diff --git a/sycl/include/sycl/ext/oneapi/experimental/enqueue_functions.hpp b/sycl/include/sycl/ext/oneapi/experimental/enqueue_functions.hpp index f599078a6769e..5e2a6ec78feb4 100644 --- a/sycl/include/sycl/ext/oneapi/experimental/enqueue_functions.hpp +++ b/sycl/include/sycl/ext/oneapi/experimental/enqueue_functions.hpp @@ -110,6 +110,16 @@ event submit_with_event_impl(const queue &Q, PropertiesT Props, return Q.submit_with_event<__SYCL_USE_FALLBACK_ASSERT>( Props, detail::type_erased_cgfo_ty{CGF}, nullptr, CodeLoc); } + +template +event submit_with_event_impl(const queue &Q, PropertiesT Props, + nd_range Range, + const KernelType &KernelFunc, + const sycl::detail::code_location &CodeLoc) { + return Q.submit_with_event<__SYCL_USE_FALLBACK_ASSERT, PropertiesT, + KernelName, KernelType, Dims>(Props, Range, KernelFunc, CodeLoc); +} } // namespace detail template @@ -144,6 +154,19 @@ event submit_with_event(const queue &Q, CommandGroupFunc &&CGF, std::forward(CGF), CodeLoc); } +#ifdef __DPCPP_ENABLE_UNFINISHED_NO_CGH_SUBMIT +template +event submit_with_event(const queue &Q, PropertiesT Props, + nd_range Range, + const KernelType &KernelFunc, + const sycl::detail::code_location &CodeLoc = + sycl::detail::code_location::current()) { + return sycl::ext::oneapi::experimental::detail::submit_with_event_impl + (Q, Props, Range, KernelFunc, CodeLoc); +} +#endif + template void single_task(handler &CGH, const KernelType &KernelObj) { CGH.single_task(KernelObj); diff --git a/sycl/include/sycl/queue.hpp b/sycl/include/sycl/queue.hpp index a0dbdf5c540e8..b2a1106aa051b 100644 --- a/sycl/include/sycl/queue.hpp +++ b/sycl/include/sycl/queue.hpp @@ -149,6 +149,36 @@ class __SYCL_EXPORT SubmissionInfo { ext::oneapi::experimental::event_mode_enum::none; }; +using KernelParamDescGetterFuncPtr = detail::kernel_param_desc_t (*)(int); + +// This class is intended to store the kernel runtime information, +// extracted from the compile time kernel structures. +class __SYCL_EXPORT KernelRuntimeInfo { +public: + KernelRuntimeInfo() {} + + std::string_view &KernelName() { return MKernelName; } + std::unique_ptr &HostKernel() { return MHostKernel; } + int &KernelNumArgs() { return MKernelNumArgs; } + KernelParamDescGetterFuncPtr &KernelParamDescGetter() { + return MKernelParamDescGetter; + } + bool &KernelIsESIMD() { return MKernelIsESIMD; } + bool &KernelHasSpecialCaptures() { return MKernelHasSpecialCaptures; } + detail::KernelNameBasedCacheT *&KernelNameBasedCachePtr() { + return MKernelNameBasedCachePtr; + } + +private: + std::string_view MKernelName; + std::unique_ptr MHostKernel; + int MKernelNumArgs = 0; + KernelParamDescGetterFuncPtr MKernelParamDescGetter = nullptr; + bool MKernelIsESIMD = false; + bool MKernelHasSpecialCaptures = true; + detail::KernelNameBasedCacheT *MKernelNameBasedCachePtr = nullptr; +}; + } // namespace v1 } // namespace detail @@ -167,6 +197,13 @@ template event submit_with_event_impl(const queue &Q, PropertiesT Props, CommandGroupFunc &&CGF, const sycl::detail::code_location &CodeLoc); + +template +event submit_with_event_impl(const queue &Q, PropertiesT Props, + nd_range Range, + const KernelType &KernelFunc, + const sycl::detail::code_location &CodeLoc); } // namespace detail } // namespace ext::oneapi::experimental @@ -3215,11 +3252,17 @@ class __SYCL_EXPORT queue : public detail::OwnerLessBase { parallel_for(nd_range Range, RestT &&...Rest) { constexpr detail::code_location CodeLoc = getCodeLocation(); detail::tls_code_loc_t TlsCodeLocCapture(CodeLoc); +#ifdef __DPCPP_ENABLE_UNFINISHED_NO_CGH_SUBMIT + return submit_with_event(sycl::ext::oneapi::experimental::empty_properties_t{}, + Range, Rest..., CodeLoc); +#else return submit( [&](handler &CGH) { CGH.template parallel_for(Range, Rest...); }, TlsCodeLocCapture.query()); +#endif } /// parallel_for version with a kernel represented as a lambda + nd_range that @@ -3596,6 +3639,13 @@ class __SYCL_EXPORT queue : public detail::OwnerLessBase { const queue &Q, PropertiesT Props, CommandGroupFunc &&CGF, const sycl::detail::code_location &CodeLoc); + template + friend event ext::oneapi::experimental::detail::submit_with_event_impl( + const queue &Q, PropertiesT Props, nd_range Range, + const KernelType &KernelFunc, + const sycl::detail::code_location &CodeLoc); + template void ProcessSubmitProperties(PropertiesT Props, detail::v1::SubmissionInfo &SI) const { @@ -3609,6 +3659,36 @@ class __SYCL_EXPORT queue : public detail::OwnerLessBase { } } + template struct TransformUserItemType { + using type = std::conditional_t< + std::is_convertible_v, LambdaArgType>, nd_item, + std::conditional_t, LambdaArgType>, + item, LambdaArgType>>; + }; + + template + void ProcessKernelRuntimeInfo( + PropertiesT Props, const KernelType &KernelFunc, + detail::v1::KernelRuntimeInfo &KRInfo) const { + using NameT = + typename detail::get_kernel_name_t::name; + using LambdaArgType = sycl::detail::lambda_arg_type>; + using TransformedArgType = std::conditional_t< + std::is_integral::value && Dims == 1, item, + typename TransformUserItemType::type>; + + KRInfo.HostKernel().reset( + new detail::HostKernel( + KernelFunc)); + KRInfo.KernelName() = detail::getKernelName(); + KRInfo.KernelNumArgs() = detail::getKernelNumParams(); + KRInfo.KernelParamDescGetter() = &(detail::getKernelParamDesc); + KRInfo.KernelIsESIMD() = detail::isKernelESIMD(); + KRInfo.KernelHasSpecialCaptures() = detail::hasSpecialCaptures(); + KRInfo.KernelNameBasedCachePtr() = detail::getKernelNameBasedCache(); + } + #ifndef __INTEL_PREVIEW_BREAKING_CHANGES /// TODO: Unused. Remove these when ABI-break window is open. /// Not using `type_erased_cgfo_ty` on purpose. @@ -3680,6 +3760,24 @@ class __SYCL_EXPORT queue : public detail::OwnerLessBase { const detail::code_location &CodeLoc, bool IsTopCodeLoc) const; + event submit_with_event_impl( + nd_range<1> Range, + const detail::v1::SubmissionInfo &ExtSubmitInfo, + const detail::v1::KernelRuntimeInfo &KRInfo, + const detail::code_location &CodeLoc, bool IsTopCodeLoc) const; + + event submit_with_event_impl( + nd_range<2> Range, + const detail::v1::SubmissionInfo &ExtSubmitInfo, + const detail::v1::KernelRuntimeInfo &KRInfo, + const detail::code_location &CodeLoc, bool IsTopCodeLoc) const; + + event submit_with_event_impl( + nd_range<3> Range, + const detail::v1::SubmissionInfo &ExtSubmitInfo, + const detail::v1::KernelRuntimeInfo &KRInfo, + const detail::code_location &CodeLoc, bool IsTopCodeLoc) const; + /// A template-free version of submit_without_event as const member function. void submit_without_event_impl(const detail::type_erased_cgfo_ty &CGH, const detail::v1::SubmissionInfo &SubmitInfo, @@ -3763,6 +3861,26 @@ class __SYCL_EXPORT queue : public detail::OwnerLessBase { TlsCodeLocCapture.isToplevel()); } + template + event submit_with_event(PropertiesT Props, nd_range Range, + const KernelType &KernelFunc, + const detail::code_location &CodeLoc = + detail::code_location::current()) const { + detail::tls_code_loc_t TlsCodeLocCapture(CodeLoc); + detail::v1::SubmissionInfo SI{}; + detail::v1::KernelRuntimeInfo KRInfo{}; + + ProcessSubmitProperties(Props, SI); + ProcessKernelRuntimeInfo(Props, KernelFunc, KRInfo); + + // TODO UseFallbackAssert + + return submit_with_event_impl(Range, SI, KRInfo, + TlsCodeLocCapture.query(), TlsCodeLocCapture.isToplevel()); + } + /// Submits a command group function object to the queue, in order to be /// scheduled for execution on the device. /// diff --git a/sycl/source/detail/queue_impl.hpp b/sycl/source/detail/queue_impl.hpp index d6892011b991e..7bff56d1b8e3a 100644 --- a/sycl/source/detail/queue_impl.hpp +++ b/sycl/source/detail/queue_impl.hpp @@ -370,6 +370,45 @@ class queue_impl : public std::enable_shared_from_this { return createSyclObjFromImpl(ResEvent); } + event + submit_with_event(nd_range<1> Range, + const detail::v1::SubmissionInfo &SubmitInfo, + const detail::v1::KernelRuntimeInfo &KRInfo, + const detail::code_location &CodeLoc, bool IsTopCodeLoc) { + (void)Range; + (void)SubmitInfo; + (void)KRInfo; + (void)CodeLoc; + (void)IsTopCodeLoc; + return event(); + } + + event + submit_with_event(nd_range<2> Range, + const detail::v1::SubmissionInfo &SubmitInfo, + const detail::v1::KernelRuntimeInfo &KRInfo, + const detail::code_location &CodeLoc, bool IsTopCodeLoc) { + (void)Range; + (void)SubmitInfo; + (void)KRInfo; + (void)CodeLoc; + (void)IsTopCodeLoc; + return event(); + } + + event + submit_with_event(nd_range<3> Range, + const detail::v1::SubmissionInfo &SubmitInfo, + const detail::v1::KernelRuntimeInfo &KRInfo, + const detail::code_location &CodeLoc, bool IsTopCodeLoc) { + (void)Range; + (void)SubmitInfo; + (void)KRInfo; + (void)CodeLoc; + (void)IsTopCodeLoc; + return event(); + } + void submit_without_event(const detail::type_erased_cgfo_ty &CGF, const v1::SubmissionInfo &SubmitInfo, const detail::code_location &Loc, diff --git a/sycl/source/queue.cpp b/sycl/source/queue.cpp index ed0b0e42e6e6f..422148c0ec8f9 100644 --- a/sycl/source/queue.cpp +++ b/sycl/source/queue.cpp @@ -312,6 +312,33 @@ event queue::submit_with_event_impl( return impl->submit_with_event(CGH, SubmitInfo, CodeLoc, IsTopCodeLoc); } +event queue::submit_with_event_impl( + nd_range<1> Range, + const detail::v1::SubmissionInfo &SubmitInfo, + const detail::v1::KernelRuntimeInfo &KRInfo, + const detail::code_location &CodeLoc, bool IsTopCodeLoc) const { + return impl->submit_with_event(Range, SubmitInfo, KRInfo, + CodeLoc, IsTopCodeLoc); +} + +event queue::submit_with_event_impl( + nd_range<2> Range, + const detail::v1::SubmissionInfo &SubmitInfo, + const detail::v1::KernelRuntimeInfo &KRInfo, + const detail::code_location &CodeLoc, bool IsTopCodeLoc) const { + return impl->submit_with_event(Range, SubmitInfo, KRInfo, + CodeLoc, IsTopCodeLoc); +} + +event queue::submit_with_event_impl( + nd_range<3> Range, + const detail::v1::SubmissionInfo &SubmitInfo, + const detail::v1::KernelRuntimeInfo &KRInfo, + const detail::code_location &CodeLoc, bool IsTopCodeLoc) const { + return impl->submit_with_event(Range, SubmitInfo, KRInfo, + CodeLoc, IsTopCodeLoc); +} + void queue::submit_without_event_impl( const detail::type_erased_cgfo_ty &CGH, const detail::v1::SubmissionInfo &SubmitInfo,