Merge pull request #3924 from psychocoderHPC/topic-pmaccHeaderChecks

PMacc: header include validation
ComputationalRadiationPhysics · Mar 6, 2023 · c96fc6c · c96fc6c
2 parents 0d674eb + dd1627d
commit c96fc6c
Show file tree

Hide file tree

Showing 80 changed files with 933 additions and 533 deletions.
diff --git a/include/pmacc/Environment.hpp b/include/pmacc/Environment.hpp
diff --git a/include/pmacc/Environment.tpp b/include/pmacc/Environment.tpp
diff --git a/include/pmacc/algorithms/TypeCast.hpp b/include/pmacc/algorithms/TypeCast.hpp
@@ -21,6 +21,8 @@
 
 #pragma once
 
+#include "pmacc/attribute/FunctionSpecifier.hpp"
+
 namespace pmacc
 {
     namespace algorithms

diff --git a/include/pmacc/communication/AsyncCommunication.hpp b/include/pmacc/communication/AsyncCommunication.hpp
@@ -21,6 +21,8 @@
 
 #pragma once
 
+#include "pmacc/eventSystem/events/EventTask.hpp"
+
 namespace pmacc
 {
     namespace communication

diff --git a/include/pmacc/communication/manager_common.hpp b/include/pmacc/communication/manager_common.hpp
@@ -25,6 +25,8 @@
 #include <stdexcept>
 #include <string>
 
+#include <mpi.h>
+
 const int GridManagerRank = 0;
 
 enum

diff --git a/include/pmacc/dataManagement/DataConnector.hpp b/include/pmacc/dataManagement/DataConnector.hpp
@@ -21,6 +21,7 @@
 
 #pragma once
 
+#include "pmacc/Environment.def"
 #include "pmacc/assert.hpp"
 #include "pmacc/dataManagement/AbstractInitialiser.hpp"
 #include "pmacc/dataManagement/ISimulationData.hpp"

diff --git a/include/pmacc/device/MemoryInfo.hpp b/include/pmacc/device/MemoryInfo.hpp
@@ -22,8 +22,11 @@
 #pragma once
 
 
+#include "pmacc/Environment.def"
 #include "pmacc/types.hpp"
 
+#include <pmacc/communication/manager_common.hpp>
+
 #include <cstring> // memset
 
 #include <mpi.h>

diff --git a/include/pmacc/eventSystem/events/EventPool.hpp b/include/pmacc/eventSystem/events/EventPool.hpp
@@ -22,6 +22,7 @@
 
 #pragma once
 
+#include "pmacc/Environment.def"
 #include "pmacc/debug/VerboseLog.hpp"
 #include "pmacc/eventSystem/events/CudaEvent.hpp"
 #include "pmacc/eventSystem/events/CudaEventHandle.hpp"

diff --git a/include/pmacc/eventSystem/events/kernelEvents.hpp b/include/pmacc/eventSystem/events/kernelEvents.hpp
@@ -24,84 +24,19 @@
 
 #include "pmacc/Environment.hpp"
 #include "pmacc/dimensions/DataSpace.hpp"
+#include "pmacc/exec/KernelLauncher.hpp"
+#include "pmacc/exec/KernelMetaData.hpp"
+#include "pmacc/exec/KernelWithDynSharedMem.hpp"
 #include "pmacc/traits/GetNComponents.hpp"
 #include "pmacc/types.hpp"
 
 #include <string>
 
 
-/* No namespace in this file since we only declare macro defines */
-
-/*if this flag is defined all kernel calls would be checked and synchronize
- * this flag must set by the compiler or inside of the Makefile
- */
-#if(PMACC_SYNC_KERNEL == 1)
-#    define CUDA_CHECK_KERNEL_MSG(...) CUDA_CHECK_MSG(__VA_ARGS__)
-#else
-/*no synchronize and check of kernel calls*/
-#    define CUDA_CHECK_KERNEL_MSG(...) ;
-#endif
-
-
 namespace pmacc::exec
 {
     namespace detail
     {
-        /** Kernel with dynamic shared memory
-         *
-         * This implements the possibility to define dynamic shared memory without
-         * specializing the needed alpaka trait BlockSharedMemDynSizeBytes for each kernel with shared memory.
-         * The trait BlockSharedMemDynSizeBytes is defined by PMacc for all types of KernelWithDynSharedMem.
-         */
-        template<typename T_Kernel>
-        struct KernelWithDynSharedMem : public T_Kernel
-        {
-            size_t const m_dynSharedMemBytes;
-
-            KernelWithDynSharedMem(T_Kernel const& kernel, size_t const& dynSharedMemBytes)
-                : T_Kernel(kernel)
-                , m_dynSharedMemBytes(dynSharedMemBytes)
-            {
-            }
-        };
-
-        //! Meta data of a device kernel
-        class KernelMetaData
-        {
-            //! file name
-            std::string const m_file;
-            //! line number
-            size_t const m_line;
-
-        public:
-            KernelMetaData(std::string const& file, size_t const line) : m_file(file), m_line(line)
-            {
-            }
-
-            //! file name from where the kernel is called
-            std::string getFile() const
-            {
-                return m_file;
-            }
-
-            //! line number in the file where the kernel is called
-            size_t getLine() const
-            {
-                return m_line;
-            }
-        };
-
-
-        /** Object to launch the kernel functor on the device.
-         *
-         * This objects contains the kernel functor, kernel meta information and the launch parameters.
-         * Object is used to enqueue the kernel with user arguments on the device.
-         *
-         * @tparam T_Kernel pmacc Kernel object
-         */
-        template<typename T_Kernel>
-        struct KernelLauncher;
-
         /** Wraps a user kernel functor to prepare the execution on the device.
          *
          * This objects contains the kernel functor, kernel meta information.
@@ -149,89 +84,6 @@ namespace pmacc::exec
                 size_t const sharedMemByte) const -> KernelLauncher<KernelWithDynSharedMem<T_KernelFunctor>>;
             /**@}*/
         };
-
-
-        template<typename T_Kernel>
-        struct KernelLauncher
-        {
-            //! kernel functor
-            T_Kernel const m_kernel;
-            //! Debug meta data for the kernel functor.
-            KernelMetaData const m_metaData;
-            //! grid extents for the kernel
-            cupla::dim3 const m_gridExtent;
-            //! block extents for the kernel
-            cupla::dim3 const m_blockExtent;
-
-            /** kernel starter object
-             *
-             * @param kernel pmacc Kernel
-             */
-            template<typename T_VectorGrid, typename T_VectorBlock>
-            HINLINE KernelLauncher(
-                T_Kernel const& kernel,
-                detail::KernelMetaData const& kernelMetaData,
-                T_VectorGrid const& gridExtent,
-                T_VectorBlock const& blockExtent)
-                : m_kernel(kernel)
-                , m_metaData(kernelMetaData)
-                , m_gridExtent(DataSpace<traits::GetNComponents<T_VectorGrid>::value>(gridExtent).toDim3())
-                , m_blockExtent(DataSpace<traits::GetNComponents<T_VectorBlock>::value>(blockExtent).toDim3())
-            {
-            }
-
-            /** Enqueue the kernel functor with the given arguments for execution.
-             *
-             * The stream into which the kernel is enqueued is automatically chosen by PMacc's event system.
-             *
-             * @tparam T_Args types of the arguments
-             * @param args arguments for the kernel functor
-             */
-            template<typename... T_Args>
-            HINLINE void operator()(T_Args&&... args) const
-            {
-                std::string const kernelName = typeid(m_kernel).name();
-                std::string const kernelInfo = kernelName + std::string(" [") + m_metaData.getFile() + std::string(":")
-                    + std::to_string(m_metaData.getLine()) + std::string(" ]");
-
-                CUDA_CHECK_KERNEL_MSG(cuplaDeviceSynchronize(), std::string("Crash before kernel call ") + kernelInfo);
-
-                pmacc::TaskKernel* taskKernel = pmacc::Environment<>::get().Factory().createTaskKernel(kernelName);
-
-                /* The next 3 lines cast pmacc vectors or integral values used for the execution extents into alpaka
-                 * vectors.
-                 * Note to be cupla compatible we use currently always 3-dimensional kernels.
-                 *
-                 * @todo find a better way to write this part of code, in general alpaka understands PMacc vectors but
-                 * PMacc has no easy way t cast integral numbers to an 3-dimensional vector.
-                 */
-                auto gridExtent = cupla::IdxVec3(m_gridExtent.z, m_gridExtent.y, m_gridExtent.x);
-                auto blockExtent = cupla::IdxVec3(m_blockExtent.z, m_blockExtent.y, m_blockExtent.x);
-                auto elemExtent = cupla::IdxVec3::ones();
-                auto workDiv
-                    = ::alpaka::WorkDivMembers<cupla::KernelDim, cupla::IdxType>(gridExtent, blockExtent, elemExtent);
-
-                auto const kernelTask
-                    = ::alpaka::createTaskKernel<cupla::Acc>(workDiv, m_kernel, std::forward<T_Args>(args)...);
-
-                auto cuplaStream = taskKernel->getCudaStream();
-                auto& stream = cupla::manager::Stream<cupla::AccDev, cupla::AccStream>::get().stream(cuplaStream);
-
-                ::alpaka::enqueue(stream, kernelTask);
-
-                CUDA_CHECK_KERNEL_MSG(
-                    cuplaGetLastError(),
-                    std::string("Last error after kernel launch ") + kernelInfo);
-                CUDA_CHECK_KERNEL_MSG(
-                    cuplaDeviceSynchronize(),
-                    std::string("Crash after kernel launch ") + kernelInfo);
-                taskKernel->activateChecks();
-                CUDA_CHECK_KERNEL_MSG(
-                    cuplaDeviceSynchronize(),
-                    std::string("Crash after kernel activation") + kernelInfo);
-            }
-        };
-
     } // namespace detail
 
     /** Creates a kernel object.
@@ -292,3 +144,4 @@ namespace alpaka
 
 
 #include "pmacc/eventSystem/events/kernelEvents.tpp"
+#include "pmacc/exec/KernelLauncher.tpp"
diff --git a/include/pmacc/eventSystem/tasks/Factory.hpp b/include/pmacc/eventSystem/tasks/Factory.hpp
@@ -22,6 +22,8 @@
 
 #pragma once
 
+#include "pmacc/Environment.def"
+#include "pmacc/eventSystem/events/EventTask.hpp"
 #include "pmacc/eventSystem/streams/EventStream.hpp"
 #include "pmacc/eventSystem/tasks/ITask.hpp"
 #include "pmacc/eventSystem/tasks/TaskKernel.hpp"

diff --git a/include/pmacc/eventSystem/tasks/TaskReceive.hpp b/include/pmacc/eventSystem/tasks/TaskReceive.hpp
@@ -23,6 +23,7 @@
 #pragma once
 
 #include "pmacc/Environment.hpp"
+#include "pmacc/eventSystem/Manager.hpp"
 #include "pmacc/eventSystem/events/EventDataReceive.hpp"
 #include "pmacc/eventSystem/tasks/MPITask.hpp"
 

diff --git a/include/pmacc/eventSystem/tasks/TaskReceiveMPI.hpp b/include/pmacc/eventSystem/tasks/TaskReceiveMPI.hpp
@@ -26,7 +26,6 @@
 #include "pmacc/communication/ICommunicator.hpp"
 #include "pmacc/eventSystem/events/EventDataReceive.hpp"
 #include "pmacc/eventSystem/tasks/MPITask.hpp"
-#include "pmacc/memory/buffers/Buffer.hpp"
 
 #include <memory>
 
@@ -47,7 +46,7 @@ namespace pmacc
 
         void init() override
         {
-            Buffer<TYPE, DIM>* dst = exchange->getCommunicationBuffer();
+            auto* dst = exchange->getCommunicationBuffer();
 
             this->request = Environment<DIM>::get().EnvironmentController().getCommunicator().startReceive(
                 exchange->getExchangeType(),

diff --git a/include/pmacc/eventSystem/tasks/TaskSendMPI.hpp b/include/pmacc/eventSystem/tasks/TaskSendMPI.hpp
@@ -25,7 +25,6 @@
 #include "pmacc/Environment.hpp"
 #include "pmacc/communication/ICommunicator.hpp"
 #include "pmacc/eventSystem/tasks/MPITask.hpp"
-#include "pmacc/memory/buffers/Buffer.hpp"
 
 #include <mpi.h>
 
@@ -44,7 +43,7 @@ namespace pmacc
 
         void init() override
         {
-            Buffer<TYPE, DIM>* src = exchange->getCommunicationBuffer();
+            auto* src = exchange->getCommunicationBuffer();
 
             this->request = Environment<DIM>::get().EnvironmentController().getCommunicator().startSend(
                 exchange->getExchangeType(),

diff --git a/include/pmacc/eventSystem/tasks/TaskSetValue.hpp b/include/pmacc/eventSystem/tasks/TaskSetValue.hpp
@@ -25,7 +25,7 @@
 #include "pmacc/dimensions/DataSpace.hpp"
 #include "pmacc/eventSystem/tasks/StreamTask.hpp"
 #include "pmacc/lockstep.hpp"
-#include "pmacc/lockstep/lockstep.hpp"
+#include "pmacc/lockstep/WorkerCfg.hpp"
 #include "pmacc/mappings/simulation/EnvironmentController.hpp"
 #include "pmacc/memory/boxes/DataBox.hpp"
 #include "pmacc/memory/buffers/DeviceBuffer.hpp"

diff --git a/include/pmacc/exec/KernelLauncher.hpp b/include/pmacc/exec/KernelLauncher.hpp
@@ -0,0 +1,36 @@
+/* Copyright 2022 Rene Widera
+ *
+ * This file is part of PMacc.
+ *
+ * PMacc is free software: you can redistribute it and/or modify
+ * it under the terms of either the GNU General Public License or
+ * the GNU Lesser General Public License as published by
+ * the Free Software Foundation, either version 3 of the License, or
+ * (at your option) any later version.
+ *
+ * PMacc is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License and the GNU Lesser General Public License
+ * for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * and the GNU Lesser General Public License along with PMacc.
+ * If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#pragma once
+
+
+namespace pmacc::exec::detail
+{
+    /** Object to launch the kernel functor on the device.
+     *
+     * This objects contains the kernel functor, kernel meta information and the launch parameters.
+     * Object is used to enqueue the kernel with user arguments on the device.
+     *
+     * @tparam T_Kernel pmacc Kernel object
+     */
+    template<typename T_Kernel>
+    struct KernelLauncher;
+} // namespace pmacc::exec::detail