[CIR] Backport TargetAddressSpaceAttr and Support both existing Lang AS and target AS attributes in pointer types and Global Ops

clang/include/clang/CIR/Dialect/Builder/CIRBaseBuilder.h

@@ -11,6 +11,7 @@
 
 #include "clang/AST/Decl.h"
 #include "clang/AST/Type.h"
+#include "clang/Basic/AddressSpaces.h"
 #include "clang/CIR/Dialect/IR/CIRAttrs.h"
 #include "clang/CIR/Dialect/IR/CIRDialect.h"
 #include "clang/CIR/Dialect/IR/CIROpsEnums.h"
@@ -105,20 +106,35 @@ class CIRBaseBuilderTy : public mlir::OpBuilder {
     return cir::PointerType::get(ty);
   }
 
-  cir::PointerType getPointerTo(mlir::Type ty, cir::AddressSpace as) {
-    return cir::PointerType::get(ty, as);
+  /// Create a pointer type with an address space attribute.
+  /// Either a cir::ClangAddressSpaceAttr or cir::TargetAddressSpaceAttr is
+  /// expected.
+  cir::PointerType getPointerTo(mlir::Type ty, mlir::Attribute addrSpaceAttr) {
+    if (!addrSpaceAttr)
+      return cir::PointerType::get(ty);
+
+    assert((mlir::isa<cir::ClangAddressSpaceAttr>(addrSpaceAttr) ||
+            mlir::isa<cir::TargetAddressSpaceAttr>(addrSpaceAttr)) &&
+           "expected address space attribute");
+
+    return cir::PointerType::get(ty, addrSpaceAttr);
   }
 
   cir::PointerType getPointerTo(mlir::Type ty, clang::LangAS langAS) {
-    return getPointerTo(ty, cir::toCIRAddressSpace(langAS));
+    if (langAS == clang::LangAS::Default)
+      return getPointerTo(ty);
+
+    mlir::Attribute addrSpaceAttr =
+        cir::toCIRClangAddressSpaceAttr(getContext(), langAS);
+    return getPointerTo(ty, addrSpaceAttr);
   }
 
   cir::PointerType getVoidPtrTy(clang::LangAS langAS = clang::LangAS::Default) {
     return getPointerTo(cir::VoidType::get(getContext()), langAS);
   }
 
-  cir::PointerType getVoidPtrTy(cir::AddressSpace as) {
-    return getPointerTo(cir::VoidType::get(getContext()), as);
+  cir::PointerType getVoidPtrTy(mlir::Attribute addrSpaceAttr) {
+    return getPointerTo(cir::VoidType::get(getContext()), addrSpaceAttr);
   }
 
   cir::MethodAttr getMethodAttr(cir::MethodType ty, cir::FuncOp methodFuncOp) {

clang/include/clang/CIR/Dialect/IR/CIRAttrConstraints.td

@@ -89,4 +89,24 @@ def CIR_AnyTBAAAttr : AnyAttrOf<[
   string cppType = "::mlir::Attribute";
 }
 
+//===----------------------------------------------------------------------===//
+// AddressSpaceAttr constraints
+//===----------------------------------------------------------------------===//
+
+// NOTE: We might end up using this only for GlobalOps, as we cannot apply constraints
+// to types.
+def CIR_AddressSpaceAttrConstraint
+  : CIR_AttrConstraint<"::cir::ClangAddressSpaceAttr", "clang address space attribute">;
+
+def CIR_TargetAddressSpaceAttrConstraint
+  : CIR_AttrConstraint<"::cir::TargetAddressSpaceAttr", "target address space attribute">;
+
+def CIR_AnyAddressSpaceAttr : AnyAttrOf<[
+  CIR_AddressSpaceAttrConstraint,
+  CIR_TargetAddressSpaceAttrConstraint
+]> {
+  string cppType = "::mlir::Attribute";
+  let constBuilderCall = "nullptr";
+}
+
 #endif // CLANG_CIR_DIALECT_IR_CIRATTRCONSTRAINTS_TD

clang/include/clang/CIR/Dialect/IR/CIRAttrs.td

@@ -963,43 +963,71 @@ def CIR_DynamicCastInfoAttr : CIR_Attr<"DynamicCastInfo", "dyn_cast_info"> {
 // AddressSpaceAttr
 //===----------------------------------------------------------------------===//
 
-def CIR_AddressSpaceAttr :  CIR_EnumAttr<CIR_AddressSpace, "address_space"> {
+def CIR_ClangAddressSpaceAttr :  CIR_EnumAttr<CIR_ClangAddressSpace, "clang_address_space"> {
+
+  let summary = "Represents a language/Clang-level address space";
+  let description = [{
+    Encodes the semantic address spaces defined by the front-end language
+    (e.g. `__shared__`, `__constant__`, `__local__`). Values are stored using the
+    `cir::ClangAddressSpace` enum, keeping the representation compact while and
+     preserves the qualifier until it is mapped onto target/LLVM address-space numbers.
+
+     Example: 
+     ``` mlir
+     !cir.ptr<!s32i, clang_address_space(offload_local)>
+     cir.global constant external clang_address_space(offload_constant)
+
+     ```
+  }];
+
   let builders = [
     AttrBuilder<(ins "clang::LangAS":$langAS), [{
-      return $_get($_ctxt, cir::toCIRAddressSpace(langAS));
+      return $_get($_ctxt, cir::toCIRClangAddressSpace(langAS));
     }]>
   ];
 
   let assemblyFormat = [{
     `(` custom<AddressSpaceValue>($value) `)`
   }];
 
-  let defaultValue = "cir::AddressSpace::Default";
+  let defaultValue = "cir::ClangAddressSpace::Default";
 
   let extraClassDeclaration = [{
-    bool isLang() const;
-    bool isTarget() const;
-    unsigned getTargetValue() const;
     unsigned getAsUnsignedValue() const;
   }];
 
   let extraClassDefinition = [{
     unsigned $cppClass::getAsUnsignedValue() const {
       return static_cast<unsigned>(getValue());
     }
+  }];
+}
 
-    bool $cppClass::isLang() const {
-      return cir::isLangAddressSpace(getValue());
-    }
+//===----------------------------------------------------------------------===//
+// TargetAddressSpaceAttr
+//===----------------------------------------------------------------------===//
 
-    bool $cppClass::isTarget() const {
-      return cir::isTargetAddressSpace(getValue());
-    }
+def CIR_TargetAddressSpaceAttr : CIR_Attr< "TargetAddressSpace",
+                                         "target_address_space"> {
+  let summary = "Represents a target-specific numeric address space";
+  let description = [{
+    The TargetAddressSpaceAttr represents a target-specific numeric address space,
+    corresponding to the LLVM IR `addressspace` qualifier and the clang
+     `address_space` attribute.
+
+    A value of zero represents the default address space. The semantics of non-zero
+    address spaces are target-specific.
 
-    unsigned $cppClass::getTargetValue() const {
-      return cir::getTargetAddressSpaceValue(getValue());
-    }
+    Example:
+    ```mlir
+    // Target-specific numeric address spaces
+    !cir.ptr<!s32i, target_address_space(1)>
+    !cir.ptr<!s32i, target_address_space(1)>
+    ```
   }];
+
+  let parameters = (ins "unsigned":$value);
+  let assemblyFormat = "`<` `target` `<` $value `>` `>`";
 }
 
 //===----------------------------------------------------------------------===//

clang/include/clang/CIR/Dialect/IR/CIREnumAttr.td

@@ -41,15 +41,14 @@ class CIR_DefaultValuedEnumParameter<EnumAttrInfo info, string value = "">
   let defaultValue = value;
 }
 
-def CIR_AddressSpace : CIR_I32EnumAttr<
-  "AddressSpace", "address space kind", [
+def CIR_ClangAddressSpace : CIR_I32EnumAttr<
+  "ClangAddressSpace", "clang address space kind", [
   I32EnumAttrCase<"Default", 0, "default">,
   I32EnumAttrCase<"OffloadPrivate", 1, "offload_private">,
   I32EnumAttrCase<"OffloadLocal", 2, "offload_local">,
   I32EnumAttrCase<"OffloadGlobal", 3, "offload_global">,
   I32EnumAttrCase<"OffloadConstant", 4, "offload_constant">,
-  I32EnumAttrCase<"OffloadGeneric", 5, "offload_generic">,
-  I32EnumAttrCase<"Target", 6, "target">
+  I32EnumAttrCase<"OffloadGeneric", 5, "offload_generic">
 ]> {
   let description = [{
     The `address_space` attribute is used to represent address spaces for

clang/include/clang/CIR/Dialect/IR/CIROps.td

@@ -2617,8 +2617,8 @@ def CIR_GlobalOp : CIR_Op<"global", [
     TypeAttr:$sym_type,
     CIR_GlobalLinkageKind:$linkage,
     DefaultValuedAttr<
-      CIR_AddressSpaceAttr,
-      "AddressSpace::Default"
+      CIR_AnyAddressSpaceAttr,
+      "{}"
     >:$addr_space,
     OptionalAttr<CIR_TLSModel>:$tls_model,
     // Note this can also be a FlatSymbolRefAttr
@@ -2642,7 +2642,7 @@ def CIR_GlobalOp : CIR_Op<"global", [
        (`comdat` $comdat^)?
        ($tls_model^)?
        (`dso_local` $dso_local^)?
-       (`addrspace` `` $addr_space^)?
+       (` ` custom<GlobalAddressSpaceValue>($addr_space)^ )?
        $sym_name
        custom<GlobalOpTypeAndInitialValue>($sym_type, $initial_value, $ctorRegion, $dtorRegion)
        ($annotations^)?
@@ -2666,7 +2666,7 @@ def CIR_GlobalOp : CIR_Op<"global", [
       // CIR defaults to external linkage.
       CArg<"cir::GlobalLinkageKind",
             "cir::GlobalLinkageKind::ExternalLinkage">:$linkage,
-      CArg<"cir::AddressSpace", "cir::AddressSpace::Default">:$addrSpace,
+      CArg<"mlir::Attribute", "{}">:$addrSpace,
       CArg<"llvm::function_ref<void(mlir::OpBuilder &, mlir::Location)>",
            "nullptr">:$ctorBuilder,
       CArg<"llvm::function_ref<void(mlir::OpBuilder &, mlir::Location)>",

clang/include/clang/CIR/Dialect/IR/CIRTypes.h

@@ -34,37 +34,16 @@ bool isSized(mlir::Type ty);
 // AddressSpace helpers
 //===----------------------------------------------------------------------===//
 
-cir::AddressSpace toCIRAddressSpace(clang::LangAS langAS);
+cir::ClangAddressSpace toCIRClangAddressSpace(clang::LangAS langAS);
 
-constexpr unsigned getAsUnsignedValue(cir::AddressSpace as) {
-  return static_cast<unsigned>(as);
-}
-
-inline constexpr unsigned TargetAddressSpaceOffset =
-    cir::getMaxEnumValForAddressSpace();
-
-// Target address space is used for target-specific address spaces that are not
-// part of the enum. Its value is represented as an offset from the maximum
-// value of the enum. Make sure that it is always the last enum value.
-static_assert(getAsUnsignedValue(cir::AddressSpace::Target) ==
-                  cir::getMaxEnumValForAddressSpace(),
-              "Target address space must be the last enum value");
-
-constexpr bool isTargetAddressSpace(cir::AddressSpace as) {
-  return getAsUnsignedValue(as) >= cir::getMaxEnumValForAddressSpace();
-}
+/// Convert a LangAS to the appropriate address space attribute.
+/// Returns ClangAddressSpaceAttr for clang/language-specific address spaces,
+/// or TargetAddressSpaceAttr for target-specific address spaces.
+mlir::Attribute toCIRClangAddressSpaceAttr(mlir::MLIRContext *ctx,
+                                           clang::LangAS langAS);
 
-constexpr bool isLangAddressSpace(cir::AddressSpace as) {
-  return !isTargetAddressSpace(as);
-}
-
-constexpr unsigned getTargetAddressSpaceValue(cir::AddressSpace as) {
-  assert(isTargetAddressSpace(as) && "expected target address space");
-  return getAsUnsignedValue(as) - TargetAddressSpaceOffset;
-}
-
-constexpr cir::AddressSpace computeTargetAddressSpace(unsigned v) {
-  return static_cast<cir::AddressSpace>(v + TargetAddressSpaceOffset);
+constexpr unsigned getAsUnsignedValue(cir::ClangAddressSpace as) {
+  return static_cast<unsigned>(as);
 }
 
 } // namespace cir

clang/include/clang/CIR/Dialect/IR/CIRTypes.td

@@ -16,11 +16,13 @@
 include "clang/CIR/Dialect/IR/CIRDialect.td"
 include "clang/CIR/Dialect/IR/CIREnumAttr.td"
 include "clang/CIR/Dialect/IR/CIRTypeConstraints.td"
+include "clang/CIR/Dialect/IR/CIRAttrConstraints.td"
 include "clang/CIR/Interfaces/ASTAttrInterfaces.td"
 include "clang/CIR/Interfaces/CIRTypeInterfaces.td"
 include "mlir/Interfaces/DataLayoutInterfaces.td"
 include "mlir/IR/AttrTypeBase.td"
 include "mlir/IR/EnumAttr.td"
+include "mlir/IR/CommonAttrConstraints.td"
 
 //===----------------------------------------------------------------------===//
 // CIR Types
@@ -231,32 +233,31 @@ def CIR_PointerType : CIR_Type<"Pointer", "ptr", [
     ```
   }];
 
+  let genVerifyDecl = 1;
+
   let parameters = (ins
     "mlir::Type":$pointee,
-    CIR_DefaultValuedEnumParameter<
-      CIR_AddressSpace,
-      "cir::AddressSpace::Default"
-    >:$addrSpace
+    OptionalParameter<"mlir::Attribute">:$addrSpace
   );
 
   let skipDefaultBuilders = 1;
   let builders = [
     TypeBuilderWithInferredContext<(ins
       "mlir::Type":$pointee,
-      CArg<"cir::AddressSpace", "cir::AddressSpace::Default">:$addrSpace), [{
+      CArg<"mlir::Attribute", "{}">:$addrSpace), [{
         return $_get(pointee.getContext(), pointee, addrSpace);
     }]>,
     TypeBuilder<(ins
       "mlir::Type":$pointee,
-      CArg<"cir::AddressSpace", "cir::AddressSpace::Default">:$addrSpace), [{
+      CArg<"mlir::Attribute", "{}">:$addrSpace), [{
         return $_get($_ctxt, pointee, addrSpace);
     }]>
   ];
 
   let assemblyFormat = [{
     `<`
       $pointee
-      ( `,` `addrspace` `(` `` custom<AddressSpaceValue>($addrSpace)^ `)` )?
+      ( `,` ` ` custom<AddressSpaceValue>($addrSpace)^ )?
     `>`
   }];
 

clang/include/clang/CIR/MissingFeatures.h

@@ -476,6 +476,8 @@ struct MissingFeatures {
   static bool mustProgress() { return false; }
 
   static bool skipTempCopy() { return false; }
+
+  static bool dataLayoutPtrHandlingBasedOnLangAS() { return false; }
 };
 
 } // namespace cir

clang/lib/CIR/CodeGen/CIRGenBuilder.h

@@ -749,7 +749,7 @@ class CIRGenBuilderTy : public cir::CIRBaseBuilderTy {
   [[nodiscard]] cir::GlobalOp
   createGlobal(mlir::ModuleOp module, mlir::Location loc, mlir::StringRef name,
                mlir::Type type, bool isConst, cir::GlobalLinkageKind linkage,
-               cir::AddressSpace addrSpace = cir::AddressSpace::Default) {
+               mlir::Attribute addrSpace = {}) {
     mlir::OpBuilder::InsertionGuard guard(*this);
     setInsertionPointToStart(module.getBody());
     return cir::GlobalOp::create(*this, loc, name, type, isConst, linkage,
@@ -759,10 +759,11 @@ class CIRGenBuilderTy : public cir::CIRBaseBuilderTy {
   /// Creates a versioned global variable. If the symbol is already taken, an ID
   /// will be appended to the symbol. The returned global must always be queried
   /// for its name so it can be referenced correctly.
-  [[nodiscard]] cir::GlobalOp createVersionedGlobal(
-      mlir::ModuleOp module, mlir::Location loc, mlir::StringRef name,
-      mlir::Type type, bool isConst, cir::GlobalLinkageKind linkage,
-      cir::AddressSpace addrSpace = cir::AddressSpace::Default) {
+  [[nodiscard]] cir::GlobalOp
+  createVersionedGlobal(mlir::ModuleOp module, mlir::Location loc,
+                        mlir::StringRef name, mlir::Type type, bool isConst,
+                        cir::GlobalLinkageKind linkage,
+                        mlir::Attribute addrSpace = {}) {
     // Create a unique name if the given name is already taken.
     std::string uniqueName;
     if (unsigned version = GlobalsVersioning[name.str()]++)

clang/lib/CIR/CodeGen/CIRGenBuiltin.cpp

@@ -34,6 +34,7 @@
 #include "clang/Frontend/FrontendDiagnostic.h"
 
 #include "mlir/Dialect/Func/IR/FuncOps.h"
+#include "mlir/IR/Attributes.h"
 #include "mlir/IR/BuiltinAttributes.h"
 #include "mlir/IR/Value.h"
 #include "mlir/Support/LLVM.h"
@@ -1727,9 +1728,9 @@ RValue CIRGenFunction::emitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
     // the AST level this is handled within CreateTempAlloca et al., but for the
     // builtin / dynamic alloca we have to handle it here.
     assert(!cir::MissingFeatures::addressSpace());
-    cir::AddressSpace AAS = getCIRAllocaAddressSpace();
-    cir::AddressSpace EAS = cir::toCIRAddressSpace(
-        E->getType()->getPointeeType().getAddressSpace());
+    mlir::Attribute AAS = getCIRAllocaAddressSpace();
+    mlir::Attribute EAS = cir::toCIRClangAddressSpaceAttr(
+        &getMLIRContext(), E->getType()->getPointeeType().getAddressSpace());
     if (EAS != AAS) {
       assert(false && "Non-default address space for alloca NYI");
     }

clang/lib/CIR/CodeGen/CIRGenCUDANV.cpp

@@ -329,7 +329,7 @@ mlir::Operation *CIRGenNVCUDARuntime::getKernelHandle(cir::FuncOp fn,
         return CIRGenModule::createGlobalOp(
             cgm, fn->getLoc(), globalName,
             builder.getPointerTo(fn.getFunctionType()), true,
-            cir::AddressSpace::Default,
+            /*addrSpace=*/{},
             /*insertPoint=*/nullptr);
       });
 

clang/lib/CIR/CodeGen/CIRGenDecl.cpp

@@ -479,7 +479,8 @@ CIRGenModule::getOrCreateStaticVarDecl(const VarDecl &D,
     Name = getStaticDeclName(*this, D);
 
   mlir::Type LTy = getTypes().convertTypeForMem(Ty);
-  cir::AddressSpace AS = cir::toCIRAddressSpace(getGlobalVarAddressSpace(&D));
+  mlir::Attribute AS = cir::toCIRClangAddressSpaceAttr(
+      &getMLIRContext(), getGlobalVarAddressSpace(&D));
 
   // OpenCL variables in local address space and CUDA shared
   // variables cannot have an initializer.