1 files changed, 35 insertions, 5 deletions
diff --git a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
index 468bc1d..4863b45 100644
--- a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
+++ b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
@@ -707,6 +707,23 @@ static void instantiateDependentAMDGPUMaxNumWorkGroupsAttr(
     S.AMDGPU().addAMDGPUMaxNumWorkGroupsAttr(New, Attr, XExpr, YExpr, ZExpr);
 }
 
+static void instantiateDependentCUDAClusterDimsAttr(
+    Sema &S, const MultiLevelTemplateArgumentList &TemplateArgs,
+    const CUDAClusterDimsAttr &Attr, Decl *New) {
+  EnterExpressionEvaluationContext Unevaluated(
+      S, Sema::ExpressionEvaluationContext::ConstantEvaluated);
+
+  auto SubstElt = [&S, &TemplateArgs](Expr *E) {
+    return E ? S.SubstExpr(E, TemplateArgs).get() : nullptr;
+  };
+
+  Expr *XExpr = SubstElt(Attr.getX());
+  Expr *YExpr = SubstElt(Attr.getY());
+  Expr *ZExpr = SubstElt(Attr.getZ());
+
+  S.addClusterDimsAttr(New, Attr, XExpr, YExpr, ZExpr);
+}
+
 // This doesn't take any template parameters, but we have a custom action that
 // needs to happen when the kernel itself is instantiated. We need to run the
 // ItaniumMangler to mark the names required to name this kernel.
@@ -765,10 +782,18 @@ static bool isRelevantAttr(Sema &S, const Decl *D, const Attr *A) {
 
 static void instantiateDependentHLSLParamModifierAttr(
     Sema &S, const MultiLevelTemplateArgumentList &TemplateArgs,
-    const HLSLParamModifierAttr *Attr, Decl *New) {
-  ParmVarDecl *P = cast<ParmVarDecl>(New);
-  P->addAttr(Attr->clone(S.getASTContext()));
-  P->setType(S.HLSL().getInoutParameterType(P->getType()));
+    const HLSLParamModifierAttr *Attr, const Decl *Old, Decl *New) {
+  ParmVarDecl *NewParm = cast<ParmVarDecl>(New);
+  NewParm->addAttr(Attr->clone(S.getASTContext()));
+
+  const Type *OldParmTy = cast<ParmVarDecl>(Old)->getType().getTypePtr();
+  if (OldParmTy->isDependentType() && Attr->isAnyOut())
+    NewParm->setType(S.HLSL().getInoutParameterType(NewParm->getType()));
+
+  assert(
+      (!Attr->isAnyOut() || (NewParm->getType().isRestrictQualified() &&
+                             NewParm->getType()->isReferenceType())) &&
+      "out or inout parameter type must be a reference and restrict qualified");
 }
 
 void Sema::InstantiateAttrsForDecl(
@@ -921,9 +946,14 @@ void Sema::InstantiateAttrs(const MultiLevelTemplateArgumentList &TemplateArgs,
           *this, TemplateArgs, *AMDGPUMaxNumWorkGroups, New);
     }
 
+    if (const auto *CUDAClusterDims = dyn_cast<CUDAClusterDimsAttr>(TmplAttr)) {
+      instantiateDependentCUDAClusterDimsAttr(*this, TemplateArgs,
+                                              *CUDAClusterDims, New);
+    }
+
     if (const auto *ParamAttr = dyn_cast<HLSLParamModifierAttr>(TmplAttr)) {
       instantiateDependentHLSLParamModifierAttr(*this, TemplateArgs, ParamAttr,
-                                                New);
+                                                Tmpl, New);
       continue;
     }