Merge llvm/main into amd-debug

Author: SC llvm team

clang/lib/AST/ByteCode/Descriptor.cpp

@@ -246,6 +246,25 @@ static void dtorRecord(Block *B, std::byte *Ptr, const Descriptor *D) {
     destroyBase(B, Ptr, F.Desc, F.Offset);
 }
 
+/// Whether a record needs its descriptor dtor function called.
+static bool needsRecordDtor(const Record *R) {
+  for (const auto &B : R->bases()) {
+    if (B.Desc->DtorFn)
+      return true;
+  }
+
+  for (const auto &F : R->fields()) {
+    if (F.Desc->DtorFn)
+      return true;
+  }
+
+  for (const auto &V : R->virtual_bases()) {
+    if (V.Desc->DtorFn)
+      return true;
+  }
+  return false;
+}
+
 static BlockCtorFn getCtorPrim(PrimType Type) {
   // Floating types are special. They are primitives, but need their
   // constructor called.
@@ -336,7 +355,7 @@ Descriptor::Descriptor(const DeclTy &D, const Type *SourceTy,
       AllocSize(std::max<size_t>(alignof(void *), Size) + MDSize),
       ElemDesc(Elem), IsConst(IsConst), IsMutable(IsMutable),
       IsTemporary(IsTemporary), IsArray(true), CtorFn(ctorArrayDesc),
-      DtorFn(dtorArrayDesc) {
+      DtorFn(Elem->DtorFn ? dtorArrayDesc : nullptr) {
   assert(Source && "Missing source");
 }
 
@@ -347,7 +366,7 @@ Descriptor::Descriptor(const DeclTy &D, const Descriptor *Elem, MetadataSize MD,
       Size(UnknownSizeMark), MDSize(MD.value_or(0)),
       AllocSize(MDSize + alignof(void *)), ElemDesc(Elem), IsConst(true),
       IsMutable(false), IsTemporary(IsTemporary), IsArray(true),
-      CtorFn(ctorArrayDesc), DtorFn(dtorArrayDesc) {
+      CtorFn(ctorArrayDesc), DtorFn(Elem->DtorFn ? dtorArrayDesc : nullptr) {
   assert(Source && "Missing source");
 }
 
@@ -359,7 +378,7 @@ Descriptor::Descriptor(const DeclTy &D, const Record *R, MetadataSize MD,
       Size(ElemSize), MDSize(MD.value_or(0)), AllocSize(Size + MDSize),
       ElemRecord(R), IsConst(IsConst), IsMutable(IsMutable),
       IsTemporary(IsTemporary), IsVolatile(IsVolatile), CtorFn(ctorRecord),
-      DtorFn(dtorRecord) {
+      DtorFn(needsRecordDtor(R) ? dtorRecord : nullptr) {
   assert(Source && "Missing source");
 }
 

clang/lib/AST/ByteCode/Integral.h

@@ -318,6 +318,11 @@ template <unsigned Bits, bool Signed> class Integral final {
   template <typename T> static bool CheckMulUB(T A, T B, T &R) {
     if constexpr (std::is_signed_v<T>) {
       return llvm::MulOverflow<T>(A, B, R);
+    } else if constexpr (sizeof(T) < sizeof(int)) {
+      // Silly integer promotion rules will convert both A and B to int,
+      // even it T is unsigned. Prevent that by manually casting to uint first.
+      R = static_cast<T>(static_cast<unsigned>(A) * static_cast<unsigned>(B));
+      return false;
     } else {
       R = A * B;
       return false;

clang/lib/AST/ByteCode/Interp.cpp

@@ -1207,17 +1207,15 @@ static bool runRecordDestructor(InterpState &S, CodePtr OpPC,
   }
 
   // Destructor of this record.
-  if (const CXXDestructorDecl *Dtor = R->getDestructor();
-      Dtor && !Dtor->isTrivial()) {
-    const Function *DtorFunc = S.getContext().getOrCreateFunction(Dtor);
-    if (!DtorFunc)
-      return false;
+  const CXXDestructorDecl *Dtor = R->getDestructor();
+  assert(Dtor);
+  assert(!Dtor->isTrivial());
+  const Function *DtorFunc = S.getContext().getOrCreateFunction(Dtor);
+  if (!DtorFunc)
+    return false;
 
-    S.Stk.push<Pointer>(BasePtr);
-    if (!Call(S, OpPC, DtorFunc, 0))
-      return false;
-  }
-  return true;
+  S.Stk.push<Pointer>(BasePtr);
+  return Call(S, OpPC, DtorFunc, 0);
 }
 
 static bool RunDestructors(InterpState &S, CodePtr OpPC, const Block *B) {
@@ -1229,6 +1227,9 @@ static bool RunDestructors(InterpState &S, CodePtr OpPC, const Block *B) {
 
   assert(Desc->isRecord() || Desc->isCompositeArray());
 
+  if (Desc->hasTrivialDtor())
+    return true;
+
   if (Desc->isCompositeArray()) {
     unsigned N = Desc->getNumElems();
     if (N == 0)

clang/lib/AST/ByteCode/InterpBuiltin.cpp

@@ -2514,9 +2514,9 @@ static bool interp__builtin_is_within_lifetime(InterpState &S, CodePtr OpPC,
   return true;
 }
 
-static bool interp__builtin_elementwise_sat(InterpState &S, CodePtr OpPC,
-                                            const CallExpr *Call,
-                                            unsigned BuiltinID) {
+static bool interp__builtin_elementwise_int_binop(InterpState &S, CodePtr OpPC,
+                                                  const CallExpr *Call,
+                                                  unsigned BuiltinID) {
   assert(Call->getNumArgs() == 2);
 
   // Single integer case.
@@ -2553,6 +2553,8 @@ static bool interp__builtin_elementwise_sat(InterpState &S, CodePtr OpPC,
   const Pointer &LHS = S.Stk.pop<Pointer>();
   const Pointer &Dst = S.Stk.peek<Pointer>();
   PrimType ElemT = *S.getContext().classify(VT->getElementType());
+  bool DestUnsigned =
+      VT->getElementType()->isUnsignedIntegerOrEnumerationType();
   unsigned NumElems = VT->getNumElements();
   for (unsigned I = 0; I != NumElems; ++I) {
     APSInt Elem1;
@@ -2586,6 +2588,34 @@ static bool interp__builtin_elementwise_sat(InterpState &S, CodePtr OpPC,
       Result = APSInt(llvm::APIntOps::mulhs(Elem1, Elem2),
                       /*isUnsigned=*/false);
       break;
+    case clang::X86::BI__builtin_ia32_psllv2di:
+    case clang::X86::BI__builtin_ia32_psllv4di:
+    case clang::X86::BI__builtin_ia32_psllv4si:
+    case clang::X86::BI__builtin_ia32_psllv8si:
+      if (Elem2.uge(Elem2.getBitWidth())) {
+        Result = APSInt(APInt::getZero(Elem2.getBitWidth()), DestUnsigned);
+        break;
+      }
+      Result = APSInt(Elem1.shl(Elem2.getZExtValue()), DestUnsigned);
+      break;
+    case clang::X86::BI__builtin_ia32_psrav4si:
+    case clang::X86::BI__builtin_ia32_psrav8si:
+      if (Elem2.uge(Elem2.getBitWidth())) {
+        Result = APSInt(Elem1.ashr(Elem2.getBitWidth() - 1), DestUnsigned);
+        break;
+      }
+      Result = APSInt(Elem1.ashr(Elem2.getZExtValue()), DestUnsigned);
+      break;
+    case clang::X86::BI__builtin_ia32_psrlv2di:
+    case clang::X86::BI__builtin_ia32_psrlv4di:
+    case clang::X86::BI__builtin_ia32_psrlv4si:
+    case clang::X86::BI__builtin_ia32_psrlv8si:
+      if (Elem2.uge(Elem2.getBitWidth())) {
+        Result = APSInt(APInt::getZero(Elem2.getBitWidth()), DestUnsigned);
+        break;
+      }
+      Result = APSInt(Elem1.lshr(Elem2.getZExtValue()), DestUnsigned);
+      break;
     default:
       llvm_unreachable("Wrong builtin ID");
     }
@@ -3232,7 +3262,17 @@ bool InterpretBuiltin(InterpState &S, CodePtr OpPC, const CallExpr *Call,
   case clang::X86::BI__builtin_ia32_pmulhw128:
   case clang::X86::BI__builtin_ia32_pmulhw256:
   case clang::X86::BI__builtin_ia32_pmulhw512:
-    return interp__builtin_elementwise_sat(S, OpPC, Call, BuiltinID);
+  case clang::X86::BI__builtin_ia32_psllv2di:
+  case clang::X86::BI__builtin_ia32_psllv4di:
+  case clang::X86::BI__builtin_ia32_psllv4si:
+  case clang::X86::BI__builtin_ia32_psllv8si:
+  case clang::X86::BI__builtin_ia32_psrav4si:
+  case clang::X86::BI__builtin_ia32_psrav8si:
+  case clang::X86::BI__builtin_ia32_psrlv2di:
+  case clang::X86::BI__builtin_ia32_psrlv4di:
+  case clang::X86::BI__builtin_ia32_psrlv4si:
+  case clang::X86::BI__builtin_ia32_psrlv8si:
+    return interp__builtin_elementwise_int_binop(S, OpPC, Call, BuiltinID);
 
   case Builtin::BI__builtin_elementwise_max:
   case Builtin::BI__builtin_elementwise_min:

clang/lib/AST/ByteCode/Program.cpp

@@ -101,7 +101,7 @@ unsigned Program::createGlobalString(const StringLiteral *S, const Expr *Base) {
       }
     }
   }
-  Ptr.initialize();
+  Ptr.initializeAllElements();
 
   return GlobalIndex;
 }

clang/test/CodeGen/X86/avx2-builtins.c

@@ -7,6 +7,14 @@
 // RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -ffreestanding %s -triple=i386-apple-darwin -target-feature +avx2 -emit-llvm -o - -Wall -Werror | FileCheck %s --check-prefixes=CHECK,X86
 // RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -ffreestanding %s -triple=i386-apple-darwin -target-feature +avx2 -fno-signed-char -emit-llvm -o - -Wall -Werror | FileCheck %s --check-prefixes=CHECK,X86
 
+// RUN: %clang_cc1 -x c -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +avx2 -emit-llvm -o - -Wall -Werror -fexperimental-new-constant-interpreter | FileCheck %s --check-prefixes=CHECK,X64
+// RUN: %clang_cc1 -x c -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +avx2 -fno-signed-char -emit-llvm -o - -Wall -Werror -fexperimental-new-constant-interpreter | FileCheck %s --check-prefixes=CHECK,X64
+// RUN: %clang_cc1 -x c -flax-vector-conversions=none -ffreestanding %s -triple=i386-apple-darwin -target-feature +avx2 -emit-llvm -o - -Wall -Werror -fexperimental-new-constant-interpreter | FileCheck %s --check-prefixes=CHECK,X86
+// RUN: %clang_cc1 -x c -flax-vector-conversions=none -ffreestanding %s -triple=i386-apple-darwin -target-feature +avx2 -fno-signed-char -emit-llvm -o - -Wall -Werror -fexperimental-new-constant-interpreter | FileCheck %s --check-prefixes=CHECK,X86
+// RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +avx2 -emit-llvm -o - -Wall -Werror -fexperimental-new-constant-interpreter | FileCheck %s --check-prefixes=CHECK,X64
+// RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +avx2 -fno-signed-char -emit-llvm -o - -Wall -Werror -fexperimental-new-constant-interpreter | FileCheck %s --check-prefixes=CHECK,X64
+// RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -ffreestanding %s -triple=i386-apple-darwin -target-feature +avx2 -emit-llvm -o - -Wall -Werror -fexperimental-new-constant-interpreter | FileCheck %s --check-prefixes=CHECK,X86
+// RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -ffreestanding %s -triple=i386-apple-darwin -target-feature +avx2 -fno-signed-char -emit-llvm -o - -Wall -Werror -fexperimental-new-constant-interpreter | FileCheck %s --check-prefixes=CHECK,X86
 
 #include <immintrin.h>
 #include "builtin_test_helpers.h"

flang/include/flang/Parser/parse-tree.h

@@ -4893,8 +4893,11 @@ struct OpenMPSectionsConstruct {
   CharBlock source;
   // Each of the OpenMPConstructs in the list below contains an
   // OpenMPSectionConstruct. This is guaranteed by the parser.
+  // The end sections directive is optional here because it is difficult to
+  // generate helpful error messages for a missing end directive within the
+  // parser. Semantics will generate an error if this is absent.
   std::tuple<OmpBeginSectionsDirective, std::list<OpenMPConstruct>,
-      OmpEndSectionsDirective>
+      std::optional<OmpEndSectionsDirective>>
       t;
 };
 

flang/lib/Lower/OpenMP/OpenMP.cpp

@@ -3979,9 +3979,12 @@ static void genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,
   List<Clause> clauses = makeClauses(
       std::get<parser::OmpClauseList>(beginSectionsDirective.t), semaCtx);
   const auto &endSectionsDirective =
-      std::get<parser::OmpEndSectionsDirective>(sectionsConstruct.t);
+      std::get<std::optional<parser::OmpEndSectionsDirective>>(
+          sectionsConstruct.t);
+  assert(endSectionsDirective &&
+         "Missing end section directive should have been handled in semantics");
   clauses.append(makeClauses(
-      std::get<parser::OmpClauseList>(endSectionsDirective.t), semaCtx));
+      std::get<parser::OmpClauseList>(endSectionsDirective->t), semaCtx));
   mlir::Location currentLocation = converter.getCurrentLocation();
 
   llvm::omp::Directive directive =

flang/lib/Parser/openmp-parsers.cpp

@@ -1921,7 +1921,7 @@ TYPE_PARSER(sourced(construct<OpenMPSectionsConstruct>(
             construct<OpenMPSectionConstruct>(maybe(sectionDir), block))),
         many(construct<OpenMPConstruct>(
             sourced(construct<OpenMPSectionConstruct>(sectionDir, block))))),
-    Parser<OmpEndSectionsDirective>{} / endOmpLine)))
+    maybe(Parser<OmpEndSectionsDirective>{} / endOmpLine))))
 
 static bool IsExecutionPart(const OmpDirectiveName &name) {
   return name.IsExecutionPart();

flang/lib/Parser/unparse.cpp

@@ -2788,7 +2788,7 @@ class UnparseVisitor {
     Walk(std::get<std::list<OpenMPConstruct>>(x.t), "");
     BeginOpenMP();
     Word("!$OMP END ");
-    Walk(std::get<OmpEndSectionsDirective>(x.t));
+    Walk(std::get<std::optional<OmpEndSectionsDirective>>(x.t));
     Put("\n");
     EndOpenMP();
   }