Merge "merge main into amd-staging" into amd-staging

Author: ronlieb

.github/workflows/libcxx-build-and-test.yaml

@@ -79,6 +79,7 @@ jobs:
           path: |
             **/test-results.xml
             **/*.abilist
+            **/CMakeConfigureLog.yaml
             **/CMakeError.log
             **/CMakeOutput.log
             **/crash_diagnostics/*
@@ -123,6 +124,7 @@ jobs:
           path: |
             **/test-results.xml
             **/*.abilist
+            **/CMakeConfigureLog.yaml
             **/CMakeError.log
             **/CMakeOutput.log
             **/crash_diagnostics/*
@@ -188,6 +190,7 @@ jobs:
           path: |
             **/test-results.xml
             **/*.abilist
+            **/CMakeConfigureLog.yaml
             **/CMakeError.log
             **/CMakeOutput.log
             **/crash_diagnostics/*
@@ -230,6 +233,7 @@ jobs:
           path: |
             **/test-results.xml
             **/*.abilist
+            **/CMakeConfigureLog.yaml
             **/CMakeError.log
             **/CMakeOutput.log
             **/crash_diagnostics/*

clang-tools-extra/clang-tidy/readability/FunctionCognitiveComplexityCheck.cpp

@@ -126,7 +126,7 @@ struct CognitiveComplexity final {
   // Limit of 25 is the "upstream"'s default.
   static constexpr unsigned DefaultLimit = 25U;
 
-  // Based on the publicly-avaliable numbers for some big open-source projects
+  // Based on the publicly-available numbers for some big open-source projects
   // https://sonarcloud.io/projects?languages=c%2Ccpp&size=5   we can estimate:
   // value ~20 would result in no allocs for 98% of functions, ~12 for 96%, ~10
   // for 91%, ~8 for 88%, ~6 for 84%, ~4 for 77%, ~2 for 64%, and ~1 for 37%.

clang-tools-extra/test/clang-tidy/checkers/readability/function-cognitive-complexity.cpp

@@ -69,10 +69,10 @@ void unittest_false() {
 //----------------------------------------------------------------------------//
 
 // break does not increase cognitive complexity.
-// only  break LABEL  does, but it is unavaliable in C or C++
+// only  break LABEL  does, but it is unavailable in C or C++
 
 // continue does not increase cognitive complexity.
-// only  continue LABEL  does, but it is unavaliable in C or C++
+// only  continue LABEL  does, but it is unavailable in C or C++
 
 void unittest_b1_00() {
 // CHECK-NOTES: :[[@LINE-1]]:6: warning: function 'unittest_b1_00' has cognitive complexity of 33 (threshold 0) [readability-function-cognitive-complexity]

clang/include/clang/Basic/DiagnosticSemaKinds.td

@@ -12718,7 +12718,18 @@ def err_acc_num_arg_conflict
 def err_acc_clause_in_clause_region
     : Error<"loop with a '%0' clause may not exist in the region of a '%1' "
             "clause%select{| on a 'kernels' compute construct}2">;
-
+def err_acc_gang_reduction_conflict
+    : Error<"%select{OpenACC 'gang' clause with a 'dim' value greater than "
+            "1|OpenACC 'reduction' clause}0 cannot "
+            "appear on the same 'loop' construct as a %select{'reduction' "
+            "clause|'gang' clause with a 'dim' value greater than 1}0">;
+def err_acc_gang_reduction_numgangs_conflict
+    : Error<"OpenACC '%0' clause cannot appear on the same 'loop' construct "
+            "as a '%1' clause inside a compute construct with a "
+            "'num_gangs' clause with more than one argument">;
+def err_reduction_op_mismatch
+    : Error<"OpenACC 'reduction' variable must have the same operator in all "
+            "nested constructs (%0 vs %1)">;
 // AMDGCN builtins diagnostics
 def err_amdgcn_global_load_lds_size_invalid_value : Error<"invalid size value">;
 def note_amdgcn_global_load_lds_size_valid_value : Note<"size must be 1, 2, or 4">;

clang/include/clang/Sema/SemaOpenACC.h

@@ -68,6 +68,7 @@ class SemaOpenACC : public SemaBase {
     /// used to diagnose if there are multiple 'for' loops at any one level.
     LLVM_PREFERRED_TYPE(bool)
     unsigned CurLevelHasLoopAlready : 1;
+
   } LoopInfo{/*TopLevelLoopSeen=*/false, /*CurLevelHasLoopAlready=*/false};
 
   /// The 'collapse' clause requires quite a bit of checking while
@@ -109,6 +110,14 @@ class SemaOpenACC : public SemaBase {
     bool TileDepthSatisfied = true;
   } TileInfo;
 
+  /// A list of the active reduction clauses, which allows us to check that all
+  /// vars on nested constructs for the same reduction var have the same
+  /// reduction operator. Currently this is enforced against all constructs
+  /// despite the rule being in the 'loop' section. By current reading, this
+  /// should apply to all anyway, but we may need to make this more like the
+  /// 'loop' clause enforcement, where this is 'blocked' by a compute construct.
+  llvm::SmallVector<OpenACCReductionClause *> ActiveReductionClauses;
+
 public:
   ComputeConstructInfo &getActiveComputeConstructInfo() {
     return ActiveComputeConstructInfo;
@@ -615,7 +624,9 @@ class SemaOpenACC : public SemaBase {
 
   /// Called while semantically analyzing the reduction clause, ensuring the var
   /// is the correct kind of reference.
-  ExprResult CheckReductionVar(Expr *VarExpr);
+  ExprResult CheckReductionVar(OpenACCDirectiveKind DirectiveKind,
+                               OpenACCReductionOperator ReductionOp,
+                               Expr *VarExpr);
 
   /// Called to check the 'var' type is a variable of pointer type, necessary
   /// for 'deviceptr' and 'attach' clauses. Returns true on success.
@@ -634,6 +645,22 @@ class SemaOpenACC : public SemaBase {
   // Check a single expression on a gang clause.
   ExprResult CheckGangExpr(OpenACCGangKind GK, Expr *E);
 
+  // Does the checking for a 'gang' clause that needs to be done in dependent
+  // and not dependent cases.
+  OpenACCClause *
+  CheckGangClause(ArrayRef<const OpenACCClause *> ExistingClauses,
+                  SourceLocation BeginLoc, SourceLocation LParenLoc,
+                  ArrayRef<OpenACCGangKind> GangKinds,
+                  ArrayRef<Expr *> IntExprs, SourceLocation EndLoc);
+  // Does the checking for a 'reduction ' clause that needs to be done in
+  // dependent and not dependent cases.
+  OpenACCClause *
+  CheckReductionClause(ArrayRef<const OpenACCClause *> ExistingClauses,
+                       OpenACCDirectiveKind DirectiveKind,
+                       SourceLocation BeginLoc, SourceLocation LParenLoc,
+                       OpenACCReductionOperator ReductionOp,
+                       ArrayRef<Expr *> Vars, SourceLocation EndLoc);
+
   ExprResult BuildOpenACCAsteriskSizeExpr(SourceLocation AsteriskLoc);
   ExprResult ActOnOpenACCAsteriskSizeExpr(SourceLocation AsteriskLoc);
 
@@ -686,6 +713,7 @@ class SemaOpenACC : public SemaBase {
     SourceLocation OldLoopWorkerClauseLoc;
     SourceLocation OldLoopVectorClauseLoc;
     llvm::SmallVector<OpenACCLoopConstruct *> ParentlessLoopConstructs;
+    llvm::SmallVector<OpenACCReductionClause *> ActiveReductionClauses;
     LoopInConstructRAII LoopRAII;
 
   public:

clang/lib/AST/ByteCode/Interp.h

@@ -3062,7 +3062,8 @@ inline bool BitCast(InterpState &S, CodePtr OpPC, bool TargetIsUCharOrByte,
     return false;
 
   if constexpr (std::is_same_v<T, Floating>) {
-    assert(false && "Implement bitcasting to a floating type");
+    assert(Sem);
+    S.Stk.push<Floating>(T::bitcastFromMemory(Buff.data(), *Sem));
   } else {
     assert(!Sem);
     S.Stk.push<T>(T::bitcastFromMemory(Buff.data(), ResultBitWidth));

clang/lib/AST/ByteCode/InterpBuiltin.cpp

@@ -1254,7 +1254,7 @@ static bool interp__builtin_ia32_bextr(InterpState &S, CodePtr OpPC,
                                        const InterpFrame *Frame,
                                        const Function *Func,
                                        const CallExpr *Call) {
-  if (!Call->getArg(0)->getType()->isIntegerType() ||
+  if (Call->getNumArgs() != 2 || !Call->getArg(0)->getType()->isIntegerType() ||
       !Call->getArg(1)->getType()->isIntegerType())
     return false;
 
@@ -1286,7 +1286,9 @@ static bool interp__builtin_ia32_bzhi(InterpState &S, CodePtr OpPC,
                                       const Function *Func,
                                       const CallExpr *Call) {
   QualType CallType = Call->getType();
-  if (!CallType->isIntegerType())
+  if (Call->getNumArgs() != 2 || !Call->getArg(0)->getType()->isIntegerType() ||
+      !Call->getArg(1)->getType()->isIntegerType() ||
+      !CallType->isIntegerType())
     return false;
 
   PrimType ValT = *S.Ctx.classify(Call->getArg(0));
@@ -1311,7 +1313,8 @@ static bool interp__builtin_ia32_lzcnt(InterpState &S, CodePtr OpPC,
                                        const Function *Func,
                                        const CallExpr *Call) {
   QualType CallType = Call->getType();
-  if (!CallType->isIntegerType())
+  if (!CallType->isIntegerType() ||
+      !Call->getArg(0)->getType()->isIntegerType())
     return false;
 
   APSInt Val = peekToAPSInt(S.Stk, *S.Ctx.classify(Call->getArg(0)));
@@ -1324,7 +1327,8 @@ static bool interp__builtin_ia32_tzcnt(InterpState &S, CodePtr OpPC,
                                        const Function *Func,
                                        const CallExpr *Call) {
   QualType CallType = Call->getType();
-  if (!CallType->isIntegerType())
+  if (!CallType->isIntegerType() ||
+      !Call->getArg(0)->getType()->isIntegerType())
     return false;
 
   APSInt Val = peekToAPSInt(S.Stk, *S.Ctx.classify(Call->getArg(0)));
@@ -1336,7 +1340,7 @@ static bool interp__builtin_ia32_pdep(InterpState &S, CodePtr OpPC,
                                       const InterpFrame *Frame,
                                       const Function *Func,
                                       const CallExpr *Call) {
-  if (!Call->getArg(0)->getType()->isIntegerType() ||
+  if (Call->getNumArgs() != 2 || !Call->getArg(0)->getType()->isIntegerType() ||
       !Call->getArg(1)->getType()->isIntegerType())
     return false;
 
@@ -1361,7 +1365,7 @@ static bool interp__builtin_ia32_pext(InterpState &S, CodePtr OpPC,
                                       const InterpFrame *Frame,
                                       const Function *Func,
                                       const CallExpr *Call) {
-  if (!Call->getArg(0)->getType()->isIntegerType() ||
+  if (Call->getNumArgs() != 2 || !Call->getArg(0)->getType()->isIntegerType() ||
       !Call->getArg(1)->getType()->isIntegerType())
     return false;
 

clang/lib/Driver/ToolChains/PS4CPU.cpp

@@ -248,8 +248,9 @@ void tools::PS5cpu::Linker::ConstructJob(Compilation &C, const JobAction &JA,
       Args.MakeArgString("--sysroot=" + TC.getSDKLibraryRootDir()));
 
   // Default to PIE for non-static executables.
-  const bool PIE = !Relocatable && !Shared && !Static;
-  if (Args.hasFlag(options::OPT_pie, options::OPT_no_pie, PIE))
+  const bool PIE = Args.hasFlag(options::OPT_pie, options::OPT_no_pie,
+                                !Relocatable && !Shared && !Static);
+  if (PIE)
     CmdArgs.push_back("-pie");
 
   if (!Relocatable) {
@@ -276,6 +277,12 @@ void tools::PS5cpu::Linker::ConstructJob(Compilation &C, const JobAction &JA,
     CmdArgs.push_back("-z");
     CmdArgs.push_back("start-stop-visibility=hidden");
 
+    CmdArgs.push_back("-z");
+    CmdArgs.push_back("common-page-size=0x4000");
+
+    CmdArgs.push_back("-z");
+    CmdArgs.push_back("max-page-size=0x4000");
+
     // Patch relocated regions of DWARF whose targets are eliminated at link
     // time with specific tombstones, such that they're recognisable by the
     // PlayStation debugger.
@@ -295,6 +302,11 @@ void tools::PS5cpu::Linker::ConstructJob(Compilation &C, const JobAction &JA,
   if (Shared)
     CmdArgs.push_back("--shared");
 
+  // Provide a base address for non-PIE executables. This includes cases where
+  // -static is supplied without -pie.
+  if (!Relocatable && !Shared && !PIE)
+    CmdArgs.push_back("--image-base=0x400000");
+
   assert((Output.isFilename() || Output.isNothing()) && "Invalid output.");
   if (Output.isFilename()) {
     CmdArgs.push_back("-o");

clang/lib/Sema/CheckExprLifetime.cpp

@@ -1261,12 +1261,12 @@ static void checkExprLifetimeImpl(Sema &SemaRef,
         if (pathContainsInit(Path))
           return false;
 
+        auto *DRE = dyn_cast<DeclRefExpr>(L);
         // Suppress false positives for code like the one below:
-        //   Ctor(unique_ptr<T> up) : member(*up), member2(move(up)) {}
-        if (IsLocalGslOwner && pathOnlyHandlesGslPointer(Path))
+        //   Ctor(unique_ptr<T> up) : pointer(up.get()), owner(move(up)) {}
+        if (DRE && isRecordWithAttr<OwnerAttr>(DRE->getType()))
           return false;
 
-        auto *DRE = dyn_cast<DeclRefExpr>(L);
         auto *VD = DRE ? dyn_cast<VarDecl>(DRE->getDecl()) : nullptr;
         if (!VD) {
           // A member was initialized to a local block.