fix up function definition, add TiledResources shader flag

Author: bob80905

clang/lib/CodeGen/CGHLSLBuiltins.cpp

@@ -357,7 +357,7 @@ Value *CodeGenFunction::EmitHLSLBuiltinExpr(unsigned BuiltinID,
     Value *HandleOp = EmitScalarExpr(E->getArg(0));
     Value *IndexOp = EmitScalarExpr(E->getArg(1));
 
-    // Get the *address* of the status argument (since it's a reference)
+    // Get the *address* of the status argument to write to it by reference
     LValue StatusLVal = EmitLValue(E->getArg(2));
     Address StatusAddr = StatusLVal.getAddress();
 
@@ -370,11 +370,9 @@ Value *CodeGenFunction::EmitHLSLBuiltinExpr(unsigned BuiltinID,
                                ? llvm::Intrinsic::dx_resource_load_rawbuffer
                                : llvm::Intrinsic::dx_resource_load_typedbuffer;
 
-    QualType BuiltinRetTy = E->getType();
-    llvm::Type *DataTy = ConvertType(BuiltinRetTy->getPointeeType());
-
-    llvm::Type *IntrinsicRetTy = llvm::StructType::get(
-        Builder.getContext(), {DataTy, Builder.getInt1Ty()});
+    llvm::Type *DataTy = ConvertType(E->getType());
+    llvm::Type *RetTy = llvm::StructType::get(Builder.getContext(),
+                                              {DataTy, Builder.getInt1Ty()});
 
     SmallVector<Value *, 3> Args;
     Args.push_back(HandleOp);
@@ -386,7 +384,7 @@ Value *CodeGenFunction::EmitHLSLBuiltinExpr(unsigned BuiltinID,
 
     // Call the intrinsic (returns a struct)
     Value *ResRet =
-        Builder.CreateIntrinsic(IntrinsicRetTy, IntrID, Args, {}, "ld.struct");
+        Builder.CreateIntrinsic(RetTy, IntrID, Args, {}, "ld.struct");
 
     // Extract the loaded data (first element of the struct)
     Value *LoadedValue = Builder.CreateExtractValue(ResRet, {0}, "ld.value");

clang/lib/Sema/HLSLBuiltinTypeDeclBuilder.cpp

@@ -202,7 +202,7 @@ struct BuiltinTypeMethodBuilder {
   BuiltinTypeMethodBuilder &declareLocalVar(LocalVar &Var);
   template <typename... Ts>
   BuiltinTypeMethodBuilder &callBuiltin(StringRef BuiltinName,
-                                        QualType ReturnType, Ts... ArgSpecs);
+                                        QualType ReturnType, Ts &&...ArgSpecs);
   template <typename TLHS, typename TRHS>
   BuiltinTypeMethodBuilder &assign(TLHS LHS, TRHS RHS);
   template <typename T> BuiltinTypeMethodBuilder &dereference(T Ptr);
@@ -572,7 +572,7 @@ BuiltinTypeMethodBuilder &BuiltinTypeMethodBuilder::returnThis() {
 template <typename... Ts>
 BuiltinTypeMethodBuilder &
 BuiltinTypeMethodBuilder::callBuiltin(StringRef BuiltinName,
-                                      QualType ReturnType, Ts... ArgSpecs) {
+                                      QualType ReturnType, Ts &&...ArgSpecs) {
   ensureCompleteDecl();
 
   std::array<Expr *, sizeof...(ArgSpecs)> Args{
@@ -1240,23 +1240,20 @@ BuiltinTypeDeclBuilder::addLoadWithStatusFunction(DeclarationName &Name,
   ASTContext &AST = SemaRef.getASTContext();
   using PH = BuiltinTypeMethodBuilder::PlaceHolder;
 
-  QualType ElemTy = getHandleElementType();
-  QualType AddrSpaceElemTy =
-      AST.getAddrSpaceQualType(ElemTy, LangAS::hlsl_device);
-  QualType ElemPtrTy = AST.getPointerType(AddrSpaceElemTy);
-  QualType ReturnTy;
-
-  ReturnTy = ElemTy;
-  if (IsConst)
-    ReturnTy.addConst();
+  QualType ReturnTy = getHandleElementType();
+  BuiltinTypeMethodBuilder::LocalVar ResultVar("Result", ReturnTy);
+  BuiltinTypeMethodBuilder::LocalVar StatusVar("StatusBool", AST.BoolTy);
 
-  QualType StatusRefTy = AST.getLValueReferenceType(AST.UnsignedIntTy);
   return BuiltinTypeMethodBuilder(*this, Name, ReturnTy, IsConst)
       .addParam("Index", AST.UnsignedIntTy)
-      .addParam("Status", StatusRefTy)
-      .callBuiltin("__builtin_hlsl_resource_load_with_status", ElemPtrTy,
-                   PH::Handle, PH::_0, PH::_1)
-      .dereference(PH::LastStmt)
+      .addParam("Status", AST.UnsignedIntTy, HLSLParamModifierAttr::Keyword_out)
+      .declareLocalVar(ResultVar)
+      .declareLocalVar(StatusVar)
+      .callBuiltin("__builtin_hlsl_resource_load_with_status", ReturnTy,
+                   PH::Handle, PH::_0, StatusVar)
+      .assign(ResultVar, PH::LastStmt)
+      .assign(PH::_1, StatusVar)
+      .returnValue(ResultVar)
       .finalize();
 }
 

clang/lib/Sema/SemaHLSL.cpp

@@ -3016,17 +3016,13 @@ bool SemaHLSL::CheckBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
         CheckArgTypeMatches(&SemaRef, TheCall->getArg(1),
                             SemaRef.getASTContext().UnsignedIntTy) ||
         CheckArgTypeMatches(&SemaRef, TheCall->getArg(2),
-                            SemaRef.getASTContext().UnsignedIntTy))
+                            SemaRef.getASTContext().BoolTy))
       return true;
 
     auto *ResourceTy =
         TheCall->getArg(0)->getType()->castAs<HLSLAttributedResourceType>();
-    QualType ContainedTy = ResourceTy->getContainedType();
-    auto ReturnType =
-        SemaRef.Context.getAddrSpaceQualType(ContainedTy, LangAS::hlsl_device);
-    ReturnType = SemaRef.Context.getPointerType(ReturnType);
+    QualType ReturnType = ResourceTy->getContainedType();
     TheCall->setType(ReturnType);
-    TheCall->setValueKind(VK_LValue);
 
     break;
   }

clang/test/CodeGenHLSL/resources/TypedBuffers-methods.hlsl

@@ -47,12 +47,12 @@ export float TestLoadWithStatus() {
 }
 
 // CHECK: define noundef nofpclass(nan inf) float @TestLoadWithStatus()()
-// CHECK: call {{.*}} float @hlsl::Buffer<float>::Load(unsigned int, unsigned int&)(ptr {{.*}} @Buf, i32 noundef 1, ptr noundef nonnull align 4 dereferenceable(4) %s1)
-// CHECK: call {{.*}} <4 x i32> @hlsl::RWBuffer<unsigned int vector[4]>::Load(unsigned int, unsigned int&)(ptr {{.*}} @RWBuf, i32 noundef 2, ptr noundef nonnull align 4 dereferenceable(4) %s2)
+// CHECK: call {{.*}} float @hlsl::Buffer<float>::Load(unsigned int, unsigned int&)(ptr {{.*}} @Buf, i32 noundef 1, ptr noalias noundef nonnull align 4 dereferenceable(4) %tmp)
+// CHECK: call {{.*}} <4 x i32> @hlsl::RWBuffer<unsigned int vector[4]>::Load(unsigned int, unsigned int&)(ptr {{.*}} @RWBuf, i32 noundef 2, ptr noalias noundef nonnull align 4 dereferenceable(4) %tmp1)
 // CHECK: add
 // CHECK: ret float
 
-// CHECK: define {{.*}} float @hlsl::Buffer<float>::Load(unsigned int, unsigned int&)(ptr {{.*}} %this, i32 noundef %Index, ptr noundef nonnull align 4 dereferenceable(4) %Status)
+// CHECK: define {{.*}} float @hlsl::Buffer<float>::Load(unsigned int, unsigned int&)(ptr {{.*}} %this, i32 noundef %Index, ptr noalias noundef nonnull align 4 dereferenceable(4) %Status)
 // CHECK: %__handle = getelementptr inbounds nuw %"class.hlsl::Buffer", ptr %{{.*}}, i32 0, i32 0
 // DXIL-NEXT: %[[HANDLE:.*]] = load target("dx.TypedBuffer", float, 0, 0, 0), ptr %__handle
 // CHECK-NEXT: %[[INDEX:.*]] = load i32, ptr %Index.addr

llvm/lib/Target/DirectX/DXILShaderFlags.cpp

@@ -100,6 +100,27 @@ static bool checkWaveOps(Intrinsic::ID IID) {
   }
 }
 
+// Checks to see if the status bit from a load with status
+// instruction is ever extracted.
+// This is our proof that the module requires TiledResources
+// to be set, as if check access fully mapped was used.
+bool checkIfStatusIsExtracted(const Instruction &I) {
+  // Iterate over all uses of the instruction
+  for (const Use &U : I.uses()) {
+    const User *UserInst = U.getUser();
+
+    // Check if the user is an ExtractValue instruction
+    if (const ExtractValueInst *EVI = dyn_cast<ExtractValueInst>(UserInst)) {
+      // ExtractValueInst has a list of indices; check if it extracts index 1
+      if (EVI->getNumIndices() == 1 && EVI->getIndices()[0] == 1) {
+        return true;
+      }
+    }
+  }
+
+  return false;
+}
+
 /// Update the shader flags mask based on the given instruction.
 /// \param CSF Shader flags mask to update.
 /// \param I Instruction to check.
@@ -192,6 +213,13 @@ void ModuleShaderFlags::updateFunctionFlags(ComputedShaderFlags &CSF,
           DRTM[cast<TargetExtType>(II->getArgOperand(0)->getType())];
       if (RTI.isTyped())
         CSF.TypedUAVLoadAdditionalFormats |= RTI.getTyped().ElementCount > 1;
+      if (!CSF.TiledResources && checkIfStatusIsExtracted(I))
+        CSF.TiledResources = true;
+      break;
+    }
+    case Intrinsic::dx_resource_load_rawbuffer: {
+      if (!CSF.TiledResources && checkIfStatusIsExtracted(I))
+        CSF.TiledResources = true;
       break;
     }
     }