aboutsummaryrefslogtreecommitdiff
path: root/PhysX_3.4/Snippets/SnippetVehicleMultiThreading/SnippetVehicleMultiThreading.cpp
diff options
context:
space:
mode:
authorgit perforce import user <a@b>2016-10-25 12:29:14 -0600
committerSheikh Dawood Abdul Ajees <Sheikh Dawood Abdul Ajees>2016-10-25 18:56:37 -0500
commit3dfe2108cfab31ba3ee5527e217d0d8e99a51162 (patch)
treefa6485c169e50d7415a651bf838f5bcd0fd3bfbd /PhysX_3.4/Snippets/SnippetVehicleMultiThreading/SnippetVehicleMultiThreading.cpp
downloadphysx-3.4-3dfe2108cfab31ba3ee5527e217d0d8e99a51162.tar.xz
physx-3.4-3dfe2108cfab31ba3ee5527e217d0d8e99a51162.zip
Initial commit:
PhysX 3.4.0 Update @ 21294896 APEX 1.4.0 Update @ 21275617 [CL 21300167]
Diffstat (limited to 'PhysX_3.4/Snippets/SnippetVehicleMultiThreading/SnippetVehicleMultiThreading.cpp')
-rw-r--r--PhysX_3.4/Snippets/SnippetVehicleMultiThreading/SnippetVehicleMultiThreading.cpp612
1 files changed, 612 insertions, 0 deletions
diff --git a/PhysX_3.4/Snippets/SnippetVehicleMultiThreading/SnippetVehicleMultiThreading.cpp b/PhysX_3.4/Snippets/SnippetVehicleMultiThreading/SnippetVehicleMultiThreading.cpp
new file mode 100644
index 00000000..007748fb
--- /dev/null
+++ b/PhysX_3.4/Snippets/SnippetVehicleMultiThreading/SnippetVehicleMultiThreading.cpp
@@ -0,0 +1,612 @@
+// This code contains NVIDIA Confidential Information and is disclosed to you
+// under a form of NVIDIA software license agreement provided separately to you.
+//
+// Notice
+// NVIDIA Corporation and its licensors retain all intellectual property and
+// proprietary rights in and to this software and related documentation and
+// any modifications thereto. Any use, reproduction, disclosure, or
+// distribution of this software and related documentation without an express
+// license agreement from NVIDIA Corporation is strictly prohibited.
+//
+// ALL NVIDIA DESIGN SPECIFICATIONS, CODE ARE PROVIDED "AS IS.". NVIDIA MAKES
+// NO WARRANTIES, EXPRESSED, IMPLIED, STATUTORY, OR OTHERWISE WITH RESPECT TO
+// THE MATERIALS, AND EXPRESSLY DISCLAIMS ALL IMPLIED WARRANTIES OF NONINFRINGEMENT,
+// MERCHANTABILITY, AND FITNESS FOR A PARTICULAR PURPOSE.
+//
+// Information and code furnished is believed to be accurate and reliable.
+// However, NVIDIA Corporation assumes no responsibility for the consequences of use of such
+// information or for any infringement of patents or other rights of third parties that may
+// result from its use. No license is granted by implication or otherwise under any patent
+// or patent rights of NVIDIA Corporation. Details are subject to change without notice.
+// This code supersedes and replaces all information previously supplied.
+// NVIDIA Corporation products are not authorized for use as critical
+// components in life support devices or systems without express written approval of
+// NVIDIA Corporation.
+//
+// Copyright (c) 2008-2016 NVIDIA Corporation. All rights reserved.
+// Copyright (c) 2004-2008 AGEIA Technologies, Inc. All rights reserved.
+// Copyright (c) 2001-2004 NovodeX AG. All rights reserved.
+
+// ****************************************************************************
+// This snippet illustrates multi-threaded vehicles.
+//
+// It creates multiple vehicles on a plane and then concurrently simulates them
+// in parallel across multiple threads.
+
+// The concurrent vehicle simulation is split into four steps:
+// 1. Suspension raycasts
+// 2. Vehicle updates
+// 3. Vehicle post-updates
+// 4. SDK simulate
+
+// Steps 1 and 2 above both make use of concurrency to improve performance.
+
+// Step 3 above is an extra step that is necessary when PxVehicleUpdates is performed concurrently.
+// The vehicle post-update step must be performed sequentially.
+
+// The 4 steps above are timed with profile zones. The total accumulated time spent in each of
+// these steps is recorded and printed at the end of the simulation. With PVD attached, profiles
+// can be further analyzed in the "Profile Zones" view in PVD.
+
+// The number of threads can be set by changing NUM_WORKER_THREADS, while the number of
+// vehicles can be set by changing NUM_VEHICLES. The number of vehicles processed per task
+// can be modified by setting RAYCAST_BATCH_SIZE and UPDATE_BATCH_SIZE. It is worthwhile
+// experimenting with these parameters to get a feel for the performance gains possible in
+// different scenarios and across various platforms.
+
+// To avoid the overhead of PVD affecting the integrity of the profiling data the default
+// behavior is for PVD to only record profile data.
+// Visualizing the scene will reduce the integrity of the performance numbers.
+
+// Performance statistics are only generated in "profile" build.
+// PVD is not active at all in "release" build.
+
+// ****************************************************************************
+
+#include <ctype.h>
+
+#include "PxPhysicsAPI.h"
+
+#include "vehicle/PxVehicleUtil.h"
+
+#include "../SnippetVehicleCommon/SnippetVehicleCreate.h"
+#include "../SnippetVehicleCommon/SnippetVehicleSceneQuery.h"
+#include "../SnippetVehicleCommon/SnippetVehicleFilterShader.h"
+#include "../SnippetVehicleCommon/SnippetVehicleTireFriction.h"
+#include "../SnippetVehicleCommon/SnippetVehicleWheelQueryResult.h"
+#include "../SnippetVehicleCommon/SnippetVehicleConcurrency.h"
+
+#include "../SnippetUtils/SnippetUtils.h"
+#include "../SnippetCommon/SnippetPrint.h"
+#include "../SnippetCommon/SnippetPVD.h"
+
+#include "foundation/PxProfiler.h"
+
+
+using namespace physx;
+using namespace snippetvehicle;
+
+PxDefaultAllocator gAllocator;
+PxDefaultErrorCallback gErrorCallback;
+
+PxFoundation* gFoundation = NULL;
+PxPhysics* gPhysics = NULL;
+
+PxDefaultCpuDispatcher* gDispatcher = NULL;
+PxScene* gScene = NULL;
+
+PxCooking* gCooking = NULL;
+
+PxMaterial* gMaterial = NULL;
+
+PxPvd* gPvd = NULL;
+
+#if 1
+PxPvdInstrumentationFlags gConnectionFlags = PxPvdInstrumentationFlag::ePROFILE;
+#else
+PxPvdInstrumentationFlags gConnectionFlags = PxPvdInstrumentationFlag::eALL;
+#endif
+
+PxTaskManager* gTaskManager = NULL;
+
+PxRigidStatic* gGroundPlane = NULL;
+
+#define NUM_VEHICLES 1024
+PxVehicleWheels* gVehicles[NUM_VEHICLES];
+PxBatchQuery* gBatchQueries[NUM_VEHICLES];
+PxVehicleDrivableSurfaceToTireFrictionPairs* gFrictionPairs = NULL;
+VehicleSceneQueryData* gVehicleSceneQueryData = NULL;
+VehicleWheelQueryResults* gVehicleWheelQueryResults = NULL;
+VehicleConcurrency* gVehicleConcurrency = NULL;
+
+static const int gNumNames = 4;
+static const char* gNames[gNumNames] =
+{
+ "concurrentVehicleRaycasts",
+ "concurrentVehicleUpdates",
+ "concurrentVehiclePostUpdates",
+ "Basic.simulate"
+};
+
+struct ProfilerCallback : public physx::PxProfilerCallback
+{
+ PxU64 times[gNumNames];
+
+ ProfilerCallback()
+ {
+ for (int i = 0; i < gNumNames; ++i)
+ times[i] = 0;
+ }
+
+ ~ProfilerCallback()
+ {
+ for (int i = 0; i < gNumNames; ++i)
+ {
+ float ms = SnippetUtils::getElapsedTimeInMilliseconds(times[i]);
+ printf("%s: %f ms\n", gNames[i], PxF64(ms));
+ }
+ }
+
+ virtual void* zoneStart(const char* eventName, bool, uint64_t)
+ {
+ for (int i = 0; i < gNumNames; ++i)
+ {
+ if (!strcmp(gNames[i], eventName))
+ {
+ times[i] -= SnippetUtils::getCurrentTimeCounterValue();
+ break;
+ }
+ }
+ return NULL;
+ }
+ virtual void zoneEnd(void* /*profilerData*/, const char* eventName, bool, uint64_t)
+ {
+ PxU64 time = SnippetUtils::getCurrentTimeCounterValue();
+
+ for (int i = 0; i < gNumNames; ++i)
+ {
+ if (!strcmp(gNames[i], eventName))
+ {
+ times[i] += time;
+ break;
+ }
+ }
+ }
+};
+ProfilerCallback gProfilerCallback;
+
+#define NUM_WORKER_THREADS 1
+
+#define RAYCAST_BATCH_SIZE 1
+
+#define UPDATE_BATCH_SIZE 1
+
+VehicleDesc initVehicleDesc()
+{
+ //Set up the chassis mass, dimensions, moment of inertia, and center of mass offset.
+ //The moment of inertia is just the moment of inertia of a cuboid but modified for easier steering.
+ //Center of mass offset is 0.65m above the base of the chassis and 0.25m towards the front.
+ const PxF32 chassisMass = 1500.0f;
+ const PxVec3 chassisDims(2.5f,2.0f,5.0f);
+ const PxVec3 chassisMOI
+ ((chassisDims.y*chassisDims.y + chassisDims.z*chassisDims.z)*chassisMass/12.0f,
+ (chassisDims.x*chassisDims.x + chassisDims.z*chassisDims.z)*0.8f*chassisMass/12.0f,
+ (chassisDims.x*chassisDims.x + chassisDims.y*chassisDims.y)*chassisMass/12.0f);
+ const PxVec3 chassisCMOffset(0.0f, -chassisDims.y*0.5f + 0.65f, 0.25f);
+
+ //Set up the wheel mass, radius, width, moment of inertia, and number of wheels.
+ //Moment of inertia is just the moment of inertia of a cylinder.
+ const PxF32 wheelMass = 20.0f;
+ const PxF32 wheelRadius = 0.5f;
+ const PxF32 wheelWidth = 0.4f;
+ const PxF32 wheelMOI = 0.5f*wheelMass*wheelRadius*wheelRadius;
+ const PxU32 nbWheels = 6;
+
+ VehicleDesc vehicleDesc;
+
+ vehicleDesc.chassisMass = chassisMass;
+ vehicleDesc.chassisDims = chassisDims;
+ vehicleDesc.chassisMOI = chassisMOI;
+ vehicleDesc.chassisCMOffset = chassisCMOffset;
+ vehicleDesc.chassisMaterial = gMaterial;
+ vehicleDesc.chassisSimFilterData = PxFilterData(COLLISION_FLAG_CHASSIS, COLLISION_FLAG_CHASSIS_AGAINST, 0, 0);
+
+ vehicleDesc.wheelMass = wheelMass;
+ vehicleDesc.wheelRadius = wheelRadius;
+ vehicleDesc.wheelWidth = wheelWidth;
+ vehicleDesc.wheelMOI = wheelMOI;
+ vehicleDesc.numWheels = nbWheels;
+ vehicleDesc.wheelMaterial = gMaterial;
+ vehicleDesc.chassisSimFilterData = PxFilterData(COLLISION_FLAG_WHEEL, COLLISION_FLAG_WHEEL_AGAINST, 0, 0);
+
+ return vehicleDesc;
+}
+
+void initPhysics()
+{
+ /////////////////////////////////////////////
+ //Initialise the sdk and scene
+ /////////////////////////////////////////////
+
+ gFoundation = PxCreateFoundation(PX_FOUNDATION_VERSION, gAllocator, gErrorCallback);
+
+ gPvd = PxCreatePvd(*gFoundation);
+ PxPvdTransport* transport = PxDefaultPvdSocketTransportCreate(PVD_HOST, 5425, 10);
+ gPvd->connect(*transport, gConnectionFlags);
+
+ // PVD sets itself up as the profiler during the "connect" call above. We override this with
+ // our own callback. If we wanted both our profiling and PVD's at the same time, we would
+ // just call the PVD functions (available in PxPvd's base class) from our own profiler callback.
+ PxSetProfilerCallback(&gProfilerCallback);
+
+ gPhysics = PxCreatePhysics(PX_PHYSICS_VERSION, *gFoundation, PxTolerancesScale(), true, gPvd);
+
+ PxSceneDesc sceneDesc(gPhysics->getTolerancesScale());
+ sceneDesc.gravity = PxVec3(0.0f, -9.81f, 0.0f);
+
+ gDispatcher = PxDefaultCpuDispatcherCreate(NUM_WORKER_THREADS);
+
+ sceneDesc.cpuDispatcher = gDispatcher;
+ sceneDesc.filterShader = VehicleFilterShader;
+ gScene = gPhysics->createScene(sceneDesc);
+
+ PxPvdSceneClient* pvdClient = gScene->getScenePvdClient();
+ if(pvdClient)
+ {
+ pvdClient->setScenePvdFlag(PxPvdSceneFlag::eTRANSMIT_CONSTRAINTS, true);
+ pvdClient->setScenePvdFlag(PxPvdSceneFlag::eTRANSMIT_CONTACTS, false);
+ pvdClient->setScenePvdFlag(PxPvdSceneFlag::eTRANSMIT_SCENEQUERIES, false);
+ }
+ gCooking = PxCreateCooking(PX_PHYSICS_VERSION, *gFoundation, PxCookingParams(PxTolerancesScale()));
+
+ /////////////////////////////////////////////
+ //Create a task manager that will be used to
+ //update the vehicles concurrently across
+ //multiple threads.
+ /////////////////////////////////////////////
+
+ gTaskManager = PxTaskManager::createTaskManager(gFoundation->getErrorCallback(), gDispatcher);
+
+ /////////////////////////////////////////////
+ //Initialise the vehicle sdk and create
+ //vehicles that will drive on a plane
+ /////////////////////////////////////////////
+
+ PxInitVehicleSDK(*gPhysics);
+ PxVehicleSetBasisVectors(PxVec3(0,1,0), PxVec3(0,0,1));
+ PxVehicleSetUpdateMode(PxVehicleUpdateMode::eVELOCITY_CHANGE);
+
+ //Create the batched scene queries for the suspension raycasts.
+ gVehicleSceneQueryData = VehicleSceneQueryData::allocate(NUM_VEHICLES, PX_MAX_NB_WHEELS, 1, 1, WheelSceneQueryPreFilterBlocking , NULL, gAllocator);
+ for(PxU32 i = 0; i < NUM_VEHICLES; i++)
+ {
+ gBatchQueries[i] = VehicleSceneQueryData::setUpBatchedSceneQuery(i, *gVehicleSceneQueryData, gScene);
+ }
+
+ //Create the friction table for each combination of tire and surface type.
+ //For simplicity we only have a single surface type.
+ gMaterial = gPhysics->createMaterial(0.5f, 0.5f, 0.6f);
+ gFrictionPairs = createFrictionPairs(gMaterial);
+
+ //Create a plane to drive on.
+ PxFilterData groundPlaneSimFilterData(COLLISION_FLAG_GROUND, COLLISION_FLAG_GROUND_AGAINST, 0, 0);
+ gGroundPlane = createDrivablePlane(groundPlaneSimFilterData, gMaterial, gPhysics);
+ gScene->addActor(*gGroundPlane);
+
+ //Create vehicles that will drive on the plane.
+ for(PxU32 i = 0; i < NUM_VEHICLES; i++)
+ {
+ VehicleDesc vehicleDesc = initVehicleDesc();
+ PxVehicleDrive4W* vehicle = createVehicle4W(vehicleDesc, gPhysics, gCooking);
+ PxTransform startTransform(PxVec3(vehicleDesc.chassisDims.x*3.0f*i, (vehicleDesc.chassisDims.y*0.5f + vehicleDesc.wheelRadius + 1.0f), 0), PxQuat(PxIdentity));
+ vehicle->getRigidDynamicActor()->setGlobalPose(startTransform);
+ gScene->addActor(*vehicle->getRigidDynamicActor());
+
+ //Set the vehicle to rest in first gear.
+ //Set the vehicle to use auto-gears.
+ vehicle->setToRestState();
+ vehicle->mDriveDynData.forceGearChange(PxVehicleGearsData::eFIRST);
+ vehicle->mDriveDynData.setUseAutoGears(true);
+
+ //Set each car to accelerate forwards
+ vehicle->mDriveDynData.setAnalogInput(PxVehicleDrive4WControl::eANALOG_INPUT_ACCEL, 1.0f);
+
+ gVehicles[i] = vehicle;
+ }
+
+ //Set up the wheel query results that are used to query the state of the vehicle after calling PxVehicleUpdates
+ gVehicleWheelQueryResults = VehicleWheelQueryResults::allocate(NUM_VEHICLES, PX_MAX_NB_WHEELS, gAllocator);
+
+ //Set up the data required for concurrent calls to PxVehicleUpdates
+ gVehicleConcurrency = VehicleConcurrency::allocate(NUM_VEHICLES, PX_MAX_NB_WHEELS, gAllocator);
+
+ //Set up the profile zones so that the advantages of parallelism can be measured in pvd.
+}
+
+//TaskVehicleRaycasts allows vehicle suspension raycasts to be performed concurrently across
+//multiple threads.
+class TaskVehicleRaycasts: public PxLightCpuTask
+{
+public:
+
+ TaskVehicleRaycasts()
+ : mThreadId(0xffffffff)
+ {
+ }
+
+ void setThreadId(const PxU32 threadId)
+ {
+ mThreadId = threadId;
+ }
+
+ virtual void run()
+ {
+ PxU32 vehicleId = mThreadId*RAYCAST_BATCH_SIZE;
+ while(vehicleId < NUM_VEHICLES)
+ {
+ const PxU32 numToRaycast = PxMin(NUM_VEHICLES - vehicleId, static_cast<PxU32>(RAYCAST_BATCH_SIZE));
+ for(PxU32 i = 0; i < numToRaycast; i++)
+ {
+ PxVehicleWheels* vehicles[1] = {gVehicles[vehicleId + i]};
+ PxBatchQuery* batchQuery = gBatchQueries[vehicleId + i];
+ const PxU32 raycastQueryResultsSize = gVehicleSceneQueryData->getQueryResultBufferSize();
+ PxRaycastQueryResult* raycastQueryResults = gVehicleSceneQueryData->getRaycastQueryResultBuffer(vehicleId + i);
+ PxVehicleSuspensionRaycasts(batchQuery, 1, vehicles, raycastQueryResultsSize, raycastQueryResults);
+ }
+ vehicleId += NUM_WORKER_THREADS*RAYCAST_BATCH_SIZE;
+ }
+ }
+
+ virtual const char* getName() const { return "TaskVehicleRaycasts"; }
+
+private:
+
+ PxU32 mThreadId;
+};
+
+//TaskVehicleUpdates allows vehicle updates to be performed concurrently across
+//multiple threads.
+class TaskVehicleUpdates: public PxLightCpuTask
+{
+public:
+
+ TaskVehicleUpdates()
+ : PxLightCpuTask(),
+ mTimestep(0),
+ mGravity(PxVec3(0,0,0)),
+ mThreadId(0xffffffff)
+ {
+ }
+
+ void setThreadId(const PxU32 threadId)
+ {
+ mThreadId = threadId;
+ }
+
+ void setTimestep(const PxF32 timestep)
+ {
+ mTimestep = timestep;
+ }
+
+ void setGravity(const PxVec3& gravity)
+ {
+ mGravity = gravity;
+ }
+
+ virtual void run()
+ {
+ PxU32 vehicleId = mThreadId*UPDATE_BATCH_SIZE;
+ while(vehicleId < NUM_VEHICLES)
+ {
+ const PxU32 numToUpdate = PxMin(NUM_VEHICLES - vehicleId, static_cast<PxU32>(UPDATE_BATCH_SIZE));
+ for(PxU32 i = 0; i < numToUpdate; i++)
+ {
+ PxVehicleWheels* vehicles[1] = {gVehicles[vehicleId +i]};
+ PxVehicleWheelQueryResult* vehicleWheelQueryResults = gVehicleWheelQueryResults->getVehicleWheelQueryResults(vehicleId + i);
+ PxVehicleConcurrentUpdateData* concurrentUpdates = gVehicleConcurrency->getVehicleConcurrentUpdate(vehicleId + i);
+ PxVehicleUpdates(mTimestep, mGravity, *gFrictionPairs, 1, vehicles, vehicleWheelQueryResults, concurrentUpdates);
+ }
+ vehicleId += NUM_WORKER_THREADS*UPDATE_BATCH_SIZE;
+ }
+ }
+
+ virtual const char* getName() const { return "TaskVehicleUpdates"; }
+
+private:
+
+ PxF32 mTimestep;
+ PxVec3 mGravity;
+
+ PxU32 mThreadId;
+};
+
+//TaskWait runs after all concurrent raycasts and updates have completed.
+class TaskWait: public PxLightCpuTask
+{
+public:
+
+ TaskWait(SnippetUtils::Sync* syncHandle)
+ : PxLightCpuTask(),
+ mSyncHandle(syncHandle)
+ {
+ }
+
+ virtual void run()
+ {
+ }
+
+ PX_INLINE void release()
+ {
+ PxLightCpuTask::release();
+ SnippetUtils::syncSet(mSyncHandle);
+ }
+
+ virtual const char* getName() const { return "TaskWait"; }
+
+private:
+
+ SnippetUtils::Sync* mSyncHandle;
+};
+
+void concurrentVehicleRaycasts()
+{
+ SnippetUtils::Sync* vehicleRaycastsComplete = SnippetUtils::syncCreate();
+ SnippetUtils::syncReset(vehicleRaycastsComplete);
+
+ //Create tasks that will update the vehicles concurrently then wait until all vehicles
+ //have completed their update.
+ TaskWait taskWait(vehicleRaycastsComplete);
+ TaskVehicleRaycasts taskVehicleRaycasts[NUM_WORKER_THREADS];
+ for(PxU32 i = 0; i < NUM_WORKER_THREADS; i++)
+ {
+ taskVehicleRaycasts[i].setThreadId(i);
+ }
+
+ //Start the task manager.
+ gTaskManager->resetDependencies();
+ gTaskManager->startSimulation();
+
+ //Start the profiler.
+ PX_PROFILE_ZONE("concurrentVehicleRaycasts",0);
+
+ //Update the raycasts concurrently then wait until all vehicles
+ //have completed their raycasts.
+ taskWait.setContinuation(*gTaskManager, NULL);
+ for(PxU32 i = 0; i < NUM_WORKER_THREADS; i++)
+ {
+ taskVehicleRaycasts[i].setContinuation(&taskWait);
+ }
+ taskWait.removeReference();
+ for(PxU32 i = 0; i < NUM_WORKER_THREADS; i++)
+ {
+ taskVehicleRaycasts[i].removeReference();
+ }
+
+ //Wait for the signal that the work has been completed.
+ SnippetUtils::syncWait(vehicleRaycastsComplete);
+
+ //Release the sync handle
+ SnippetUtils::syncRelease(vehicleRaycastsComplete);
+
+}
+
+void concurrentVehicleUpdates(const PxReal timestep)
+{
+ SnippetUtils::Sync* vehicleUpdatesComplete = SnippetUtils::syncCreate();
+ SnippetUtils::syncReset(vehicleUpdatesComplete);
+
+ //Create tasks that will update the vehicles concurrently then wait until all vehicles
+ //have completed their update.
+ TaskWait taskWait(vehicleUpdatesComplete);
+ TaskVehicleUpdates taskVehicleUpdates[NUM_WORKER_THREADS];
+ for(PxU32 i = 0; i < NUM_WORKER_THREADS; i++)
+ {
+ taskVehicleUpdates[i].setThreadId(i);
+ taskVehicleUpdates[i].setTimestep(timestep);
+ taskVehicleUpdates[i].setGravity(gScene->getGravity());
+ }
+
+ //Start the task manager.
+ gTaskManager->resetDependencies();
+ gTaskManager->startSimulation();
+
+ //Start the profiler.
+ {
+ PX_PROFILE_ZONE("concurrentVehicleUpdates",0);
+
+ //Update the vehicles concurrently then wait until all vehicles
+ //have completed their update.
+ taskWait.setContinuation(*gTaskManager, NULL);
+ for(PxU32 i = 0; i < NUM_WORKER_THREADS; i++)
+ {
+ taskVehicleUpdates[i].setContinuation(&taskWait);
+ }
+ taskWait.removeReference();
+ for(PxU32 i = 0; i < NUM_WORKER_THREADS; i++)
+ {
+ taskVehicleUpdates[i].removeReference();
+ }
+
+ //Wait for the signal that the work has been completed.
+ SnippetUtils::syncWait(vehicleUpdatesComplete);
+
+ //Release the sync handle
+ SnippetUtils::syncRelease(vehicleUpdatesComplete);
+
+ //End the profiler
+ }
+
+ //When PxVehicleUpdates is executed concurrently a secondary step is required to complete the
+ //update of the vehicles.
+ PX_PROFILE_ZONE("concurrentVehiclePostUpdates",0);
+ PxVehiclePostUpdates(gVehicleConcurrency->getVehicleConcurrentUpdateBuffer(), NUM_VEHICLES, gVehicles);
+}
+
+
+void stepPhysics()
+{
+ const PxF32 timestep = 1.0f/60.0f;
+
+ //Concurrent vehicle raycasts.
+ concurrentVehicleRaycasts();
+
+ //Concurrent vehicle updates.
+ concurrentVehicleUpdates(timestep);
+
+ //Scene update.
+ PX_PROFILE_ZONE("VehicleStepPhysics",0);
+ gScene->simulate(timestep);
+ gScene->fetchResults(true);
+ }
+
+void cleanupPhysics()
+{
+ //Clean up the vehicles and scene objects
+ gVehicleConcurrency->free(gAllocator);
+ gVehicleWheelQueryResults->free(gAllocator);
+ for(PxU32 i = 0; i < NUM_VEHICLES; i++)
+ {
+ gVehicles[i]->getRigidDynamicActor()->release();
+ static_cast<PxVehicleDrive4W*>(gVehicles[i])->free();
+ }
+ gGroundPlane->release();
+ gFrictionPairs->release();
+ for(PxU32 i = 0; i < NUM_VEHICLES; i++)
+ {
+ gBatchQueries[i]->release();
+ }
+ gVehicleSceneQueryData->free(gAllocator);
+ PxCloseVehicleSDK();
+
+ //Clean up the task manager used for concurrent vehicle updates.
+ gTaskManager->release();
+
+ //Clean up the scene and sdk.
+ gMaterial->release();
+ gCooking->release();
+ gScene->release();
+ gDispatcher->release();
+ gPhysics->release();
+ PxPvdTransport* transport = gPvd->getTransport();
+ gPvd->release();
+ transport->release();
+ gFoundation->release();
+
+ printf("SnippetVehicleMultiThreading done.\n");
+}
+
+int snippetMain(int, const char*const*)
+{
+ printf("Initialising ... \n");
+ initPhysics();
+
+ printf("Simulating ... \n");
+ for(PxU32 i = 0; i < 256; i++)
+ {
+ stepPhysics();
+ }
+
+ cleanupPhysics();
+
+ return 0;
+}