Merge remote-tracking branch 'origin/main' into pr/zext-knownbits

Author: topperc

clang/lib/AST/ByteCode/DynamicAllocator.h

@@ -97,7 +97,7 @@ class DynamicAllocator final {
 private:
   llvm::DenseMap<const Expr *, AllocationSite> AllocationSites;
 
-  using PoolAllocTy = llvm::BumpPtrAllocatorImpl<llvm::MallocAllocator>;
+  using PoolAllocTy = llvm::BumpPtrAllocator;
   PoolAllocTy DescAllocator;
 
   /// Allocates a new descriptor.

clang/lib/AST/ByteCode/Program.h

@@ -171,7 +171,7 @@ class Program final {
   llvm::DenseMap<const void *, unsigned> NativePointerIndices;
 
   /// Custom allocator for global storage.
-  using PoolAllocTy = llvm::BumpPtrAllocatorImpl<llvm::MallocAllocator>;
+  using PoolAllocTy = llvm::BumpPtrAllocator;
 
   /// Descriptor + storage for a global object.
   ///

flang/include/flang/Common/Fortran.h

@@ -118,7 +118,8 @@ static constexpr IgnoreTKRSet ignoreTKRAll{IgnoreTKR::Type, IgnoreTKR::Kind,
 std::string AsFortran(IgnoreTKRSet);
 
 bool AreCompatibleCUDADataAttrs(std::optional<CUDADataAttr>,
-    std::optional<CUDADataAttr>, IgnoreTKRSet, bool allowUnifiedMatchingRule,
+    std::optional<CUDADataAttr>, IgnoreTKRSet, std::optional<std::string> *,
+    bool allowUnifiedMatchingRule,
     const LanguageFeatureControl *features = nullptr);
 
 static constexpr char blankCommonObjectName[] = "__BLNK__";

flang/lib/Common/Fortran.cpp

@@ -103,7 +103,8 @@ std::string AsFortran(IgnoreTKRSet tkr) {
 /// dummy argument attribute while `y` represents the actual argument attribute.
 bool AreCompatibleCUDADataAttrs(std::optional<CUDADataAttr> x,
     std::optional<CUDADataAttr> y, IgnoreTKRSet ignoreTKR,
-    bool allowUnifiedMatchingRule, const LanguageFeatureControl *features) {
+    std::optional<std::string> *warning, bool allowUnifiedMatchingRule,
+    const LanguageFeatureControl *features) {
   bool isCudaManaged{features
           ? features->IsEnabled(common::LanguageFeature::CudaManaged)
           : false};
@@ -134,8 +135,12 @@ bool AreCompatibleCUDADataAttrs(std::optional<CUDADataAttr> x,
     } else {
       if (*x == CUDADataAttr::Device) {
         if ((y &&
-                (*y == CUDADataAttr::Managed || *y == CUDADataAttr::Unified)) ||
+                (*y == CUDADataAttr::Managed || *y == CUDADataAttr::Unified ||
+                    *y == CUDADataAttr::Shared)) ||
             (!y && (isCudaUnified || isCudaManaged))) {
+          if (y && *y == CUDADataAttr::Shared && warning) {
+            *warning = "SHARED attribute ignored"s;
+          }
           return true;
         }
       } else if (*x == CUDADataAttr::Managed) {

flang/lib/Evaluate/characteristics.cpp

@@ -371,7 +371,7 @@ bool DummyDataObject::IsCompatibleWith(const DummyDataObject &actual,
   }
   if (!attrs.test(Attr::Value) &&
       !common::AreCompatibleCUDADataAttrs(cudaDataAttr, actual.cudaDataAttr,
-          ignoreTKR,
+          ignoreTKR, warning,
           /*allowUnifiedMatchingRule=*/false)) {
     if (whyNot) {
       *whyNot = "incompatible CUDA data attributes";
@@ -1771,7 +1771,7 @@ bool DistinguishUtils::Distinguishable(
       x.intent != common::Intent::In) {
     return true;
   } else if (!common::AreCompatibleCUDADataAttrs(x.cudaDataAttr, y.cudaDataAttr,
-                 x.ignoreTKR | y.ignoreTKR,
+                 x.ignoreTKR | y.ignoreTKR, nullptr,
                  /*allowUnifiedMatchingRule=*/false)) {
     return true;
   } else if (features_.IsEnabled(

flang/lib/Semantics/check-call.cpp

@@ -976,8 +976,9 @@ static void CheckExplicitDataArg(const characteristics::DummyDataObject &dummy,
         actualDataAttr = common::CUDADataAttr::Device;
       }
     }
+    std::optional<std::string> warning;
     if (!common::AreCompatibleCUDADataAttrs(dummyDataAttr, actualDataAttr,
-            dummy.ignoreTKR,
+            dummy.ignoreTKR, &warning,
             /*allowUnifiedMatchingRule=*/true, &context.languageFeatures())) {
       auto toStr{[](std::optional<common::CUDADataAttr> x) {
         return x ? "ATTRIBUTES("s +
@@ -988,6 +989,10 @@ static void CheckExplicitDataArg(const characteristics::DummyDataObject &dummy,
           "%s has %s but its associated actual argument has %s"_err_en_US,
           dummyName, toStr(dummyDataAttr), toStr(actualDataAttr));
     }
+    if (warning && context.ShouldWarn(common::UsageWarning::CUDAUsage)) {
+      messages.Say(common::UsageWarning::CUDAUsage, "%s"_warn_en_US,
+          std::move(*warning));
+    }
   }
 
   // Warning for breaking F'2023 change with character allocatables

flang/test/Semantics/cuf17.cuf

@@ -0,0 +1,18 @@
+! RUN: bbc -emit-hlfir -fcuda %s 2>&1 | FileCheck %s
+
+module mod1
+contains
+
+attributes(device) subroutine sub1(adev)
+  real, device :: adev(10)
+end
+
+attributes(global) subroutine sub2()
+  real, shared :: adev(10)
+  !WARNING: SHARED attribute ignored
+  call sub1(adev)
+end subroutine
+
+end module
+
+! CHECK: warning: SHARED attribute ignored

llvm/include/llvm/Target/GlobalISel/Combine.td

@@ -566,7 +566,8 @@ def binop_same_val: GICombineRule<
 // Fold (0 op x) - > 0
 def binop_left_to_zero: GICombineRule<
   (defs root:$root),
-  (match (wip_match_opcode G_SDIV, G_UDIV, G_SREM, G_UREM):$root,
+  (match (wip_match_opcode G_SHL, G_LSHR, G_ASHR, G_SDIV, G_UDIV, G_SREM,
+                           G_UREM):$root,
     [{ return Helper.matchOperandIsZero(*${root}, 1); }]),
   (apply [{ Helper.replaceSingleDefInstWithOperand(*${root}, 1); }])
 >;

llvm/lib/Analysis/ValueTracking.cpp

@@ -6713,11 +6713,11 @@ static bool isSameUnderlyingObjectInLoop(const PHINode *PN,
 }
 
 const Value *llvm::getUnderlyingObject(const Value *V, unsigned MaxLookup) {
-  if (!V->getType()->isPointerTy())
-    return V;
   for (unsigned Count = 0; MaxLookup == 0 || Count < MaxLookup; ++Count) {
     if (auto *GEP = dyn_cast<GEPOperator>(V)) {
       V = GEP->getPointerOperand();
+      if (!V->getType()->isPointerTy()) // Only handle scalar pointer base.
+        return nullptr;
     } else if (Operator::getOpcode(V) == Instruction::BitCast ||
                Operator::getOpcode(V) == Instruction::AddrSpaceCast) {
       Value *NewV = cast<Operator>(V)->getOperand(0);

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

@@ -1339,9 +1339,9 @@ RISCVTargetLowering::RISCVTargetLowering(const TargetMachine &TM,
                             ISD::VECTOR_SHUFFLE, ISD::VECTOR_COMPRESS},
                            VT, Custom);
 
-        // FIXME: mload, mstore, vp_gather/scatter can be
-        // hoisted to here.
-        setOperationAction({ISD::LOAD, ISD::STORE, ISD::MGATHER, ISD::MSCATTER},
+        // FIXME: vp_gather/scatter can be hoisted to here.
+        setOperationAction({ISD::LOAD, ISD::STORE, ISD::MLOAD, ISD::MSTORE,
+                            ISD::MGATHER, ISD::MSCATTER},
                            VT, Custom);
         setOperationAction({ISD::VP_LOAD, ISD::VP_STORE,
                             ISD::EXPERIMENTAL_VP_STRIDED_LOAD,
@@ -1409,8 +1409,6 @@ RISCVTargetLowering::RISCVTargetLowering(const TargetMachine &TM,
         setOperationAction({ISD::BUILD_VECTOR, ISD::SCALAR_TO_VECTOR}, VT,
                            Custom);
 
-        setOperationAction({ISD::MLOAD, ISD::MSTORE}, VT, Custom);
-
         setOperationAction({ISD::VP_GATHER, ISD::VP_SCATTER}, VT, Custom);
 
         setOperationAction({ISD::FADD, ISD::FSUB, ISD::FMUL, ISD::FDIV,