merge main into amd-staging

Change-Id: Iac6b5994548ea561594483180f27626ce78a3686

Author: ronlieb

clang/include/clang/Basic/Attr.td

@@ -2776,7 +2776,7 @@ def Ownership : InheritableAttr {
   let Args = [IdentifierArgument<"Module">,
               VariadicParamIdxArgument<"Args">];
   let Subjects = SubjectList<[HasFunctionProto]>;
-  let Documentation = [Undocumented];
+  let Documentation = [OwnershipDocs];
 }
 
 def Packed : InheritableAttr {

clang/include/clang/Basic/AttrDocs.td

@@ -1389,6 +1389,81 @@ Query for this attribute with ``__has_attribute(overloadable)``.
   }];
 }
 
+def OwnershipDocs : Documentation {
+  let Heading = "ownership_holds, ownership_returns, ownership_takes (Clang "
+                "Static Analyzer)";
+  let Category = DocCatFunction;
+  let Content = [{
+
+.. note::
+
+  In order for the Clang Static Analyzer to acknowledge these attributes, the
+  ``Optimistic`` config needs to be set to true for the checker
+  ``unix.DynamicMemoryModeling``:
+
+  ``-Xclang -analyzer-config -Xclang unix.DynamicMemoryModeling:Optimistic=true``
+
+These attributes are used by the Clang Static Analyzer's dynamic memory modeling
+facilities to mark custom allocating/deallocating functions.
+
+All 3 attributes' first parameter of type string is the type of the allocation:
+``malloc``, ``new``, etc. to allow for catching :ref:`mismatched deallocation
+<unix-MismatchedDeallocator>` bugs. The allocation type can be any string, e.g.
+a function annotated with
+returning a piece of memory of type ``lasagna`` but freed with a function
+annotated to release ``cheese`` typed memory will result in mismatched
+deallocation warning.
+
+The (currently) only allocation type having special meaning is ``malloc`` --
+the Clang Static Analyzer makes sure that allocating functions annotated with
+``malloc`` are treated like they used the standard ``malloc()``, and can be
+safely deallocated with the standard ``free()``.
+
+* Use ``ownership_returns`` to mark a function as an allocating function. Takes
+  1 parameter to denote the allocation type.
+* Use ``ownership_takes`` to mark a function as a deallocating function. Takes 2
+  parameters: the allocation type, and the index of the parameter that is being
+  deallocated (counting from 1).
+* Use ``ownership_holds`` to mark that a function takes over the ownership of a
+  piece of memory and will free it at some unspecified point in the future. Like
+  ``ownership_takes``, this takes 2 parameters: the allocation type, and the
+  index of the parameter whose ownership will be taken over (counting from 1).
+
+The annotations ``ownership_takes`` and ``ownership_holds`` both prevent memory
+leak reports (concerning the specified argument); the difference between them
+is that using taken memory is a use-after-free error, while using held memory
+is assumed to be legitimate.
+
+Example:
+
+.. code-block:: c
+
+  // Denotes that my_malloc will return with a dynamically allocated piece of
+  // memory using malloc().
+  void __attribute((ownership_returns(malloc))) *my_malloc(size_t);
+
+  // Denotes that my_free will deallocate its parameter using free().
+  void __attribute((ownership_takes(malloc, 1))) my_free(void *);
+
+  // Denotes that my_hold will take over the ownership of its parameter that was
+  // allocated via malloc().
+  void __attribute((ownership_holds(malloc, 1))) my_hold(void *);
+
+Further reading about dynamic memory modeling in the Clang Static Analyzer is
+found in these checker docs:
+:ref:`unix.Malloc <unix-Malloc>`, :ref:`unix.MallocSizeof <unix-MallocSizeof>`,
+:ref:`unix.MismatchedDeallocator <unix-MismatchedDeallocator>`,
+:ref:`cplusplus.NewDelete <cplusplus-NewDelete>`,
+:ref:`cplusplus.NewDeleteLeaks <cplusplus-NewDeleteLeaks>`,
+:ref:`optin.taint.TaintedAlloc <optin-taint-TaintedAlloc>`.
+Mind that many more checkers are affected by dynamic memory modeling changes to
+some extent.
+
+Further reading for other annotations:
+`Source Annotations in the Clang Static Analyzer <https://clang-analyzer.llvm.org/annotations.html>`_.
+  }];
+}
+
 def ObjCMethodFamilyDocs : Documentation {
   let Category = DocCatFunction;
   let Content = [{

clang/include/clang/Basic/arm_sme.td

@@ -716,6 +716,8 @@ let SMETargetGuard = "sme2" in {
   def SVZERO_ZT : Inst<"svzero_zt", "vi", "", MergeNone, "aarch64_sme_zero_zt", [IsOverloadNone, IsStreamingCompatible, IsOutZT0], [ImmCheck<0, ImmCheck0_0>]>;
 }
 
+def IN_STREAMING_MODE :  Inst<"__arm_in_streaming_mode", "sv", "Pc", MergeNone, "aarch64_sme_in_streaming_mode", [IsOverloadNone, IsStreamingCompatible], []>;
+
 //
 // lookup table expand four contiguous registers
 //

clang/lib/AST/ASTContext.cpp

@@ -14357,7 +14357,7 @@ QualType ASTContext::getCorrespondingSignedFixedPointType(QualType Ty) const {
 // corresponding backend features (which may contain duplicates).
 static std::vector<std::string> getFMVBackendFeaturesFor(
     const llvm::SmallVectorImpl<StringRef> &FMVFeatStrings) {
-  std::vector<std::string> BackendFeats{{"+fmv"}};
+  std::vector<std::string> BackendFeats;
   llvm::AArch64::ExtensionSet FeatureBits;
   for (StringRef F : FMVFeatStrings)
     if (auto FMVExt = llvm::AArch64::parseFMVExtension(F))

clang/lib/Basic/Targets/OSTargets.cpp

@@ -114,9 +114,6 @@ void getDarwinDefines(MacroBuilder &Builder, const LangOptions &Opts,
     assert(OsVersion.getMinor().value_or(0) < 100 &&
            OsVersion.getSubminor().value_or(0) < 100 && "Invalid version!");
     Builder.defineMacro("__ENVIRONMENT_OS_VERSION_MIN_REQUIRED__", Str);
-
-    // Tell users about the kernel if there is one.
-    Builder.defineMacro("__MACH__");
   }
 
   PlatformMinVersion = OsVersion;

clang/lib/CodeGen/CGBuiltin.cpp

@@ -11324,6 +11324,19 @@ Value *CodeGenFunction::EmitAArch64SMEBuiltinExpr(unsigned BuiltinID,
   if (Builtin->LLVMIntrinsic == 0)
     return nullptr;
 
+  if (BuiltinID == SME::BI__builtin_sme___arm_in_streaming_mode) {
+    // If we already know the streaming mode, don't bother with the intrinsic
+    // and emit a constant instead
+    const auto *FD = cast<FunctionDecl>(CurFuncDecl);
+    if (const auto *FPT = FD->getType()->getAs<FunctionProtoType>()) {
+      unsigned SMEAttrs = FPT->getAArch64SMEAttributes();
+      if (!(SMEAttrs & FunctionType::SME_PStateSMCompatibleMask)) {
+        bool IsStreaming = SMEAttrs & FunctionType::SME_PStateSMEnabledMask;
+        return ConstantInt::getBool(Builder.getContext(), IsStreaming);
+      }
+    }
+  }
+
   // Predicates must match the main datatype.
   for (unsigned i = 0, e = Ops.size(); i != e; ++i)
     if (auto PredTy = dyn_cast<llvm::VectorType>(Ops[i]->getType()))

clang/lib/CodeGen/CodeGenModule.cpp

@@ -2758,7 +2758,21 @@ bool CodeGenModule::GetCPUAndFeaturesAttributes(GlobalDecl GD,
     Attrs.addAttribute("target-features", llvm::join(Features, ","));
     AddedAttr = true;
   }
-
+  if (getTarget().getTriple().isAArch64()) {
+    llvm::SmallVector<StringRef, 8> Feats;
+    if (TV)
+      TV->getFeatures(Feats);
+    else if (TC)
+      TC->getFeatures(Feats, GD.getMultiVersionIndex());
+    if (!Feats.empty()) {
+      llvm::sort(Feats);
+      std::string FMVFeatures;
+      for (StringRef F : Feats)
+        FMVFeatures.append(",+" + F.str());
+      Attrs.addAttribute("fmv-features", FMVFeatures.substr(1));
+      AddedAttr = true;
+    }
+  }
   return AddedAttr;
 }
 

clang/lib/CodeGen/Targets/SPIR.cpp

@@ -64,6 +64,8 @@ class SPIRVTargetCodeGenInfo : public CommonSPIRTargetCodeGenInfo {
   void setCUDAKernelCallingConvention(const FunctionType *&FT) const override;
   LangAS getGlobalVarAddressSpace(CodeGenModule &CGM,
                                   const VarDecl *D) const override;
+  void setTargetAttributes(const Decl *D, llvm::GlobalValue *GV,
+                           CodeGen::CodeGenModule &M) const override;
   llvm::SyncScope::ID getLLVMSyncScopeID(const LangOptions &LangOpts,
                                          SyncScope Scope,
                                          llvm::AtomicOrdering Ordering,
@@ -245,6 +247,41 @@ SPIRVTargetCodeGenInfo::getGlobalVarAddressSpace(CodeGenModule &CGM,
   return DefaultGlobalAS;
 }
 
+void SPIRVTargetCodeGenInfo::setTargetAttributes(
+    const Decl *D, llvm::GlobalValue *GV, CodeGen::CodeGenModule &M) const {
+  if (!M.getLangOpts().HIP ||
+      M.getTarget().getTriple().getVendor() != llvm::Triple::AMD)
+    return;
+  if (GV->isDeclaration())
+    return;
+
+  auto F = dyn_cast<llvm::Function>(GV);
+  if (!F)
+    return;
+
+  auto FD = dyn_cast_or_null<FunctionDecl>(D);
+  if (!FD)
+    return;
+  if (!FD->hasAttr<CUDAGlobalAttr>())
+    return;
+
+  unsigned N = M.getLangOpts().GPUMaxThreadsPerBlock;
+  if (auto FlatWGS = FD->getAttr<AMDGPUFlatWorkGroupSizeAttr>())
+    N = FlatWGS->getMax()->EvaluateKnownConstInt(M.getContext()).getExtValue();
+
+  // We encode the maximum flat WG size in the first component of the 3D
+  // max_work_group_size attribute, which will get reverse translated into the
+  // original AMDGPU attribute when targeting AMDGPU.
+  auto Int32Ty = llvm::IntegerType::getInt32Ty(M.getLLVMContext());
+  llvm::Metadata *AttrMDArgs[] = {
+      llvm::ConstantAsMetadata::get(llvm::ConstantInt::get(Int32Ty, N)),
+      llvm::ConstantAsMetadata::get(llvm::ConstantInt::get(Int32Ty, 1)),
+      llvm::ConstantAsMetadata::get(llvm::ConstantInt::get(Int32Ty, 1))};
+
+  F->setMetadata("max_work_group_size",
+                 llvm::MDNode::get(M.getLLVMContext(), AttrMDArgs));
+}
+
 llvm::SyncScope::ID
 SPIRVTargetCodeGenInfo::getLLVMSyncScopeID(const LangOptions &, SyncScope Scope,
                                            llvm::AtomicOrdering,

clang/lib/Driver/Driver.cpp

@@ -6747,6 +6747,8 @@ const ToolChain &Driver::getToolChain(const ArgList &Args,
           TC = std::make_unique<toolchains::BareMetal>(*this, Target, Args);
         else if (Target.isOSBinFormatELF())
           TC = std::make_unique<toolchains::Generic_ELF>(*this, Target, Args);
+        else if (Target.isAppleMachO())
+          TC = std::make_unique<toolchains::AppleMachO>(*this, Target, Args);
         else if (Target.isOSBinFormatMachO())
           TC = std::make_unique<toolchains::MachO>(*this, Target, Args);
         else