JoltPhysics

JoltPhysics 是一个高性能的物理引擎，专为游戏和实时应用设计。
它提供了丰富的功能，包括碰撞检测、刚体动力学、约束系统等。
JoltPhysics 的设计目标是提供高效且易于使用的物理模拟解决方案。

特点

• 高性能：JoltPhysics 经过优化，能够处理大量物体的物理模拟，适用于大型场景和复杂交互。
• 确定性：引擎提供确定性的物理模拟，确保在相同的输入下每次都能得到相同的结果。
  • Level1:本机确定性
  • Level2:跨平台确定性
• 易于集成：提供了简单的 API，易于集成。
• 跨平台支持：支持多种平台，包括 Windows、Linux 、Android、IOS 和 macOS
• 丰富的功能：包括碰撞检测、刚体动力学、约束系统、粒子系统等
• 开源

Hello World

这里是官方的CPP示例代码，展示了如何使用 JoltPhysics 创建一个简单的物理世界

• ObjectLayerPairFilterTable

  • 首先我们要设置Layer，和Unity类似，新增一些层级，然后设置层级之间的碰撞关系

• BroadPhaseLayerInterfaceTable

  • 这一步是设置物理引擎的BroadPhase层级，BroadPhase层级是物理引擎中用于快速检测碰撞的层级结构

• MaxBodies

  • 设置物理引擎中最大刚体数量

• MaxBodyPairs

  • 设置物理引擎中最大刚体对数量

• MaxContactConstraints

  • 设置物理引擎中最大碰撞对和最大接触约束数量

// Jolt Physics Library (https://github.com/jrouwe/JoltPhysics)
// SPDX-FileCopyrightText: 2025 Jorrit Rouwe
// SPDX-License-Identifier: CC0-1.0
// This file is in the public domain. It serves as an example to start building your own application using Jolt Physics. Feel free to copy paste without attribution!

// The Jolt headers don't include Jolt.h. Always include Jolt.h before including any other Jolt header.
// You can use Jolt.h in your precompiled header to speed up compilation.
#include <Jolt/Jolt.h>

// Jolt includes
#include <Jolt/RegisterTypes.h>
#include <Jolt/Core/Factory.h>
#include <Jolt/Core/TempAllocator.h>
#include <Jolt/Core/JobSystemThreadPool.h>
#include <Jolt/Physics/PhysicsSettings.h>
#include <Jolt/Physics/PhysicsSystem.h>
#include <Jolt/Physics/Collision/Shape/BoxShape.h>
#include <Jolt/Physics/Collision/Shape/SphereShape.h>
#include <Jolt/Physics/Body/BodyCreationSettings.h>
#include <Jolt/Physics/Body/BodyActivationListener.h>

// STL includes
#include <iostream>
#include <cstdarg>
#include <thread>

// Disable common warnings triggered by Jolt, you can use JPH_SUPPRESS_WARNING_PUSH / JPH_SUPPRESS_WARNING_POP to store and restore the warning state
JPH_SUPPRESS_WARNINGS

// All Jolt symbols are in the JPH namespace
using namespace JPH;

// If you want your code to compile using single or double precision write 0.0_r to get a Real value that compiles to double or float depending if JPH_DOUBLE_PRECISION is set or not.
using namespace JPH::literals;

// We're also using STL classes in this example
using namespace std;

// Callback for traces, connect this to your own trace function if you have one
static void TraceImpl(const char *inFMT, ...)
{
	// Format the message
	va_list list;
	va_start(list, inFMT);
	char buffer[1024];
	vsnprintf(buffer, sizeof(buffer), inFMT, list);
	va_end(list);

	// Print to the TTY
	cout << buffer << endl;
}

#ifdef JPH_ENABLE_ASSERTS

// Callback for asserts, connect this to your own assert handler if you have one
static bool AssertFailedImpl(const char *inExpression, const char *inMessage, const char *inFile, uint inLine)
{
	// Print to the TTY
	cout << inFile << ":" << inLine << ": (" << inExpression << ") " << (inMessage != nullptr? inMessage : "") << endl;

	// Breakpoint
	return true;
};

#endif // JPH_ENABLE_ASSERTS

// Layer that objects can be in, determines which other objects it can collide with
// Typically you at least want to have 1 layer for moving bodies and 1 layer for static bodies, but you can have more
// layers if you want. E.g. you could have a layer for high detail collision (which is not used by the physics simulation
// but only if you do collision testing).
namespace Layers
{
	static constexpr ObjectLayer NON_MOVING = 0;
	static constexpr ObjectLayer MOVING = 1;
	static constexpr ObjectLayer NUM_LAYERS = 2;
};

/// Class that determines if two object layers can collide
class ObjectLayerPairFilterImpl : public ObjectLayerPairFilter
{
public:
	virtual bool					ShouldCollide(ObjectLayer inObject1, ObjectLayer inObject2) const override
	{
		switch (inObject1)
		{
		case Layers::NON_MOVING:
			return inObject2 == Layers::MOVING; // Non moving only collides with moving
		case Layers::MOVING:
			return true; // Moving collides with everything
		default:
			JPH_ASSERT(false);
			return false;
		}
	}
};

// Each broadphase layer results in a separate bounding volume tree in the broad phase. You at least want to have
// a layer for non-moving and moving objects to avoid having to update a tree full of static objects every frame.
// You can have a 1-on-1 mapping between object layers and broadphase layers (like in this case) but if you have
// many object layers you'll be creating many broad phase trees, which is not efficient. If you want to fine tune
// your broadphase layers define JPH_TRACK_BROADPHASE_STATS and look at the stats reported on the TTY.
namespace BroadPhaseLayers
{
	static constexpr BroadPhaseLayer NON_MOVING(0);
	static constexpr BroadPhaseLayer MOVING(1);
	static constexpr uint NUM_LAYERS(2);
};

// BroadPhaseLayerInterface implementation
// This defines a mapping between object and broadphase layers.
class BPLayerInterfaceImpl final : public BroadPhaseLayerInterface
{
public:
									BPLayerInterfaceImpl()
	{
		// Create a mapping table from object to broad phase layer
		mObjectToBroadPhase[Layers::NON_MOVING] = BroadPhaseLayers::NON_MOVING;
		mObjectToBroadPhase[Layers::MOVING] = BroadPhaseLayers::MOVING;
	}

	virtual uint					GetNumBroadPhaseLayers() const override
	{
		return BroadPhaseLayers::NUM_LAYERS;
	}

	virtual BroadPhaseLayer			GetBroadPhaseLayer(ObjectLayer inLayer) const override
	{
		JPH_ASSERT(inLayer < Layers::NUM_LAYERS);
		return mObjectToBroadPhase[inLayer];
	}

#if defined(JPH_EXTERNAL_PROFILE) || defined(JPH_PROFILE_ENABLED)
	virtual const char *			GetBroadPhaseLayerName(BroadPhaseLayer inLayer) const override
	{
		switch ((BroadPhaseLayer::Type)inLayer)
		{
		case (BroadPhaseLayer::Type)BroadPhaseLayers::NON_MOVING:	return "NON_MOVING";
		case (BroadPhaseLayer::Type)BroadPhaseLayers::MOVING:		return "MOVING";
		default:													JPH_ASSERT(false); return "INVALID";
		}
	}
#endif // JPH_EXTERNAL_PROFILE || JPH_PROFILE_ENABLED

private:
	BroadPhaseLayer					mObjectToBroadPhase[Layers::NUM_LAYERS];
};

/// Class that determines if an object layer can collide with a broadphase layer
class ObjectVsBroadPhaseLayerFilterImpl : public ObjectVsBroadPhaseLayerFilter
{
public:
	virtual bool				ShouldCollide(ObjectLayer inLayer1, BroadPhaseLayer inLayer2) const override
	{
		switch (inLayer1)
		{
		case Layers::NON_MOVING:
			return inLayer2 == BroadPhaseLayers::MOVING;
		case Layers::MOVING:
			return true;
		default:
			JPH_ASSERT(false);
			return false;
		}
	}
};

// An example contact listener
class MyContactListener : public ContactListener
{
public:
	// See: ContactListener
	virtual ValidateResult	OnContactValidate(const Body &inBody1, const Body &inBody2, RVec3Arg inBaseOffset, const CollideShapeResult &inCollisionResult) override
	{
		cout << "Contact validate callback" << endl;

		// Allows you to ignore a contact before it is created (using layers to not make objects collide is cheaper!)
		return ValidateResult::AcceptAllContactsForThisBodyPair;
	}

	virtual void			OnContactAdded(const Body &inBody1, const Body &inBody2, const ContactManifold &inManifold, ContactSettings &ioSettings) override
	{
		cout << "A contact was added" << endl;
	}

	virtual void			OnContactPersisted(const Body &inBody1, const Body &inBody2, const ContactManifold &inManifold, ContactSettings &ioSettings) override
	{
		cout << "A contact was persisted" << endl;
	}

	virtual void			OnContactRemoved(const SubShapeIDPair &inSubShapePair) override
	{
		cout << "A contact was removed" << endl;
	}
};

// An example activation listener
class MyBodyActivationListener : public BodyActivationListener
{
public:
	virtual void		OnBodyActivated(const BodyID &inBodyID, uint64 inBodyUserData) override
	{
		cout << "A body got activated" << endl;
	}

	virtual void		OnBodyDeactivated(const BodyID &inBodyID, uint64 inBodyUserData) override
	{
		cout << "A body went to sleep" << endl;
	}
};

// Program entry point
int main(int argc, char** argv)
{
	// Register allocation hook. In this example we'll just let Jolt use malloc / free but you can override these if you want (see Memory.h).
	// This needs to be done before any other Jolt function is called.
	RegisterDefaultAllocator();

	// Install trace and assert callbacks
	Trace = TraceImpl;
	JPH_IF_ENABLE_ASSERTS(AssertFailed = AssertFailedImpl;)

	// Create a factory, this class is responsible for creating instances of classes based on their name or hash and is mainly used for deserialization of saved data.
	// It is not directly used in this example but still required.
	Factory::sInstance = new Factory();

	// Register all physics types with the factory and install their collision handlers with the CollisionDispatch class.
	// If you have your own custom shape types you probably need to register their handlers with the CollisionDispatch before calling this function.
	// If you implement your own default material (PhysicsMaterial::sDefault) make sure to initialize it before this function or else this function will create one for you.
	RegisterTypes();

	// We need a temp allocator for temporary allocations during the physics update. We're
	// pre-allocating 10 MB to avoid having to do allocations during the physics update.
	// B.t.w. 10 MB is way too much for this example but it is a typical value you can use.
	// If you don't want to pre-allocate you can also use TempAllocatorMalloc to fall back to
	// malloc / free.
	TempAllocatorImpl temp_allocator(10 * 1024 * 1024);

	// We need a job system that will execute physics jobs on multiple threads. Typically
	// you would implement the JobSystem interface yourself and let Jolt Physics run on top
	// of your own job scheduler. JobSystemThreadPool is an example implementation.
	JobSystemThreadPool job_system(cMaxPhysicsJobs, cMaxPhysicsBarriers, thread::hardware_concurrency() - 1);

	// This is the max amount of rigid bodies that you can add to the physics system. If you try to add more you'll get an error.
	// Note: This value is low because this is a simple test. For a real project use something in the order of 65536.
	const uint cMaxBodies = 1024;

	// This determines how many mutexes to allocate to protect rigid bodies from concurrent access. Set it to 0 for the default settings.
	const uint cNumBodyMutexes = 0;

	// This is the max amount of body pairs that can be queued at any time (the broad phase will detect overlapping
	// body pairs based on their bounding boxes and will insert them into a queue for the narrowphase). If you make this buffer
	// too small the queue will fill up and the broad phase jobs will start to do narrow phase work. This is slightly less efficient.
	// Note: This value is low because this is a simple test. For a real project use something in the order of 65536.
	const uint cMaxBodyPairs = 1024;

	// This is the maximum size of the contact constraint buffer. If more contacts (collisions between bodies) are detected than this
	// number then these contacts will be ignored and bodies will start interpenetrating / fall through the world.
	// Note: This value is low because this is a simple test. For a real project use something in the order of 10240.
	const uint cMaxContactConstraints = 1024;

	// Create mapping table from object layer to broadphase layer
	// Note: As this is an interface, PhysicsSystem will take a reference to this so this instance needs to stay alive!
	// Also have a look at BroadPhaseLayerInterfaceTable or BroadPhaseLayerInterfaceMask for a simpler interface.
	BPLayerInterfaceImpl broad_phase_layer_interface;

	// Create class that filters object vs broadphase layers
	// Note: As this is an interface, PhysicsSystem will take a reference to this so this instance needs to stay alive!
	// Also have a look at ObjectVsBroadPhaseLayerFilterTable or ObjectVsBroadPhaseLayerFilterMask for a simpler interface.
	ObjectVsBroadPhaseLayerFilterImpl object_vs_broadphase_layer_filter;

	// Create class that filters object vs object layers
	// Note: As this is an interface, PhysicsSystem will take a reference to this so this instance needs to stay alive!
	// Also have a look at ObjectLayerPairFilterTable or ObjectLayerPairFilterMask for a simpler interface.
	ObjectLayerPairFilterImpl object_vs_object_layer_filter;

	// Now we can create the actual physics system.
	PhysicsSystem physics_system;
	physics_system.Init(cMaxBodies, cNumBodyMutexes, cMaxBodyPairs, cMaxContactConstraints, broad_phase_layer_interface, object_vs_broadphase_layer_filter, object_vs_object_layer_filter);

	// A body activation listener gets notified when bodies activate and go to sleep
	// Note that this is called from a job so whatever you do here needs to be thread safe.
	// Registering one is entirely optional.
	MyBodyActivationListener body_activation_listener;
	physics_system.SetBodyActivationListener(&body_activation_listener);

	// A contact listener gets notified when bodies (are about to) collide, and when they separate again.
	// Note that this is called from a job so whatever you do here needs to be thread safe.
	// Registering one is entirely optional.
	MyContactListener contact_listener;
	physics_system.SetContactListener(&contact_listener);

	// The main way to interact with the bodies in the physics system is through the body interface. There is a locking and a non-locking
	// variant of this. We're going to use the locking version (even though we're not planning to access bodies from multiple threads)
	BodyInterface &body_interface = physics_system.GetBodyInterface();

	// Next we can create a rigid body to serve as the floor, we make a large box
	// Create the settings for the collision volume (the shape).
	// Note that for simple shapes (like boxes) you can also directly construct a BoxShape.
	BoxShapeSettings floor_shape_settings(Vec3(100.0f, 1.0f, 100.0f));
	floor_shape_settings.SetEmbedded(); // A ref counted object on the stack (base class RefTarget) should be marked as such to prevent it from being freed when its reference count goes to 0.

	// Create the shape
	ShapeSettings::ShapeResult floor_shape_result = floor_shape_settings.Create();
	ShapeRefC floor_shape = floor_shape_result.Get(); // We don't expect an error here, but you can check floor_shape_result for HasError() / GetError()

	// Create the settings for the body itself. Note that here you can also set other properties like the restitution / friction.
	BodyCreationSettings floor_settings(floor_shape, RVec3(0.0_r, -1.0_r, 0.0_r), Quat::sIdentity(), EMotionType::Static, Layers::NON_MOVING);

	// Create the actual rigid body
	Body *floor = body_interface.CreateBody(floor_settings); // Note that if we run out of bodies this can return nullptr

	// Add it to the world
	body_interface.AddBody(floor->GetID(), EActivation::DontActivate);

	// Now create a dynamic body to bounce on the floor
	// Note that this uses the shorthand version of creating and adding a body to the world
	BodyCreationSettings sphere_settings(new SphereShape(0.5f), RVec3(0.0_r, 2.0_r, 0.0_r), Quat::sIdentity(), EMotionType::Dynamic, Layers::MOVING);
	BodyID sphere_id = body_interface.CreateAndAddBody(sphere_settings, EActivation::Activate);

	// Now you can interact with the dynamic body, in this case we're going to give it a velocity.
	// (note that if we had used CreateBody then we could have set the velocity straight on the body before adding it to the physics system)
	body_interface.SetLinearVelocity(sphere_id, Vec3(0.0f, -5.0f, 0.0f));

	// We simulate the physics world in discrete time steps. 60 Hz is a good rate to update the physics system.
	const float cDeltaTime = 1.0f / 60.0f;

	// Optional step: Before starting the physics simulation you can optimize the broad phase. This improves collision detection performance (it's pointless here because we only have 2 bodies).
	// You should definitely not call this every frame or when e.g. streaming in a new level section as it is an expensive operation.
	// Instead insert all new objects in batches instead of 1 at a time to keep the broad phase efficient.
	physics_system.OptimizeBroadPhase();

	// Now we're ready to simulate the body, keep simulating until it goes to sleep
	uint step = 0;
	while (body_interface.IsActive(sphere_id))
	{
		// Next step
		++step;

		// Output current position and velocity of the sphere
		RVec3 position = body_interface.GetCenterOfMassPosition(sphere_id);
		Vec3 velocity = body_interface.GetLinearVelocity(sphere_id);
		cout << "Step " << step << ": Position = (" << position.GetX() << ", " << position.GetY() << ", " << position.GetZ() << "), Velocity = (" << velocity.GetX() << ", " << velocity.GetY() << ", " << velocity.GetZ() << ")" << endl;

		// If you take larger steps than 1 / 60th of a second you need to do multiple collision steps in order to keep the simulation stable. Do 1 collision step per 1 / 60th of a second (round up).
		const int cCollisionSteps = 1;

		// Step the world
		physics_system.Update(cDeltaTime, cCollisionSteps, &temp_allocator, &job_system);
	}

	// Remove the sphere from the physics system. Note that the sphere itself keeps all of its state and can be re-added at any time.
	body_interface.RemoveBody(sphere_id);

	// Destroy the sphere. After this the sphere ID is no longer valid.
	body_interface.DestroyBody(sphere_id);

	// Remove and destroy the floor
	body_interface.RemoveBody(floor->GetID());
	body_interface.DestroyBody(floor->GetID());

	// Unregisters all types with the factory and cleans up the default material
	UnregisterTypes();

	// Destroy the factory
	delete Factory::sInstance;
	Factory::sInstance = nullptr;

	return 0;
}

CSharp 绑定

JoltPhysics 提供了各种语言的绑定，包括 C#、C++、Python 等。

JoltPhysics CSharp这个第三方库提供了 JoltPhysics 的 C# 绑定
使得我们用C#做服务器运行3D物理模拟成为可能，Unity 7 将会使用Core CLR替代Mono，这意味着Unity 7 可以无缝使用Jolt Physics CSharp

JoltPhysics 尝试

最近我尝试在C#服务器运行JoltPhysics，Unity作为渲染客户端，制作了一个简单的足球游戏
https://github.com/NicoIer/GameServer.git