateName());
-    mangleTemplatePrefix(Template);
+    const DependentTemplateStorage &S = DTST->getDependentTemplateName();
+    mangleSourceName(S.getName().getIdentifier());
     mangleTemplateArgs(Template, DTST->template_arguments());
     break;
   }

clang/lib/AST/RecordLayoutBuilder.cpp

@@ -1302,6 +1302,7 @@ ItaniumRecordLayoutBuilder::LayoutBase(const BaseSubobjectInfo *Base) {
     setSize(std::max(getSize(), Offset + Layout.getSize()));
 
   // Remember max struct/class alignment.
+  UnadjustedAlignment = std::max(UnadjustedAlignment, BaseAlign);
   UpdateAlignment(BaseAlign, UnpackedAlignTo, PreferredBaseAlign);
 
   return Offset;

clang/lib/CIR/CodeGen/CIRGenStmtOpenACC.cpp

@@ -124,7 +124,7 @@ class OpenACCClauseCIREmitter final
 
       for (const DeviceTypeArgument &arg : clause.getArchitectures()) {
         deviceTypes.push_back(mlir::acc::DeviceTypeAttr::get(
-            builder.getContext(), decodeDeviceType(arg.first)));
+            builder.getContext(), decodeDeviceType(arg.getIdentifierInfo())));
       }
       operation.removeDeviceTypesAttr();
       operation.setDeviceTypesAttr(
@@ -135,7 +135,7 @@ class OpenACCClauseCIREmitter final
 
       if (!clause.getArchitectures().empty())
         operation.setDeviceType(
-            decodeDeviceType(clause.getArchitectures()[0].first));
+            decodeDeviceType(clause.getArchitectures()[0].getIdentifierInfo()));
     } else {
       return clauseNotImplemented(clause);
     }

clang/lib/Driver/SanitizerArgs.cpp

@@ -262,11 +262,8 @@ static SanitizerMask setGroupBits(SanitizerMask Kinds) {
 }
 
 // Computes the sanitizer mask as:
-//     Default + Arguments (in or out)
+//     Default + Arguments (in or out) + AlwaysIn - AlwaysOut
 // with arguments parsed from left to right.
-//
-// Error messages are printed if the AlwaysIn or AlwaysOut invariants are
-// violated, but the caller must enforce these invariants themselves.
 static SanitizerMask
 parseSanitizeArgs(const Driver &D, const llvm::opt::ArgList &Args,
                   bool DiagnoseErrors, SanitizerMask Default,
@@ -316,6 +313,9 @@ parseSanitizeArgs(const Driver &D, const llvm::opt::ArgList &Args,
     }
   }
 
+  Output |= AlwaysIn;
+  Output &= ~AlwaysOut;
+
   return Output;
 }
 
@@ -325,10 +325,6 @@ static SanitizerMask parseSanitizeTrapArgs(const Driver &D,
   SanitizerMask AlwaysTrap; // Empty
   SanitizerMask NeverTrap = ~(setGroupBits(TrappingSupported));
 
-  // N.B. We do *not* enforce NeverTrap. This maintains the behavior of
-  // '-fsanitize=undefined -fsanitize-trap=undefined'
-  // (clang/test/Driver/fsanitize.c ), which is that vptr is not enabled at all
-  // (not even in recover mode) in order to avoid the need for a ubsan runtime.
   return parseSanitizeArgs(D, Args, DiagnoseErrors, TrappingDefault, AlwaysTrap,
                            NeverTrap, options::OPT_fsanitize_trap_EQ,
                            options::OPT_fno_sanitize_trap_EQ);
@@ -727,8 +723,6 @@ SanitizerArgs::SanitizerArgs(const ToolChain &TC,
       D, Args, DiagnoseErrors, RecoverableByDefault, AlwaysRecoverable,
       Unrecoverable, options::OPT_fsanitize_recover_EQ,
       options::OPT_fno_sanitize_recover_EQ);
-  RecoverableKinds |= AlwaysRecoverable;
-  RecoverableKinds &= ~Unrecoverable;
   RecoverableKinds &= Kinds;
 
   TrappingKinds &= Kinds;

clang/lib/Driver/ToolChains/Clang.cpp

@@ -9491,6 +9491,10 @@ void LinkerWrapper::ConstructJob(Compilation &C, const JobAction &JA,
               "--lto-newpm-passes=default-post-link<O" + OOpt + ">"));
         }
       }
+      // If this is OpenMP the device linker will need `-lompdevice`.
+      if (Kind == Action::OFK_OpenMP && !Args.hasArg(OPT_no_offloadlib) &&
+          (TC->getTriple().isAMDGPU() || TC->getTriple().isNVPTX()))
+        LinkerArgs.emplace_back("-lompdevice");
 
       // Forward all of these to the appropriate toolchain.
       for (StringRef Arg : CompilerArgs)