diff --git a/llvm/include/llvm/Frontend/OpenMP/OMPGridValues.h b/llvm/include/llvm/Frontend/OpenMP/OMPGridValues.h
--- a/llvm/include/llvm/Frontend/OpenMP/OMPGridValues.h
+++ b/llvm/include/llvm/Frontend/OpenMP/OMPGridValues.h
@@ -120,6 +120,18 @@
     128,  // GV_Default_WG_Size
 };
 
+/// For Intel GPUs
+static constexpr GV SPIR64GridValues64 = {
+    256,  // GV_Slot_Size
+    64,   // GV_Warp_Size
+    (1 << 16), // GV_Max_Teams
+    440,  // GV_Default_Num_Teams
+    896,  // GV_SimpleBufferSize
+    1024, // GV_Max_WG_Size,
+    256,  // GV_Default_WG_Size
+};
+
+
 } // namespace omp
 } // namespace llvm
 
diff --git a/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def b/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def
--- a/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def
+++ b/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def
@@ -1210,6 +1210,7 @@
 __OMP_TRAIT_PROPERTY(device, arch, amdgcn)
 __OMP_TRAIT_PROPERTY(device, arch, nvptx)
 __OMP_TRAIT_PROPERTY(device, arch, nvptx64)
+__OMP_TRAIT_PROPERTY(device, arch, spir64)
 
 __OMP_TRAIT_SET(implementation)
 
diff --git a/llvm/lib/Frontend/OpenMP/OMPContext.cpp b/llvm/lib/Frontend/OpenMP/OMPContext.cpp
--- a/llvm/lib/Frontend/OpenMP/OMPContext.cpp
+++ b/llvm/lib/Frontend/OpenMP/OMPContext.cpp
@@ -51,6 +51,7 @@
   case Triple::amdgcn:
   case Triple::nvptx:
   case Triple::nvptx64:
+  case Triple::spir64:
     ActiveTraits.set(unsigned(TraitProperty::device_kind_gpu));
     break;
   default:
diff --git a/openmp/libomptarget/DeviceRTL/CMakeLists.txt b/openmp/libomptarget/DeviceRTL/CMakeLists.txt
--- a/openmp/libomptarget/DeviceRTL/CMakeLists.txt
+++ b/openmp/libomptarget/DeviceRTL/CMakeLists.txt
@@ -278,6 +278,10 @@
   compileDeviceRTLLibrary(${mcpu} amdgpu amdgcn-amd-amdhsa -fopenmp-targets=amdgcn-amd-amdhsa)
 endforeach()
 
+add_custom_target(omptarget.devicertl.spir64)
+# TODO Adapt names/versions here
+compileDeviceRTLLibrary(spir64 spir64 spir64-intel-l0 -fopenmp-targets=spir64)
+
 # Archive all the object files generated above into a static library
 add_library(omptarget.devicertl STATIC)
 set_target_properties(omptarget.devicertl PROPERTIES LINKER_LANGUAGE CXX)
diff --git a/openmp/libomptarget/DeviceRTL/include/Types.h b/openmp/libomptarget/DeviceRTL/include/Types.h
--- a/openmp/libomptarget/DeviceRTL/include/Types.h
+++ b/openmp/libomptarget/DeviceRTL/include/Types.h
@@ -140,6 +140,11 @@
 using LaneMaskTy = uint64_t;
 #pragma omp end declare variant
 
+#pragma omp begin declare variant match(device = {arch(spir64)})
+// TODO verify this
+using LaneMaskTy = uint64_t;
+#pragma omp end declare variant
+
 namespace lanes {
 enum : LaneMaskTy { All = ~(LaneMaskTy)0 };
 } // namespace lanes
diff --git a/openmp/libomptarget/DeviceRTL/src/LibC.cpp b/openmp/libomptarget/DeviceRTL/src/LibC.cpp
--- a/openmp/libomptarget/DeviceRTL/src/LibC.cpp
+++ b/openmp/libomptarget/DeviceRTL/src/LibC.cpp
@@ -33,6 +33,15 @@
 } // namespace impl
 #pragma omp end declare variant
 
+/// Intel implementation
+#pragma omp begin declare variant match(device = {arch(spir64)})
+namespace impl {
+int32_t omp_vprintf(const char *Format, void *Arguments, uint32_t) {
+  return -1;
+}
+} // namespace impl
+#pragma omp end declare variant
+
 extern "C" {
 
 int memcmp(const void *lhs, const void *rhs, size_t count) {
diff --git a/openmp/libomptarget/DeviceRTL/src/Mapping.cpp b/openmp/libomptarget/DeviceRTL/src/Mapping.cpp
--- a/openmp/libomptarget/DeviceRTL/src/Mapping.cpp
+++ b/openmp/libomptarget/DeviceRTL/src/Mapping.cpp
@@ -151,6 +151,60 @@
 #pragma omp end declare variant
 ///}
 
+/// Intel implementation
+///
+///{
+#pragma omp begin declare variant match(device = {arch(spir64)})
+
+const llvm::omp::GV &getGridValue() { return llvm::omp::SPIR64GridValues64; }
+
+uint32_t getNumHardwareThreadsInBlock() {
+  return __spirv_BuiltInWorkgroupSize(0);
+}
+
+LaneMaskTy activemask() { return -1; }
+
+LaneMaskTy lanemaskLT() {
+  uint32_t Lane = mapping::getThreadIdInWarp();
+  int64_t Ballot = mapping::activemask();
+  uint64_t Mask = ((uint64_t)1 << Lane) - (uint64_t)1;
+  return Mask & Ballot;
+}
+
+LaneMaskTy lanemaskGT() {
+  uint32_t Lane = mapping::getThreadIdInWarp();
+  if (Lane == (mapping::getWarpSize() - 1))
+    return 0;
+  int64_t Ballot = mapping::activemask();
+  uint64_t Mask = (~((uint64_t)0)) << (Lane + 1);
+  return Mask & Ballot;
+}
+
+uint32_t getThreadIdInWarp() {
+  return impl::getThreadIdInBlock() & (mapping::getWarpSize() - 1);
+}
+
+uint32_t getThreadIdInBlock() { return __spirv_BuiltInLocalInvocationId(0); }
+
+uint32_t getKernelSize() { return __spirv_BuiltInGlobalSize(0); }
+
+uint32_t getBlockId() { return __spirv_BuiltInWorkgroupId(0); }
+
+uint32_t getNumberOfBlocks() {
+  return __spirv_BuiltInGlobalSize(0) / __spirv_BuiltInWorkgroupSize(0);
+}
+
+uint32_t getWarpId() {
+  return impl::getThreadIdInBlock() / mapping::getWarpSize();
+}
+
+uint32_t getNumberOfWarpsInBlock() {
+  return mapping::getBlockSize() / mapping::getWarpSize();
+}
+
+#pragma omp end declare variant
+///}
+
 uint32_t getWarpSize() { return getGridValue().GV_Warp_Size; }
 
 } // namespace impl
diff --git a/openmp/libomptarget/DeviceRTL/src/Misc.cpp b/openmp/libomptarget/DeviceRTL/src/Misc.cpp
--- a/openmp/libomptarget/DeviceRTL/src/Misc.cpp
+++ b/openmp/libomptarget/DeviceRTL/src/Misc.cpp
@@ -58,6 +58,18 @@
 
 #pragma omp end declare variant
 
+/// Intel implementation
+#pragma omp begin declare variant match(device = {arch(spir64)})
+
+double getWTick() { return ((double)1E-9); }
+
+double getWTime() {
+  // TODO
+  return 0;
+}
+
+#pragma omp end declare variant
+
 } // namespace impl
 } // namespace ompx
 
diff --git a/openmp/libomptarget/DeviceRTL/src/State.cpp b/openmp/libomptarget/DeviceRTL/src/State.cpp
--- a/openmp/libomptarget/DeviceRTL/src/State.cpp
+++ b/openmp/libomptarget/DeviceRTL/src/State.cpp
@@ -66,6 +66,23 @@
 #pragma omp end declare variant
 ///}
 
+/// Intel implementations of the shuffle sync idiom.
+///
+///{
+#pragma omp begin declare variant match(device = {arch(spir64)})
+
+extern "C" {
+void *malloc(uint64_t Size) {
+  // TODO: Use some preallocated space for dynamic malloc.
+  return nullptr;
+}
+
+void free(void *Ptr) {}
+}
+
+#pragma omp end declare variant
+///}
+
 /// A "smart" stack in shared memory.
 ///
 /// The stack exposes a malloc/free interface but works like a stack internally.
diff --git a/openmp/libomptarget/DeviceRTL/src/Synchronization.cpp b/openmp/libomptarget/DeviceRTL/src/Synchronization.cpp
--- a/openmp/libomptarget/DeviceRTL/src/Synchronization.cpp
+++ b/openmp/libomptarget/DeviceRTL/src/Synchronization.cpp
@@ -351,6 +351,62 @@
 #pragma omp end declare variant
 ///}
 
+/// Intel Implementation
+///
+///{
+#pragma omp begin declare variant match(device = {arch(spir64)})
+
+uint32_t atomicInc(uint32_t *A, uint32_t V, atomic::OrderingTy Ordering) {
+  // TODO
+}
+
+uint32_t SHARED(namedBarrierTracker);
+
+void namedBarrierInit() {
+  // TODO
+}
+
+void namedBarrier() {
+  // TODO
+}
+
+// sema checking of amdgcn_fence is aggressive. Intention is to patch clang
+// so that it is usable within a template environment and so that a runtime
+// value of the memory order is expanded to this switch within clang/llvm.
+void fenceTeam(atomic::OrderingTy Ordering) {
+  // TODO
+}
+void fenceKernel(atomic::OrderingTy Ordering) {
+  // TODO
+}
+void fenceSystem(atomic::OrderingTy Ordering) {
+  // TODO
+}
+
+void syncWarp(__kmpc_impl_lanemask_t) {
+  // TODO
+}
+
+void syncThreads() { // TODO
+}
+void syncThreadsAligned() { // TODO
+}
+
+// TODO: Don't have wavefront lane locks. Possibly can't have them.
+void unsetLock(omp_lock_t *) { // TODO
+}
+int testLock(omp_lock_t *) { // TODO
+}
+void initLock(omp_lock_t *) { // TODO
+}
+void destroyLock(omp_lock_t *) { // TODO
+}
+void setLock(omp_lock_t *) { // TODO
+}
+
+#pragma omp end declare variant
+///}
+
 } // namespace impl
 
 void synchronize::init(bool IsSPMD) {
diff --git a/openmp/libomptarget/DeviceRTL/src/Utils.cpp b/openmp/libomptarget/DeviceRTL/src/Utils.cpp
--- a/openmp/libomptarget/DeviceRTL/src/Utils.cpp
+++ b/openmp/libomptarget/DeviceRTL/src/Utils.cpp
@@ -79,6 +79,24 @@
 #pragma omp end declare variant
 ///}
 
+/// Intel Implementation
+///
+///{
+#pragma omp begin declare variant match(device = {arch(spir64)})
+
+void Unpack(uint64_t Val, uint32_t *LowBits, uint32_t *HighBits) {
+  static_assert(sizeof(unsigned long) == 8, "");
+  *LowBits = (uint32_t)(Val & 0x00000000FFFFFFFFUL);
+  *HighBits = (uint32_t)((Val & 0xFFFFFFFF00000000UL) >> 32);
+}
+
+uint64_t Pack(uint32_t LowBits, uint32_t HighBits) {
+  return (((uint64_t)HighBits) << 32) | (uint64_t)LowBits;
+}
+
+#pragma omp end declare variant
+///}
+
 int32_t shuffle(uint64_t Mask, int32_t Var, int32_t SrcLane);
 int32_t shuffleDown(uint64_t Mask, int32_t Var, uint32_t LaneDelta,
                     int32_t Width);
@@ -126,6 +144,25 @@
 
 bool isSharedMemPtr(const void *Ptr) { return __nvvm_isspacep_shared(Ptr); }
 
+#pragma omp end declare variant
+///}
+
+/// Intel Implementation
+///
+///{
+#pragma omp begin declare variant match(device = {arch(spir64)})
+
+int32_t shuffle(uint64_t Mask, int32_t Var, int32_t SrcLane) {
+  // TODO
+}
+
+int32_t shuffleDown(uint64_t Mask, int32_t Var, uint32_t Delta, int32_t Width) {
+  // TODO
+}
+
+bool isSharedMemPtr(const void *Ptr) { // TODO
+}
+
 #pragma omp end declare variant
 ///}
 } // namespace impl