[SYCL] Add global_device and global_host address spaces

clang/include/clang/AST/Type.h

@@ -480,6 +480,12 @@ class Qualifiers {
            // Otherwise in OpenCLC v2.0 s6.5.5: every address space except
            // for __constant can be used as __generic.
            (A == LangAS::opencl_generic && B != LangAS::opencl_constant) ||
+           // We also define global_device and global_host address spaces,
+           // to distinguish global pointers allocated on host to pointers
+           // allocated on device, which are a subset of __global.
+           // FIXME: add a reference to spec when ready
+           (A == LangAS::opencl_global && (B == LangAS::opencl_global_device ||
+                                           B == LangAS::opencl_global_host)) ||
            // Consider pointer size address spaces to be equivalent to default.
            ((isPtrSizeAddressSpace(A) || A == LangAS::Default) &&
             (isPtrSizeAddressSpace(B) || B == LangAS::Default));
@@ -493,7 +499,9 @@ class Qualifiers {
            (!hasAddressSpace() &&
             (other.getAddressSpace() == LangAS::opencl_private ||
              other.getAddressSpace() == LangAS::opencl_local ||
-             other.getAddressSpace() == LangAS::opencl_global));
+             other.getAddressSpace() == LangAS::opencl_global ||
+             other.getAddressSpace() == LangAS::opencl_global_device ||
+             other.getAddressSpace() == LangAS::opencl_global_host));
   }
 
   /// Determines if these qualifiers compatibly include another set.

clang/include/clang/Basic/AddressSpaces.h

@@ -36,6 +36,8 @@ enum class LangAS : unsigned {
   opencl_constant,
   opencl_private,
   opencl_generic,
+  opencl_global_device,
+  opencl_global_host,
 
   // CUDA specific address spaces.
   cuda_device,

clang/include/clang/Basic/Attr.td

@@ -1294,6 +1294,16 @@ def OpenCLGlobalAddressSpace : TypeAttr {
   let Documentation = [OpenCLAddressSpaceGlobalDocs];
 }
 
+def OpenCLGlobalDeviceAddressSpace : TypeAttr {
+  let Spellings = [Clang<"opencl_global_device">];
+  let Documentation = [OpenCLGlobalAddressSpacesDocs];
+}
+
+def OpenCLGlobalHostAddressSpace : TypeAttr {
+  let Spellings = [Clang<"opencl_global_host">];
+  let Documentation = [OpenCLGlobalAddressSpacesDocs];
+}
+
 def OpenCLLocalAddressSpace : TypeAttr {
   let Spellings = [Keyword<"__local">, Keyword<"local">, Clang<"opencl_local">];
   let Documentation = [OpenCLAddressSpaceLocalDocs];

clang/include/clang/Basic/AttrDocs.td

@@ -3559,6 +3559,17 @@ scope) variables and static local variable as well.
   }];
 }
 
+def OpenCLGlobalAddressSpacesDocs : Documentation {
+  let Category = DocOpenCLAddressSpaces;
+  let Heading = "[[clang::opencl_global_device]], [[clang::opencl_global_host]]";
+  let Content = [{
+The (global_device) and (global_host) address space attributes specify that an
+object is allocated in global memory on the device/host. It helps distinguishing
+USM pointers that access device memory and accessors that access global memory
+from those that access host memory.
+  }];
+}
+
 def OpenCLAddressSpaceLocalDocs : Documentation {
   let Category = DocOpenCLAddressSpaces;
   let Heading = "__local, local, [[clang::opencl_local]]";

clang/include/clang/Sema/ParsedAttr.h

@@ -606,6 +606,10 @@ class ParsedAttr final
       return LangAS::opencl_constant;
     case ParsedAttr::AT_OpenCLGlobalAddressSpace:
       return LangAS::opencl_global;
+    case ParsedAttr::AT_OpenCLGlobalDeviceAddressSpace:
+      return LangAS::opencl_global_device;
+    case ParsedAttr::AT_OpenCLGlobalHostAddressSpace:
+      return LangAS::opencl_global_host;
     case ParsedAttr::AT_OpenCLLocalAddressSpace:
       return LangAS::opencl_local;
     case ParsedAttr::AT_OpenCLPrivateAddressSpace:

clang/lib/AST/ASTContext.cpp

@@ -919,18 +919,20 @@ static const LangASMap *getAddressSpaceMap(const TargetInfo &T,
     // The fake address space map must have a distinct entry for each
     // language-specific address space.
     static const unsigned FakeAddrSpaceMap[] = {
-        0, // Default
-        1, // opencl_global
-        3, // opencl_local
-        2, // opencl_constant
-        0, // opencl_private
-        4, // opencl_generic
-        5, // cuda_device
-        6, // cuda_constant
-        7, // cuda_shared
-        8, // ptr32_sptr
-        9, // ptr32_uptr
-        10 // ptr64
+        0,  // Default
+        1,  // opencl_global
+        3,  // opencl_local
+        2,  // opencl_constant
+        0,  // opencl_private
+        4,  // opencl_generic
+        11, // opencl_global_device
+        12, // opencl_global_host
+        5,  // cuda_device
+        6,  // cuda_constant
+        7,  // cuda_shared
+        8,  // ptr32_sptr
+        9,  // ptr32_uptr
+        10  // ptr64
     };
     return &FakeAddrSpaceMap;
   } else {

clang/lib/AST/ItaniumMangle.cpp

@@ -2387,16 +2387,39 @@ void CXXNameMangler::mangleQualifiers(Qualifiers Quals, const DependentAddressSp
       switch (AS) {
       default: llvm_unreachable("Not a language specific address space");
       //  <OpenCL-addrspace> ::= "CL" [ "global" | "local" | "constant" |
-      //                                "private"| "generic" ]
-      case LangAS::opencl_global:   ASString = "CLglobal";   break;
-      case LangAS::opencl_local:    ASString = "CLlocal";    break;
-      case LangAS::opencl_constant: ASString = "CLconstant"; break;
-      case LangAS::opencl_private:  ASString = "CLprivate";  break;
-      case LangAS::opencl_generic:  ASString = "CLgeneric";  break;
+      //                                "private"| "generic" | "global_device" |
+      //                                "global_host" ]
+      case LangAS::opencl_global:
+        ASString = "CLglobal";
+        break;
+      case LangAS::opencl_global_device:
+        ASString = "CLDevice";
+        break;
+      case LangAS::opencl_global_host:
+        ASString = "CLHost";
+        break;
+      case LangAS::opencl_local:
+        ASString = "CLlocal";
+        break;
+      case LangAS::opencl_constant:
+        ASString = "CLconstant";
+        break;
+      case LangAS::opencl_private:
+        ASString = "CLprivate";
+        break;
+      case LangAS::opencl_generic:
+        ASString = "CLgeneric";
+        break;
       //  <CUDA-addrspace> ::= "CU" [ "device" | "constant" | "shared" ]
-      case LangAS::cuda_device:     ASString = "CUdevice";   break;
-      case LangAS::cuda_constant:   ASString = "CUconstant"; break;
-      case LangAS::cuda_shared:     ASString = "CUshared";   break;
+      case LangAS::cuda_device:
+        ASString = "CUdevice";
+        break;
+      case LangAS::cuda_constant:
+        ASString = "CUconstant";
+        break;
+      case LangAS::cuda_shared:
+        ASString = "CUshared";
+        break;
       //  <ptrsize-addrspace> ::= [ "ptr32_sptr" | "ptr32_uptr" | "ptr64" ]
       case LangAS::ptr32_sptr:
         ASString = "ptr32_sptr";

clang/lib/AST/MicrosoftMangle.cpp

@@ -1823,6 +1823,12 @@ void MicrosoftCXXNameMangler::mangleAddressSpaceType(QualType T,
     case LangAS::opencl_global:
       Extra.mangleSourceName("_ASCLglobal");
       break;
+    case LangAS::opencl_global_device:
+      Extra.mangleSourceName("_ASCLDevice");
+      break;
+    case LangAS::opencl_global_host:
+      Extra.mangleSourceName("_ASCLHost");
+      break;
     case LangAS::opencl_local:
       Extra.mangleSourceName("_ASCLlocal");
       break;

clang/lib/AST/TypePrinter.cpp

@@ -1577,6 +1577,8 @@ void TypePrinter::printAttributedAfter(const AttributedType *T,
 
   case attr::OpenCLPrivateAddressSpace:
   case attr::OpenCLGlobalAddressSpace:
+  case attr::OpenCLGlobalDeviceAddressSpace:
+  case attr::OpenCLGlobalHostAddressSpace:
   case attr::OpenCLLocalAddressSpace:
   case attr::OpenCLConstantAddressSpace:
   case attr::OpenCLGenericAddressSpace:
@@ -1880,6 +1882,10 @@ std::string Qualifiers::getAddrSpaceAsString(LangAS AS) {
     return "__constant";
   case LangAS::opencl_generic:
     return "__generic";
+  case LangAS::opencl_global_device:
+    return "__global_device";
+  case LangAS::opencl_global_host:
+    return "__global_host";
   case LangAS::cuda_device:
     return "__device__";
   case LangAS::cuda_constant:

clang/lib/Basic/Targets/AMDGPU.cpp

@@ -45,6 +45,8 @@ const LangASMap AMDGPUTargetInfo::AMDGPUDefIsGenMap = {
     Constant, // opencl_constant
     Private,  // opencl_private
     Generic,  // opencl_generic
+    Global,   // opencl_global_device
+    Global,   // opencl_global_host
     Global,   // cuda_device
     Constant, // cuda_constant
     Local,    // cuda_shared
@@ -60,6 +62,8 @@ const LangASMap AMDGPUTargetInfo::AMDGPUDefIsPrivMap = {
     Constant, // opencl_constant
     Private,  // opencl_private
     Generic,  // opencl_generic
+    Global,   // opencl_global_device
+    Global,   // opencl_global_host
     Global,   // cuda_device
     Constant, // cuda_constant
     Local,    // cuda_shared

clang/lib/Basic/Targets/NVPTX.h

@@ -30,6 +30,8 @@ static const unsigned NVPTXAddrSpaceMap[] = {
     0, // opencl_private
     // FIXME: generic has to be added to the target
     0, // opencl_generic
+    1, // opencl_global_device
+    1, // opencl_global_host
     1, // cuda_device
     4, // cuda_constant
     3, // cuda_shared

clang/lib/Basic/Targets/SPIR.h

@@ -24,33 +24,37 @@ namespace clang {
 namespace targets {
 
 static const unsigned SPIRAddrSpaceMap[] = {
-    0, // Default
-    1, // opencl_global
-    3, // opencl_local
-    2, // opencl_constant
-    0, // opencl_private
-    4, // opencl_generic
-    0, // cuda_device
-    0, // cuda_constant
-    0, // cuda_shared
-    0, // ptr32_sptr
-    0, // ptr32_uptr
-    0  // ptr64
+    0,  // Default
+    1,  // opencl_global
+    3,  // opencl_local
+    2,  // opencl_constant
+    0,  // opencl_private
+    4,  // opencl_generic
+    11, // opencl_global_device
+    12, // opencl_global_host
+    0,  // cuda_device
+    0,  // cuda_constant
+    0,  // cuda_shared
+    0,  // ptr32_sptr
+    0,  // ptr32_uptr
+    0   // ptr64
 };
 
 static const unsigned SYCLAddrSpaceMap[] = {
-    4, // Default
-    1, // opencl_global
-    3, // opencl_local
-    2, // opencl_constant
-    0, // opencl_private
-    4, // opencl_generic
-    0, // cuda_device
-    0, // cuda_constant
-    0, // cuda_shared
-    0, // ptr32_sptr
-    0, // ptr32_uptr
-    0  // ptr64
+    4,  // Default
+    1,  // opencl_global
+    3,  // opencl_local
+    2,  // opencl_constant
+    0,  // opencl_private
+    4,  // opencl_generic
+    11, // opencl_global_device
+    12, // opencl_global_host
+    0,  // cuda_device
+    0,  // cuda_constant
+    0,  // cuda_shared
+    0,  // ptr32_sptr
+    0,  // ptr32_uptr
+    0   // ptr64
 };
 
 class LLVM_LIBRARY_VISIBILITY SPIRTargetInfo : public TargetInfo {

clang/lib/Basic/Targets/TCE.h

@@ -35,6 +35,8 @@ static const unsigned TCEOpenCLAddrSpaceMap[] = {
     4, // opencl_local
     5, // opencl_constant
     0, // opencl_private
+    1, // opencl_global_device
+    1, // opencl_global_host
     // FIXME: generic has to be added to the target
     0, // opencl_generic
     0, // cuda_device

clang/lib/Basic/Targets/X86.h

@@ -29,6 +29,8 @@ static const unsigned X86AddrSpaceMap[] = {
     0,   // opencl_constant
     0,   // opencl_private
     0,   // opencl_generic
+    0,   // opencl_global_device
+    0,   // opencl_global_host
     0,   // cuda_device
     0,   // cuda_constant
     0,   // cuda_shared

clang/lib/CodeGen/CodeGenModule.cpp

@@ -1360,10 +1360,18 @@ static void removeImageAccessQualifier(std::string& TyName) {
 // (basically all single AS CPUs).
 static unsigned ArgInfoAddressSpace(LangAS AS) {
   switch (AS) {
-  case LangAS::opencl_global:   return 1;
-  case LangAS::opencl_constant: return 2;
-  case LangAS::opencl_local:    return 3;
-  case LangAS::opencl_generic:  return 4; // Not in SPIR 2.0 specs.
+  case LangAS::opencl_global:
+    return 1;
+  case LangAS::opencl_constant:
+    return 2;
+  case LangAS::opencl_local:
+    return 3;
+  case LangAS::opencl_generic:
+    return 4; // Not in SPIR 2.0 specs.
+  case LangAS::opencl_global_device:
+    return 11;
+  case LangAS::opencl_global_host:
+    return 12;
   default:
     return 0; // Assume private.
   }
@@ -3858,6 +3866,8 @@ LangAS CodeGenModule::GetGlobalVarAddressSpace(const VarDecl *D) {
   if (LangOpts.OpenCL) {
     AddrSpace = D ? D->getType().getAddressSpace() : LangAS::opencl_global;
     assert(AddrSpace == LangAS::opencl_global ||
+           AddrSpace == LangAS::opencl_global_device ||
+           AddrSpace == LangAS::opencl_global_host ||
            AddrSpace == LangAS::opencl_constant ||
            AddrSpace == LangAS::opencl_local ||
            AddrSpace >= LangAS::FirstTargetAddressSpace);

clang/lib/Sema/SemaType.cpp

@@ -7967,6 +7967,8 @@ static bool isAddressSpaceKind(const ParsedAttr &attr) {
   return attrKind == ParsedAttr::AT_AddressSpace ||
          attrKind == ParsedAttr::AT_OpenCLPrivateAddressSpace ||
          attrKind == ParsedAttr::AT_OpenCLGlobalAddressSpace ||
+         attrKind == ParsedAttr::AT_OpenCLGlobalDeviceAddressSpace ||
+         attrKind == ParsedAttr::AT_OpenCLGlobalHostAddressSpace ||
          attrKind == ParsedAttr::AT_OpenCLLocalAddressSpace ||
          attrKind == ParsedAttr::AT_OpenCLConstantAddressSpace ||
          attrKind == ParsedAttr::AT_OpenCLGenericAddressSpace;
@@ -8048,6 +8050,8 @@ static void processTypeAttrs(TypeProcessingState &state, QualType &type,
       break;
     case ParsedAttr::AT_OpenCLPrivateAddressSpace:
     case ParsedAttr::AT_OpenCLGlobalAddressSpace:
+    case ParsedAttr::AT_OpenCLGlobalDeviceAddressSpace:
+    case ParsedAttr::AT_OpenCLGlobalHostAddressSpace:
     case ParsedAttr::AT_OpenCLLocalAddressSpace:
     case ParsedAttr::AT_OpenCLConstantAddressSpace:
     case ParsedAttr::AT_OpenCLGenericAddressSpace:

clang/test/AST/language_address_space_attribute.cpp

@@ -10,6 +10,18 @@ void langas() {
   // CHECK: VarDecl {{.*}} z_global '__global int *'
   [[clang::opencl_global]] int *z_global;
 
+  // CHECK: VarDecl {{.*}} x_global_device '__global_device int *'
+  __attribute__((opencl_global_device)) int *x_global_device;
+
+  // CHECK: VarDecl {{.*}} z_global_device '__global_device int *'
+  [[clang::opencl_global_device]] int *z_global_device;
+
+  // CHECK: VarDecl {{.*}} x_global_host '__global_host int *'
+  __attribute__((opencl_global_host)) int *x_global_host;
+
+  // CHECK: VarDecl {{.*}} z_global_host '__global_host int *'
+  [[clang::opencl_global_host]] int *z_global_host;
+
   // CHECK: VarDecl {{.*}} x_local '__local int *'
   __attribute__((opencl_local)) int *x_local;
 

clang/test/CodeGenCXX/mangle-address-space.cpp

@@ -43,6 +43,10 @@ void ocl_f0(char __private *p) { }
 
 struct ocl_OpaqueType;
 typedef ocl_OpaqueType __global * ocl_OpaqueTypePtr;
+typedef ocl_OpaqueType __attribute__((opencl_global_host)) * ocl_OpaqueTypePtrH;
+typedef ocl_OpaqueType
+    __attribute__((opencl_global_device)) *
+    ocl_OpaqueTypePtrD;
 
 // CHECKOCL-LABEL: define {{.*}}void @_Z6ocl_f0PU8CLglobal14ocl_OpaqueType
 // WINOCL-LABEL: define {{.*}}void @"?ocl_f0@@YAXPEAU?$_ASCLglobal@$$CAUocl_OpaqueType@@@__clang@@@Z"
@@ -61,4 +65,12 @@ __constant float *ocl_f1(char __generic const *p) { return 0;}
 // CHECKOCL-LABEL: define {{.*}}float* @_Z6ocl_f2PU9CLgenericKc
 // WINOCL-LABEL: define {{.*}}float* @"?ocl_f2@@YAPEAU?$_ASCLgeneric@$$CAM@__clang@@QEAU?$_ASCLgeneric@$$CBD@2@@Z"
 __generic float *ocl_f2(__generic char const * const p) { return 0;}
+
+// CHECKOCL-LABEL: define {{.*}}void @_Z6ocl_f3PU6CLHost14ocl_OpaqueType
+// WINOCL-LABEL: define {{.*}}void @"?ocl_f3@@YAXPEAU?$_ASCLHost@$$CAUocl_OpaqueType@@@__clang@@@Z"
+void ocl_f3(ocl_OpaqueTypePtrH) {}
+
+// CHECKOCL-LABEL: define {{.*}}void @_Z6ocl_f4PU8CLDevice14ocl_OpaqueType
+// WINOCL-LABEL: define {{.*}}void @"?ocl_f4@@YAXPEAU?$_ASCLDevice@$$CAUocl_OpaqueType@@@__clang@@@Z"
+void ocl_f4(ocl_OpaqueTypePtrD) {}
 #endif