Revision control
++ // Copyright (c) 2001-2004 NovodeX AG. All rights reserved.
++ // Copyright (c) 2004-2008 AGEIA Technologies, Inc. All rights reserved.
++ // SPDX-FileCopyrightText: Copyright (c) 2008-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
++ // SPDX-License-Identifier: Apache-2.0
++
++ #ifndef PX_DEFAULT_CPU_DISPATCHER_H
++ #define PX_DEFAULT_CPU_DISPATCHER_H
++
++ #include "common/PxPhysXCommonConfig.h"
++ #include "task/PxCpuDispatcher.h"
++
++ #if !PX_DOXYGEN
++ namespace physx
++ {
++ #endif
++
++ /**
++ \brief A default implementation for a CPU task dispatcher.
++
++ \see PxDefaultCpuDispatcherCreate() PxCpuDispatcher
++ */
++ class PxDefaultCpuDispatcher : public PxCpuDispatcher
++ {
++ public:
++ /**
++ \brief Deletes the dispatcher.
++
++ Do not keep a reference to the deleted instance.
++
++ \see PxDefaultCpuDispatcherCreate()
++ */
++ virtual void release() = 0;
++
++ /**
++ \brief Enables profiling at task level.
++
++ \note By default enabled only in profiling builds.
++
++ \param[in] runProfiled True if tasks should be profiled.
++ */
++ virtual void setRunProfiled(bool runProfiled) = 0;
++
++ /**
++ \brief Checks if profiling is enabled at task level.
++
++ \return True if tasks should be profiled.
++ */
++ virtual bool getRunProfiled() const = 0;
++ };
++
++
++ /**
++ \brief If a thread ends up waiting for work it will find itself in a spin-wait loop until work becomes available.
++ Three strategies are available to limit wasted cycles.
++ The strategies are as follows:
++ a) wait until a work task signals the end of the spin-wait period.
++ b) yield the thread by providing a hint to reschedule thread execution, thereby allowing other threads to run.
++ c) yield the processor by informing it that it is waiting for work and requesting it to more efficiently use compute resources.
++ */
++ struct PxDefaultCpuDispatcherWaitForWorkMode
++ {
++ enum Enum
++ {
++ eWAIT_FOR_WORK,
++ eYIELD_THREAD,
++ eYIELD_PROCESSOR
++ };
++ };
++
++
++ /**
++ \brief Create default dispatcher, extensions SDK needs to be initialized first.
++
++ \param[in] numThreads Number of worker threads the dispatcher should use.
++ \param[in] affinityMasks Array with affinity mask for each thread. If not defined, default masks will be used.
++ \param[in] mode is the strategy employed when a busy-wait is encountered.
++ \param[in] yieldProcessorCount specifies the number of times a OS-specific yield processor command will be executed
++ during each cycle of a busy-wait in the event that the specified mode is eYIELD_PROCESSOR
++
++ \note numThreads may be zero in which case no worker thread are initialized and
++ simulation tasks will be executed on the thread that calls PxScene::simulate()
++
++ \note yieldProcessorCount must be greater than zero if eYIELD_PROCESSOR is the chosen mode and equal to zero for all other modes.
++
++ \note eYIELD_THREAD and eYIELD_PROCESSOR modes will use compute resources even if the simulation is not running.
++ It is left to users to keep threads inactive, if so desired, when no simulation is running.
++
++ \see PxDefaultCpuDispatcher
++ */
++ PxDefaultCpuDispatcher* PxDefaultCpuDispatcherCreate(PxU32 numThreads, PxU32* affinityMasks = NULL, PxDefaultCpuDispatcherWaitForWorkMode::Enum mode = PxDefaultCpuDispatcherWaitForWorkMode::eWAIT_FOR_WORK, PxU32 yieldProcessorCount = 0);
++
++ #if !PX_DOXYGEN
++ } // namespace physx
++ #endif
++
++ #endif