[SYCL] Handler-less kernel submit path (single_task)

sycl/include/sycl/ext/oneapi/experimental/enqueue_functions.hpp

@@ -152,9 +152,21 @@ template <typename KernelName = sycl::detail::auto_name, typename KernelType>
 void single_task(queue Q, const KernelType &KernelObj,
                  const sycl::detail::code_location &CodeLoc =
                      sycl::detail::code_location::current()) {
-  submit(
-      std::move(Q),
-      [&](handler &CGH) { single_task<KernelName>(CGH, KernelObj); }, CodeLoc);
+  // TODO The handler-less path does not support kernel function properties
+  // and kernel functions with the kernel_handler type argument yet.
+  if constexpr (!(ext::oneapi::experimental::detail::
+                      HasKernelPropertiesGetMethod<
+                          const KernelType &>::value) &&
+                !(detail::KernelLambdaHasKernelHandlerArgT<KernelType,
+                                                           void>::value)) {
+    detail::submit_kernel_direct_single_task<KernelName>(
+        std::move(Q), empty_properties_t{}, KernelObj, CodeLoc);
+  } else {
+    submit(
+        std::move(Q),
+        [&](handler &CGH) { single_task<KernelName>(CGH, KernelObj); },
+        CodeLoc);
+  }
 }
 
 template <typename... ArgsT>
@@ -268,8 +280,8 @@ void nd_launch(queue Q, nd_range<Dimensions> Range, const KernelType &KernelObj,
                           const KernelType &>::value) &&
                 !(detail::KernelLambdaHasKernelHandlerArgT<
                     KernelType, sycl::nd_item<Dimensions>>::value)) {
-    detail::submit_kernel_direct<KernelName>(std::move(Q), empty_properties_t{},
-                                             Range, KernelObj);
+    detail::submit_kernel_direct_parallel_for<KernelName>(
+        std::move(Q), empty_properties_t{}, Range, KernelObj);
   } else {
     submit(std::move(Q), [&](handler &CGH) {
       nd_launch<KernelName>(CGH, Range, KernelObj,

sycl/include/sycl/khr/free_function_commands.hpp

@@ -164,7 +164,7 @@ void launch_grouped(const queue &q, range<1> r, range<1> size, KernelType &&k,
                           const KernelType &>::value) &&
                 !(detail::KernelLambdaHasKernelHandlerArgT<
                     KernelType, sycl::nd_item<1>>::value)) {
-    detail::submit_kernel_direct(
+    detail::submit_kernel_direct_parallel_for(
         q, ext::oneapi::experimental::empty_properties_t{},
         nd_range<1>(r, size), std::forward<KernelType>(k));
   } else {
@@ -185,7 +185,7 @@ void launch_grouped(const queue &q, range<2> r, range<2> size, KernelType &&k,
                           const KernelType &>::value) &&
                 !(detail::KernelLambdaHasKernelHandlerArgT<
                     KernelType, sycl::nd_item<2>>::value)) {
-    detail::submit_kernel_direct(
+    detail::submit_kernel_direct_parallel_for(
         q, ext::oneapi::experimental::empty_properties_t{},
         nd_range<2>(r, size), std::forward<KernelType>(k));
   } else {
@@ -206,7 +206,7 @@ void launch_grouped(const queue &q, range<3> r, range<3> size, KernelType &&k,
                           const KernelType &>::value) &&
                 !(detail::KernelLambdaHasKernelHandlerArgT<
                     KernelType, sycl::nd_item<3>>::value)) {
-    detail::submit_kernel_direct(
+    detail::submit_kernel_direct_parallel_for(
         q, ext::oneapi::experimental::empty_properties_t{},
         nd_range<3>(r, size), std::forward<KernelType>(k));
   } else {
@@ -319,11 +319,24 @@ void launch_task(handler &h, const KernelType &k) {
   h.single_task(k);
 }
 
-template <typename KernelType>
-void launch_task(const sycl::queue &q, const KernelType &k,
+template <typename KernelType, typename = typename std::enable_if_t<
+                                   enable_kernel_function_overload<KernelType>>>
+void launch_task(const sycl::queue &q, KernelType &&k,
                  const sycl::detail::code_location &codeLoc =
                      sycl::detail::code_location::current()) {
-  submit(q, [&](handler &h) { launch_task<KernelType>(h, k); }, codeLoc);
+  // TODO The handler-less path does not support kernel function properties
+  // and kernel functions with the kernel_handler type argument yet.
+  if constexpr (!(ext::oneapi::experimental::detail::
+                      HasKernelPropertiesGetMethod<
+                          const KernelType &>::value) &&
+                !(detail::KernelLambdaHasKernelHandlerArgT<KernelType,
+                                                           void>::value)) {
+    detail::submit_kernel_direct_single_task(
+        q, ext::oneapi::experimental::empty_properties_t{},
+        std::forward<KernelType>(k), codeLoc);
+  } else {
+    submit(q, [&](handler &h) { launch_task<KernelType>(h, k); }, codeLoc);
+  }
 }
 
 template <typename... Args>

sycl/include/sycl/queue.hpp

@@ -157,14 +157,14 @@ class __SYCL_EXPORT SubmissionInfo {
 
 } // namespace v1
 
-template <typename KernelName = detail::auto_name, bool EventNeeded = false,
+template <detail::WrapAs WrapAs, typename LambdaArgType,
+          typename KernelName = detail::auto_name, bool EventNeeded = false,
           typename PropertiesT, typename KernelTypeUniversalRef, int Dims>
 auto submit_kernel_direct(
-    const queue &Queue, PropertiesT Props, const nd_range<Dims> &Range,
-    KernelTypeUniversalRef &&KernelFunc,
+    const queue &Queue, [[maybe_unused]] PropertiesT Props,
+    const nd_range<Dims> &Range, KernelTypeUniversalRef &&KernelFunc,
     const detail::code_location &CodeLoc = detail::code_location::current()) {
   // TODO Properties not supported yet
-  (void)Props;
   static_assert(
       std::is_same_v<PropertiesT,
                      ext::oneapi::experimental::empty_properties_t>,
@@ -176,34 +176,39 @@ auto submit_kernel_direct(
 
   using NameT =
       typename detail::get_kernel_name_t<KernelName, KernelType>::name;
-  using LambdaArgType =
-      sycl::detail::lambda_arg_type<KernelType, nd_item<Dims>>;
-  static_assert(
-      std::is_convertible_v<sycl::nd_item<Dims>, LambdaArgType>,
-      "Kernel argument of a sycl::parallel_for with sycl::nd_range "
-      "must be either sycl::nd_item or be convertible from sycl::nd_item");
-  using TransformedArgType = sycl::nd_item<Dims>;
-
-#ifndef __SYCL_DEVICE_ONLY__
-  detail::checkValueRange<Dims>(Range);
-#endif
 
-  detail::KernelWrapper<detail::WrapAs::parallel_for, NameT, KernelType,
-                        TransformedArgType, PropertiesT>::wrap(KernelFunc);
+  detail::KernelWrapper<WrapAs, NameT, KernelType, LambdaArgType,
+                        PropertiesT>::wrap(KernelFunc);
 
-  HostKernelRef<KernelType, KernelTypeUniversalRef, TransformedArgType, Dims>
+  HostKernelRef<KernelType, KernelTypeUniversalRef, LambdaArgType, Dims>
       HostKernel(std::forward<KernelTypeUniversalRef>(KernelFunc));
 
   // Instantiating the kernel on the host improves debugging.
   // Passing this pointer to another translation unit prevents optimization.
 #ifndef NDEBUG
-  // TODO: call library to prevent dropping call due to optimization
+  // TODO: call library to prevent dropping call due to optimization.
   (void)
       detail::GetInstantiateKernelOnHostPtr<KernelType, LambdaArgType, Dims>();
 #endif
 
   detail::DeviceKernelInfo *DeviceKernelInfoPtr =
       &detail::getDeviceKernelInfo<NameT>();
+  constexpr auto Info = detail::CompileTimeKernelInfo<NameT>;
+
+  assert(Info.Name != std::string_view{} && "Kernel must have a name!");
+
+  static_assert(
+      Info.Name == std::string_view{} || sizeof(KernelType) == Info.KernelSize,
+      "Unexpected kernel lambda size. This can be caused by an "
+      "external host compiler producing a lambda with an "
+      "unexpected layout. This is a limitation of the compiler."
+      "In many cases the difference is related to capturing constexpr "
+      "variables. In such cases removing constexpr specifier aligns the "
+      "captures between the host compiler and the device compiler."
+      "\n"
+      "In case of MSVC, passing "
+      "-fsycl-host-compiler-options='/std:c++latest' "
+      "might also help.");
 
   if constexpr (EventNeeded) {
     return submit_kernel_direct_with_event_impl(
@@ -216,6 +221,48 @@ auto submit_kernel_direct(
   }
 }
 
+template <typename KernelName = detail::auto_name, bool EventNeeded = false,
+          typename PropertiesT, typename KernelTypeUniversalRef, int Dims>
+auto submit_kernel_direct_parallel_for(
+    const queue &Queue, PropertiesT Props, const nd_range<Dims> &Range,
+    KernelTypeUniversalRef &&KernelFunc,
+    const detail::code_location &CodeLoc = detail::code_location::current()) {
+
+  using KernelType =
+      std::remove_const_t<std::remove_reference_t<KernelTypeUniversalRef>>;
+
+  using LambdaArgType =
+      sycl::detail::lambda_arg_type<KernelType, nd_item<Dims>>;
+  static_assert(
+      std::is_convertible_v<sycl::nd_item<Dims>, LambdaArgType>,
+      "Kernel argument of a sycl::parallel_for with sycl::nd_range "
+      "must be either sycl::nd_item or be convertible from sycl::nd_item");
+  using TransformedArgType = sycl::nd_item<Dims>;
+
+#ifndef __SYCL_DEVICE_ONLY__
+  detail::checkValueRange<Dims>(Range);
+#endif
+
+  return submit_kernel_direct<detail::WrapAs::parallel_for, TransformedArgType,
+                              KernelName, EventNeeded, PropertiesT,
+                              KernelTypeUniversalRef, Dims>(
+      Queue, Props, Range, std::forward<KernelTypeUniversalRef>(KernelFunc),
+      CodeLoc);
+}
+
+template <typename KernelName = detail::auto_name, bool EventNeeded = false,
+          typename PropertiesT, typename KernelTypeUniversalRef>
+auto submit_kernel_direct_single_task(
+    const queue &Queue, PropertiesT Props, KernelTypeUniversalRef &&KernelFunc,
+    const detail::code_location &CodeLoc = detail::code_location::current()) {
+
+  return submit_kernel_direct<detail::WrapAs::single_task, void, KernelName,
+                              EventNeeded, PropertiesT, KernelTypeUniversalRef,
+                              1>(
+      Queue, Props, nd_range<1>{1, 1},
+      std::forward<KernelTypeUniversalRef>(KernelFunc), CodeLoc);
+}
+
 } // namespace detail
 
 namespace ext ::oneapi ::experimental {
@@ -2727,12 +2774,27 @@ class __SYCL_EXPORT queue : public detail::OwnerLessBase<queue> {
         "Use queue.submit() instead");
 
     detail::tls_code_loc_t TlsCodeLocCapture(CodeLoc);
-    return submit(
-        [&](handler &CGH) {
-          CGH.template single_task<KernelName, KernelType, PropertiesT>(
-              Properties, KernelFunc);
-        },
-        TlsCodeLocCapture.query());
+
+    // TODO The handler-less path does not support kernel
+    // function properties and kernel functions with the kernel_handler
+    // type argument yet.
+    if constexpr (
+        std::is_same_v<PropertiesT,
+                       ext::oneapi::experimental::empty_properties_t> &&
+        !(ext::oneapi::experimental::detail::HasKernelPropertiesGetMethod<
+            const KernelType &>::value) &&
+        !(detail::KernelLambdaHasKernelHandlerArgT<KernelType, void>::value)) {
+      return detail::submit_kernel_direct_single_task<KernelName, true>(
+          *this, ext::oneapi::experimental::empty_properties_t{}, KernelFunc,
+          TlsCodeLocCapture.query());
+    } else {
+      return submit(
+          [&](handler &CGH) {
+            CGH.template single_task<KernelName, KernelType, PropertiesT>(
+                Properties, KernelFunc);
+          },
+          TlsCodeLocCapture.query());
+    }
   }
 
   /// single_task version with a kernel represented as a lambda.
@@ -3291,7 +3353,7 @@ class __SYCL_EXPORT queue : public detail::OwnerLessBase<queue> {
                             const KernelType &>::value) &&
                   !(detail::KernelLambdaHasKernelHandlerArgT<
                       KernelType, sycl::nd_item<Dims>>::value)) {
-      return detail::submit_kernel_direct<KernelName, true>(
+      return detail::submit_kernel_direct_parallel_for<KernelName, true>(
           *this, ext::oneapi::experimental::empty_properties_t{}, Range,
           Rest..., TlsCodeLocCapture.query());
     } else {

sycl/test-e2e/Basic/test_num_kernel_copies.cpp

@@ -36,7 +36,8 @@ int main(int argc, char **argv) {
 
   kernel<2> krn2;
   q.single_task(krn2);
-  assert(copy_count == 1);
+  // The kernel is copied on the scheduler-based path only.
+  assert(copy_count == 0);
   assert(move_count == 0);
   copy_count = 0;
 

sycl/test/basic_tests/kernel_size_mismatch.cpp

@@ -13,7 +13,7 @@ int main() {
      (void)A;
     // expected-no-diagnostics
 #else
-  // expected-error-re@sycl/handler.hpp:* {{static assertion failed due to requirement '{{.*}}': Unexpected kernel lambda size. This can be caused by an external host compiler producing a lambda with an unexpected layout. This is a limitation of the compiler.}}
+  // expected-error-re@sycl/queue.hpp:* {{static assertion failed due to requirement '{{.*}}': Unexpected kernel lambda size. This can be caused by an external host compiler producing a lambda with an unexpected layout. This is a limitation of the compiler.}}
 #endif
    }).wait();
 }

sycl/test/basic_tests/single_task_error_message.cpp

@@ -12,6 +12,9 @@ int main() {
           .single_task([&](sycl::handler &cgh) {
             // expected-error-re@sycl/queue.hpp:* {{static assertion failed due to requirement '{{.*}}': sycl::queue.single_task() requires a kernel instead of command group.{{.*}} Use queue.submit() instead}}
             // expected-error-re@sycl/detail/cg_types.hpp:* {{no matching function for call to object of type '(lambda at {{.*}}single_task_error_message.cpp:{{.*}})'}}
+            // TODO Investigate why this function template is not instantiated
+            // (if this is expected).
+            // expected-error@sycl/detail/cg_types.hpp:* {{no matching function for call to 'runKernelWithoutArg'}}
           })
           .wait();
     }

sycl/unittests/Extensions/FreeFunctionCommands/FreeFunctionCommandsEvents.cpp

@@ -78,6 +78,7 @@ class FreeFunctionCommandsEventsTests : public ::testing::Test {
 protected:
   void SetUp() override {
     counter_urEnqueueKernelLaunch = 0;
+    counter_urEnqueueKernelLaunchWithEvent = 0;
     counter_urUSMEnqueueMemcpy = 0;
     counter_urUSMEnqueueFill = 0;
     counter_urUSMEnqueuePrefetch = 0;
@@ -281,6 +282,57 @@ TEST_F(FreeFunctionCommandsEventsTests,
   ASSERT_EQ(counter_urEnqueueKernelLaunchWithEvent, size_t{1});
 }
 
+TEST_F(FreeFunctionCommandsEventsTests, LaunchTaskShortcutMoveKernel) {
+  mock::getCallbacks().set_replace_callback("urEnqueueKernelLaunch",
+                                            &redefined_urEnqueueKernelLaunch);
+
+  TestMoveFunctor::MoveCtorCalls = 0;
+  TestMoveFunctor MoveOnly;
+  std::mutex CvMutex;
+  std::condition_variable Cv;
+  bool ready = false;
+
+  // This kernel submission uses scheduler-bypass path, so the HostKernel
+  // shouldn't be constructed.
+
+  sycl::khr::launch_task(Queue, std::move(MoveOnly));
+
+  ASSERT_EQ(TestMoveFunctor::MoveCtorCalls, 0);
+  ASSERT_EQ(counter_urEnqueueKernelLaunch, size_t{1});
+
+  // Another kernel submission is queued behind a host task,
+  // to force the scheduler-based submission. In this case, the HostKernel
+  // should be constructed.
+
+  // Replace the callback with an event based one, since the scheduler
+  // needs to create an event internally
+  mock::getCallbacks().set_replace_callback(
+      "urEnqueueKernelLaunch", &redefined_urEnqueueKernelLaunchWithEvent);
+
+  Queue.submit([&](sycl::handler &CGH) {
+    CGH.host_task([&] {
+      std::unique_lock<std::mutex> lk(CvMutex);
+      Cv.wait(lk, [&ready] { return ready; });
+    });
+  });
+
+  sycl::khr::launch_task(Queue, std::move(MoveOnly));
+
+  {
+    std::unique_lock<std::mutex> lk(CvMutex);
+    ready = true;
+  }
+  Cv.notify_one();
+
+  Queue.wait();
+
+  // Move ctor for TestMoveFunctor is called during move construction of
+  // HostKernel. Copy ctor is called by InstantiateKernelOnHost, can't delete
+  // it.
+  ASSERT_EQ(TestMoveFunctor::MoveCtorCalls, 1);
+  ASSERT_EQ(counter_urEnqueueKernelLaunchWithEvent, size_t{1});
+}
+
 TEST_F(FreeFunctionCommandsEventsTests, SubmitLaunchGroupedKernelNoEvent) {
   mock::getCallbacks().set_replace_callback("urEnqueueKernelLaunch",
                                             &redefined_urEnqueueKernelLaunch);