Merged main:362b1157868c into amd-gfx:994f45d31656

Local branch amd-gfx 994f45d Merged main:781424c87273 into amd-gfx:b4f6afcfc5d9
Remote branch main 362b115 [flang][openacc] Avoid privatizing symbols during semantics (llvm#69506)

.github/workflows/release-tasks.yml

@@ -104,16 +104,18 @@ jobs:
         run: |
           cd llvm/utils/lit
           # Remove 'dev' suffix from lit version.
-          sed -i "s/ + 'dev'//g" lit/__init__.py
+          sed -i 's/ + "dev"//g' lit/__init__.py
           python3 setup.py sdist
 
       - name: Upload lit to test.pypi.org
         uses: pypa/gh-action-pypi-publish@release/v1
         with:
           password: ${{ secrets.LLVM_LIT_TEST_PYPI_API_TOKEN }}
           repository-url: https://test.pypi.org/legacy/
+          packages-dir: llvm/utils/lit/dist/
 
       - name: Upload lit to pypi.org
         uses: pypa/gh-action-pypi-publish@release/v1
         with:
           password: ${{ secrets.LLVM_LIT_PYPI_API_TOKEN }}
+          packages-dir: llvm/utils/lit/dist/

clang/lib/Format/TokenAnnotator.cpp

@@ -3497,6 +3497,14 @@ void TokenAnnotator::calculateFormattingInformation(AnnotatedLine &Line) const {
           Tok->setType(TT_TrailingReturnArrow);
           break;
         }
+        if (Tok->isNot(TT_TrailingAnnotation))
+          continue;
+        const auto *Next = Tok->Next;
+        if (!Next || Next->isNot(tok::l_paren))
+          continue;
+        Tok = Next->MatchingParen;
+        if (!Tok)
+          break;
       }
     }
   }

clang/unittests/Format/TokenAnnotatorTest.cpp

@@ -1826,6 +1826,10 @@ TEST_F(TokenAnnotatorTest, UnderstandsTrailingReturnArrow) {
   ASSERT_EQ(Tokens.size(), 15u) << Tokens;
   EXPECT_TOKEN(Tokens[12], tok::arrow, TT_Unknown);
 
+  Tokens = annotate("void f() FOO(foo->bar);");
+  ASSERT_EQ(Tokens.size(), 12u) << Tokens;
+  EXPECT_TOKEN(Tokens[7], tok::arrow, TT_Unknown);
+
   // Mixed
   Tokens = annotate("auto f() -> int { auto a = b()->c; }");
   ASSERT_EQ(Tokens.size(), 18u) << Tokens;

compiler-rt/lib/asan/asan_interceptors.cpp

@@ -251,7 +251,7 @@ INTERCEPTOR(int, pthread_create, void *thread, void *attr,
 
   u32 current_tid = GetCurrentTidOrInvalid();
 
-  __sanitizer_sigset_t sigset;
+  __sanitizer_sigset_t sigset = {};
 #    if SANITIZER_FREEBSD || SANITIZER_LINUX || SANITIZER_NETBSD || \
         SANITIZER_SOLARIS
   ScopedBlockSignals block(&sigset);

flang/lib/Semantics/resolve-directives.cpp

@@ -254,9 +254,6 @@ class AccAttributeVisitor : DirectiveAttributeVisitor<llvm::acc::Directive> {
       Symbol::Flag::AccCopyIn, Symbol::Flag::AccCopyOut,
       Symbol::Flag::AccDelete, Symbol::Flag::AccPresent};
 
-  Symbol::Flags accFlagsRequireNewSymbol{Symbol::Flag::AccPrivate,
-      Symbol::Flag::AccFirstPrivate, Symbol::Flag::AccReduction};
-
   Symbol::Flags accDataMvtFlags{
       Symbol::Flag::AccDevice, Symbol::Flag::AccHost, Symbol::Flag::AccSelf};
 
@@ -266,7 +263,7 @@ class AccAttributeVisitor : DirectiveAttributeVisitor<llvm::acc::Directive> {
       Symbol::Flag::AccDevicePtr, Symbol::Flag::AccDeviceResident,
       Symbol::Flag::AccLink, Symbol::Flag::AccPresent};
 
-  void PrivatizeAssociatedLoopIndex(const parser::OpenACCLoopConstruct &);
+  void CheckAssociatedLoopIndex(const parser::OpenACCLoopConstruct &);
   void ResolveAccObjectList(const parser::AccObjectList &, Symbol::Flag);
   void ResolveAccObject(const parser::AccObject &, Symbol::Flag);
   Symbol *ResolveAcc(const parser::Name &, Symbol::Flag, Scope &);
@@ -877,7 +874,7 @@ bool AccAttributeVisitor::Pre(const parser::OpenACCLoopConstruct &x) {
   }
   ClearDataSharingAttributeObjects();
   SetContextAssociatedLoopLevel(GetAssociatedLoopLevelFromClauses(clauseList));
-  PrivatizeAssociatedLoopIndex(x);
+  CheckAssociatedLoopIndex(x);
   return true;
 }
 
@@ -1141,13 +1138,12 @@ std::int64_t AccAttributeVisitor::GetAssociatedLoopLevelFromClauses(
   return 1; // default is outermost loop
 }
 
-void AccAttributeVisitor::PrivatizeAssociatedLoopIndex(
+void AccAttributeVisitor::CheckAssociatedLoopIndex(
     const parser::OpenACCLoopConstruct &x) {
   std::int64_t level{GetContext().associatedLoopLevel};
-  if (level <= 0) { // collpase value was negative or 0
+  if (level <= 0) { // collapse value was negative or 0
     return;
   }
-  Symbol::Flag ivDSA{Symbol::Flag::AccPrivate};
 
   const auto getNextDoConstruct =
       [this](const parser::Block &block) -> const parser::DoConstruct * {
@@ -1166,16 +1162,8 @@ void AccAttributeVisitor::PrivatizeAssociatedLoopIndex(
 
   const auto &outer{std::get<std::optional<parser::DoConstruct>>(x.t)};
   for (const parser::DoConstruct *loop{&*outer}; loop && level > 0; --level) {
-    // go through all the nested do-loops and resolve index variables
-    const parser::Name *iv{GetLoopIndex(*loop)};
-    if (iv) {
-      if (auto *symbol{ResolveAcc(*iv, ivDSA, currScope())}) {
-        symbol->set(Symbol::Flag::AccPreDetermined);
-        iv->symbol = symbol; // adjust the symbol within region
-        AddToContextObjectWithDSA(*symbol, ivDSA);
-      }
-    }
-
+    // Go through all nested loops to ensure index variable exists.
+    GetLoopIndex(*loop);
     const auto &block{std::get<parser::Block>(loop->t)};
     loop = getNextDoConstruct(block);
   }
@@ -1328,20 +1316,12 @@ void AccAttributeVisitor::ResolveAccObject(
 
 Symbol *AccAttributeVisitor::ResolveAcc(
     const parser::Name &name, Symbol::Flag accFlag, Scope &scope) {
-  if (accFlagsRequireNewSymbol.test(accFlag)) {
-    return DeclarePrivateAccessEntity(name, accFlag, scope);
-  } else {
-    return DeclareOrMarkOtherAccessEntity(name, accFlag);
-  }
+  return DeclareOrMarkOtherAccessEntity(name, accFlag);
 }
 
 Symbol *AccAttributeVisitor::ResolveAcc(
     Symbol &symbol, Symbol::Flag accFlag, Scope &scope) {
-  if (accFlagsRequireNewSymbol.test(accFlag)) {
-    return DeclarePrivateAccessEntity(symbol, accFlag, scope);
-  } else {
-    return DeclareOrMarkOtherAccessEntity(symbol, accFlag);
-  }
+  return DeclareOrMarkOtherAccessEntity(symbol, accFlag);
 }
 
 Symbol *AccAttributeVisitor::DeclareOrMarkOtherAccessEntity(
@@ -1374,11 +1354,6 @@ static bool WithMultipleAppearancesAccException(
 void AccAttributeVisitor::CheckMultipleAppearances(
     const parser::Name &name, const Symbol &symbol, Symbol::Flag accFlag) {
   const auto *target{&symbol};
-  if (accFlagsRequireNewSymbol.test(accFlag)) {
-    if (const auto *details{symbol.detailsIf<HostAssocDetails>()}) {
-      target = &details->symbol();
-    }
-  }
   if (HasDataSharingAttributeObject(*target) &&
       !WithMultipleAppearancesAccException(symbol, accFlag)) {
     context_.Say(name.source,

flang/test/Semantics/OpenACC/acc-symbols01.f90

@@ -14,11 +14,11 @@ program mm
   b = 2
  !$acc parallel present(c) firstprivate(b) private(a)
  !$acc loop
-  !DEF: /mm/OtherConstruct1/i (AccPrivate, AccPreDetermined) HostAssoc INTEGER(4)
+  !REF: /mm/i
   do i=1,10
-   !DEF: /mm/OtherConstruct1/a (AccPrivate) HostAssoc INTEGER(4)
-   !REF: /mm/OtherConstruct1/i
-   !DEF: /mm/OtherConstruct1/b (AccFirstPrivate) HostAssoc INTEGER(4)
+   !REF: /mm/a
+   !REF: /mm/i
+   !REF: /mm/b
    a(i) = b(i)
   end do
  !$acc end parallel

llvm/examples/Kaleidoscope/Chapter4/CMakeLists.txt

@@ -5,6 +5,7 @@ set(LLVM_LINK_COMPONENTS
   InstCombine
   Object
   OrcJIT
+  Passes
   RuntimeDyld
   ScalarOpts
   Support

llvm/examples/Kaleidoscope/Chapter5/CMakeLists.txt

@@ -5,6 +5,7 @@ set(LLVM_LINK_COMPONENTS
   InstCombine
   Object
   OrcJIT
+  Passes
   RuntimeDyld
   ScalarOpts
   Support

llvm/examples/Kaleidoscope/Chapter6/CMakeLists.txt

@@ -5,6 +5,7 @@ set(LLVM_LINK_COMPONENTS
   InstCombine
   Object
   OrcJIT
+  Passes
   RuntimeDyld
   ScalarOpts
   Support

llvm/examples/Kaleidoscope/Chapter7/CMakeLists.txt

@@ -5,6 +5,7 @@ set(LLVM_LINK_COMPONENTS
   InstCombine
   Object
   OrcJIT
+  Passes
   RuntimeDyld
   ScalarOpts
   Support

llvm/include/llvm/Config/llvm-config.h.cmake

@@ -16,7 +16,7 @@
 
 /* Indicate that this is LLVM compiled from the amd-gfx branch. */
 #define LLVM_HAVE_BRANCH_AMD_GFX
-#define LLVM_MAIN_REVISION 477915
+#define LLVM_MAIN_REVISION 477929
 
 /* Define if LLVM_ENABLE_DUMP is enabled */
 #cmakedefine LLVM_ENABLE_DUMP

llvm/lib/Analysis/InlineOrder.cpp

@@ -218,10 +218,10 @@ class PriorityInlineOrder : public InlineOrder<std::pair<CallBase *, int>> {
   // A call site could become less desirable for inlining because of the size
   // growth from prior inlining into the callee. This method is used to lazily
   // update the desirability of a call site if it's decreasing. It is only
-  // called on pop() or front(), not every time the desirability changes. When
-  // the desirability of the front call site decreases, an updated one would be
-  // pushed right back into the heap. For simplicity, those cases where
-  // the desirability of a call site increases are ignored here.
+  // called on pop(), not every time the desirability changes. When the
+  // desirability of the front call site decreases, an updated one would be
+  // pushed right back into the heap. For simplicity, those cases where the
+  // desirability of a call site increases are ignored here.
   void adjust() {
     std::pop_heap(Heap.begin(), Heap.end(), isLess);
     while (updateAndCheckDecreased(Heap.back())) {
@@ -255,11 +255,10 @@ class PriorityInlineOrder : public InlineOrder<std::pair<CallBase *, int>> {
     assert(size() > 0);
     adjust();
 
-    CallBase *CB = Heap.front();
+    std::pop_heap(Heap.begin(), Heap.end(), isLess);
+    CallBase *CB = Heap.pop_back_val();
     T Result = std::make_pair(CB, InlineHistoryMap[CB]);
     InlineHistoryMap.erase(CB);
-    std::pop_heap(Heap.begin(), Heap.end(), isLess);
-    Heap.pop_back();
     return Result;
   }
 
@@ -319,4 +318,4 @@ llvm::getInlineOrder(FunctionAnalysisManager &FAM, const InlineParams &Params,
                                                                M);
   }
   return getDefaultInlineOrder(FAM, Params, MAM, M);
-}
+}

llvm/lib/DWARFLinker/DWARFLinker.cpp

@@ -2682,12 +2682,12 @@ Error DWARFLinker::link() {
       continue;
     }
 
-    // In a first phase, just read in the debug info and load all clang modules.
+    // Clone all the clang modules with requires extracting the DIE units. We
+    // don't need the full debug info until the Analyze phase.
     OptContext.CompileUnits.reserve(
         OptContext.File.Dwarf->getNumCompileUnits());
-
     for (const auto &CU : OptContext.File.Dwarf->compile_units()) {
-      auto CUDie = CU->getUnitDIE(false);
+      auto CUDie = CU->getUnitDIE(/*ExtractUnitDIEOnly=*/true);
       if (Options.Verbose) {
         outs() << "Input compilation unit:";
         DIDumpOptions DumpOpts;
@@ -2728,9 +2728,9 @@ Error DWARFLinker::link() {
       return;
 
     for (const auto &CU : Context.File.Dwarf->compile_units()) {
-      // The !isClangModuleRef condition effectively skips over fully resolved
-      // skeleton units.
-      auto CUDie = CU->getUnitDIE();
+      // Previously we only extracted the unit DIEs. We need the full debug info
+      // now.
+      auto CUDie = CU->getUnitDIE(/*ExtractUnitDIEOnly=*/false);
       std::string PCMFile = getPCMFile(CUDie, Options.ObjectPrefixMap);
 
       if (!CUDie || LLVM_UNLIKELY(Options.Update) ||

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

@@ -3128,6 +3128,7 @@ bool SIInstrInfo::isFoldableCopy(const MachineInstr &MI) {
   case AMDGPU::V_MOV_B64_e64:
   case AMDGPU::S_MOV_B32:
   case AMDGPU::S_MOV_B64:
+  case AMDGPU::S_MOV_B64_IMM_PSEUDO:
   case AMDGPU::COPY:
   case AMDGPU::WWM_COPY:
   case AMDGPU::V_ACCVGPR_WRITE_B32_e64:

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

@@ -13744,6 +13744,10 @@ static SDValue performBUILD_VECTORCombine(SDNode *N, SelectionDAG &DAG,
   SmallVector<SDValue> RHSOps;
   for (SDValue Op : N->ops()) {
     if (Op.isUndef()) {
+      // We can't form a divide or remainder from undef.
+      if (!DAG.isSafeToSpeculativelyExecute(Opcode))
+        return SDValue();
+
       LHSOps.push_back(Op);
       RHSOps.push_back(Op);
       continue;

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

@@ -553,8 +553,8 @@ static std::optional<unsigned> getExtractIndex(Instruction *E) {
 /// successful, the matched scalars are replaced by poison values in \p VL for
 /// future analysis.
 static std::optional<TTI::ShuffleKind>
-tryToGatherExtractElements(SmallVectorImpl<Value *> &VL,
-                           SmallVectorImpl<int> &Mask) {
+tryToGatherSingleRegisterExtractElements(MutableArrayRef<Value *> VL,
+                                         SmallVectorImpl<int> &Mask) {
   // Scan list of gathered scalars for extractelements that can be represented
   // as shuffles.
   MapVector<Value *, SmallVector<int>> VectorOpToIdx;
@@ -641,7 +641,7 @@ tryToGatherExtractElements(SmallVectorImpl<Value *> &VL,
   if (!Res) {
     // TODO: try to check other subsets if possible.
     // Restore the original VL if attempt was not successful.
-    VL.swap(SavedVL);
+    copy(SavedVL, VL.begin());
     return std::nullopt;
   }
   // Restore unused scalars from mask, if some of the extractelements were not
@@ -7616,7 +7616,7 @@ BoUpSLP::getEntryCost(const TreeEntry *E, ArrayRef<Value *> VectorizedVals,
     std::optional<TargetTransformInfo::ShuffleKind> GatherShuffle;
     SmallVector<const TreeEntry *> Entries;
     // Check for gathered extracts.
-    ExtractShuffle = tryToGatherExtractElements(GatheredScalars, ExtractMask);
+    ExtractShuffle = tryToGatherSingleRegisterExtractElements(GatheredScalars, ExtractMask);
     SmallVector<Value *> IgnoredVals;
     if (UserIgnoreList)
       IgnoredVals.assign(UserIgnoreList->begin(), UserIgnoreList->end());
@@ -10166,7 +10166,8 @@ ResTy BoUpSLP::processBuildVector(const TreeEntry *E, Args &...Params) {
   Type *ScalarTy = GatheredScalars.front()->getType();
   if (!all_of(GatheredScalars, UndefValue::classof)) {
     // Check for gathered extracts.
-    ExtractShuffle = tryToGatherExtractElements(GatheredScalars, ExtractMask);
+    ExtractShuffle =
+        tryToGatherSingleRegisterExtractElements(GatheredScalars, ExtractMask);
     SmallVector<Value *> IgnoredVals;
     if (UserIgnoreList)
       IgnoredVals.assign(UserIgnoreList->begin(), UserIgnoreList->end());

llvm/test/CodeGen/AMDGPU/fold-short-64-bit-literals.mir

@@ -9,11 +9,11 @@ body:             |
 
     ; GCN-LABEL: name: no_fold_fp_64bit_literal_sgpr
     ; GCN: [[DEF:%[0-9]+]]:vreg_64 = IMPLICIT_DEF
-    ; GCN-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64 = S_MOV_B64 1311768467750121200
-    ; GCN-NEXT: [[V_ADD_F64_e64_:%[0-9]+]]:vreg_64 = V_ADD_F64_e64 0, [[S_MOV_B64_]], 0, [[DEF]], 0, 0, implicit $mode, implicit $exec
+    ; GCN-NEXT: [[S_MOV_B:%[0-9]+]]:sreg_64 = S_MOV_B64_IMM_PSEUDO 1311768467750121200
+    ; GCN-NEXT: [[V_ADD_F64_e64_:%[0-9]+]]:vreg_64 = V_ADD_F64_e64 0, [[S_MOV_B]], 0, [[DEF]], 0, 0, implicit $mode, implicit $exec
     ; GCN-NEXT: SI_RETURN_TO_EPILOG [[V_ADD_F64_e64_]]
     %0:vreg_64 = IMPLICIT_DEF
-    %1:sreg_64 = S_MOV_B64 1311768467750121200
+    %1:sreg_64 = S_MOV_B64_IMM_PSEUDO 1311768467750121200
     %2:vreg_64 = V_ADD_F64_e64 0, %1, 0, %0, 0, 0, implicit $mode, implicit $exec
     SI_RETURN_TO_EPILOG %2
 ...
@@ -46,7 +46,7 @@ body:             |
     ; GCN-NEXT: [[V_ADD_F64_e64_:%[0-9]+]]:vreg_64 = V_ADD_F64_e64 0, 4636737291354636288, 0, [[DEF]], 0, 0, implicit $mode, implicit $exec
     ; GCN-NEXT: SI_RETURN_TO_EPILOG [[V_ADD_F64_e64_]]
     %0:vreg_64 = IMPLICIT_DEF
-    %1:sreg_64 = S_MOV_B64 4636737291354636288
+    %1:sreg_64 = S_MOV_B64_IMM_PSEUDO 4636737291354636288
     %2:vreg_64 = V_ADD_F64_e64 0, %1, 0, %0, 0, 0, implicit $mode, implicit $exec
     SI_RETURN_TO_EPILOG %2
 ...
@@ -59,11 +59,11 @@ body:             |
 
     ; GCN-LABEL: name: no_fold_int_64bit_literal_sgpr
     ; GCN: [[DEF:%[0-9]+]]:sreg_64 = IMPLICIT_DEF
-    ; GCN-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64 = S_MOV_B64 1311768467750121200
-    ; GCN-NEXT: [[S_AND_B64_:%[0-9]+]]:sreg_64 = S_AND_B64 [[DEF]], [[S_MOV_B64_]], implicit-def $scc
+    ; GCN-NEXT: [[S_MOV_B:%[0-9]+]]:sreg_64 = S_MOV_B64_IMM_PSEUDO 1311768467750121200
+    ; GCN-NEXT: [[S_AND_B64_:%[0-9]+]]:sreg_64 = S_AND_B64 [[DEF]], [[S_MOV_B]], implicit-def $scc
     ; GCN-NEXT: SI_RETURN_TO_EPILOG [[S_AND_B64_]]
     %0:sreg_64 = IMPLICIT_DEF
-    %1:sreg_64 = S_MOV_B64 1311768467750121200
+    %1:sreg_64 = S_MOV_B64_IMM_PSEUDO 1311768467750121200
     %2:sreg_64 = S_AND_B64 %0, %1, implicit-def $scc
     SI_RETURN_TO_EPILOG %2
 ...