Archive for the ‘PhysX SDK’ tag
Quite an interesting beginning of GDC 2017 – NVIDIA has not only presented their newest flagship GPU, GeForce GTX 1080 Ti, but also announced several additions to the GameWorks libraries.
Let’s take a closer look.
FleX & Flow
NVIDIA FleX, unified particle-based solver, and NVIDIA Flow, an engine for simulation of smoke and fire, now both feature hardware agnostic DX12 implementation !
This is exciting news not only for gamers, but also for 3d party companies, already utilizing FleX in their products, such as Lucid Physics from Ephere.
NVIDIA has finally released first public version of the newest PhysX SDK 3.4 (and corresponding APEX 1.4). Please mind the pre-release status, final SDK should be available end of February 2017.
[Update - March 01 2017]: Final release of PhysX SDK 3.4 is now released on GitHub
One of the biggest features of the SDK 3.4 is the introduction of the GPU accelerated rigid body feature (GRB). It supports the entire rigid body pipeline feature-set (except for articulations). The state of GPU rigid bodies can be modified and queried using the exact same API as used for CPU rigid bodies. GPU rigid bodies can interact with clothing and particles in the same way that CPU rigid bodies can and can easily be used in conjunction with character controllers and vehicles.
Leveraging computing capabilities of a graphic processor, GPU accelerated rigid bodies can provide significant performance advantages over CPU simulation in scenes with several thousand active rigid bodies.
At Game Developer Conference 2016 (GDC), NVIDIA has announced the GameWorks 3.1 development kit, which introduces several new physics simulation solutions – PhysX GRB and NVIDIA Flow. Let’s take a look at them more closely:
PhysX GRB is the new GPU accelerated Rigid Body simulation pipeline. It is based on heavily modified branch of PhysX SDK 3.4, but has all the features of the standard SDK and almost identical API. PhysX GRB is currently utilizing CUDA and requires NVIDIA card for GPU acceleration.
Unlike previous implementations, PhysX GRB is featuring hybrid CPU/GPU rigid body solver, and the simulation can be executed either on CPU or GPU with almost no difference in behavior, supported object types or features (GPU articulations are not implemented yet, however).
GRB provides GPU accelerated broad phase, contact generation, constraint solver and body/shape management. In addition, it introduces new implementations of island management and pair management that have been optimized to tolerate the order of magnitude more complex scenes that can be simulated on GPU compared to CPU. New mechanisms to parallelize event notification callbacks and a new feature to lazily update scene query asynchronously are also provided.
Quite astonishing news are coming from Game Developers Conference 2015 – NVIDIA has just announced that latest PhysX SDK 3.3 is now available for free with full source code for Windows, Linux, OS X and Android. Source code release of APEX Destruction and APEX Clothing modules is also planned.
Update: iOS source now added !
Previously, only binary distrubutions of PhysX SDK were available for free for commercial (Windows PC) and non-commercial (Linux, OS X, Android) use. PhysX engine for consoles if still subject of paid licensing.
This desicion will certanly help to push PhysX’s already wide adoption among developers even further and put an additional stress on competitive solutions.
Full instruction on PhysX SDK source code access can be found here.
Update: Latest source code branch is located here – github.com/NVIDIAGameWorks/PhysX-3.3
New PhysX SDK 3.3.2 release is available for download.
Update March 2015: Bug-fixing PhysX SDK 3.3.3 release is available. Release Notes.
Windows, Linux, OS X, Android and iOS builds are now available for free with source code.
Among various bug-fixes, this version also features support for GPU acceleration on Linux platform (CUDA-capable devices only) and reworked documentation.
|PhysX SDK 3.3.2: Release Notes|
- The PhysXCommon/64.dll, nvcuda.dll and PhysXUpdateLoader/64.dll are loaded and checked for the NVIDIA Corporation digital signature. The signature is expected on all NVIDIA Corporation provided dlls. The application will exit if the signature check fails.
- Added the PxDefaultBufferedProfiler extension for simplified SDK profile events extraction.
- PxSceneDesc::sanityBounds allows a bounding box to be set for validating the position coordinates of inserted or updated rigid actors and articulations.
- Linux: Now supports GPU PhysX.
- Added set/getRunProfiled() for PxDefaultCpuDispatcher to control profiling at task level.
- Android: Support for x86 based devices was added.
- PxProfileEventHandler::durationToNanoseconds() added. Translates event duration in timestamp (cycles) into nanoseconds.
- Added SnippetProfileZone to show how to retrieve profiling information.
- Added SnippetCustomJoint to better illustrate custom joint implementation, and removed SnippetExtension.
- Added SnippetStepper to demonstrate kinematic updates while substepping with tasks.
- Significant revision of the user’s guide. Both structure and most content have been modified.
- PxTask::runProfiled() now takes threadId as a parameter.
- The static pruner now issues a performance warning in debug and checked configurations when a tree rebuild occurs and the tree is not empty.
- PxSceneDesc::staticStructure now defaults to PxPruningStructure::eDYNAMIC_AABB_TREE.
- Linux: Switched to shared libraries.
- Profile zone event names changed to match function calls.
- Overlapping read/write errors will now issue a PxErrorCode::eINVALID_OPERATION rather than PxErrorCode::eDEBUG_INFO.
- Improved SnippetToleranceScale to better demonstrate the intended use case.
- Increased 126 characters limit for warnings on unix platforms, 1k limit on all platforms.
- PhysXCommon dll load within PhysX dll now respects dll name. Please see the manual’s PhysXCommon DLL load section.
- Fixed search function of user’s guide.
- Foundation math classes now have in-place arithmetic operators (+= etc).
- Indexing operators taking signed integers in PxVec3, PxVec4, PxMat33, PxMat44, PxStrideIterator have been deprecated.
NVIDIA has released minor 3.3.1 update for PhysX SDK 3.3 branch.
Update: PhysX SDK 3.3.2 released
PhysX SDK 3.3.1 release is mostly focused on various bug-fixes and optimizations.
|PhysX SDK 3.3.1: Release Notes|
- The friction model can now be changed after scene instantiation with PxScene::setFrictionType. The friction model can also be queried with PxScene::getFrictionType.
- PxDefaultSimulationFilterShader now supports particles and cloth as well.
- PxSimulationFilterCallback: the provided actor and shape pointers are now defined as const. Note: this is no behavior change, it was never allowed to write to those objects from within the callback.
- The PxTriangleMeshFlag::eHAS_16BIT_TRIANGLE_INDICES and PxTriangleMeshFlag::eHAS_ADJACENCY_INFO enums have been deprecated. Please use PxTriangleMeshFlag::e16_BIT_INDICES and PxTriangleMeshFlag::eADJACENCY_INFO instead.
- Removed following functions from the API for platforms which do not support CUDA: PxGetSuggestedCudaDeviceOrdinal, PxCreateCudaContextManager, PxLoadPhysxGPUModule.
- PxShape::getMaterialFromInternalFaceIndex will now return a NULL pointer and produce a warning for input faceIndex value of 0xFFFFffff. This change only applies to mesh and heightfield shapes.
- Fixed concurrency issue on windows. Calling PxScene::simulate on multiple scenes concurrently may have caused a deadlock. This only happened if the scenes shared a single PxCpuDispatcher and the dispatcher was configured to use one worker thread only.
For those developers who are not ready yet to switch to 3.3 branch of the PhysX engine, NVIDIA has prepared several bug-fixing releases – PhysX SDK 3.2.5, APEX SDK 1.2.5 and corresponding tools.
Update: APEX SDK 1.3 is available
|PhysX SDK 3.2.5: Release Notes|
- Fixed a bug which could cause triggers to not fire touch events after a kinematic actor moved into or out of overlap and had no target set anymore in the next simulation step.
- Fixed crash when setting multiple materials on a shape followed by release/removal of the corresponding actor from the scene (all while the simulation is running).
- Rigid Bodies
- Fixed a bug that caused bodies to not respond to applied velocities or forces if the autowake parameter in those methods was set to false. This could happen for (almost) resting bodies which were close to other bodies (bounding box overlap) but did not touch them.
- Bugs in the suspension force and tire load computation have been fixed that affected handling when the vehicle was upside down.
- The tire load passed to the tire force computation is now clamped so that it never falls below zero.
- PxVehicleTireLoadFilterData now requires that mMinNormalisedLoad is greater than or equal to zero.
- PxVehicleTireLoadFilterData now has a new member variable mMinFilteredNormalisedLoad. This value describes the filtered normalised load that occurs when the normalised is less than or equal to mMinNormalisedLoad.
- PxVehicleWheelsSimData now has a new function setMinLongSlipDenominator. This can be used to tune stability issues that can arise when the vehicle slows down in the absence of brake and drive torques.
- A bug in the tank damping forces has now been fixed. Tanks now slow down more aggressively from engine and wheel damping forces.
- A new section “The Vehicle Under-steers Then Over-steers” has been added to the vehicle troubleshooting guide to describe steps to avoid twitchy handling on bumpy surfaces.
- A new section “The Vehicle Never Goes Beyond First Gear” has been added to the vehicle troubleshooting guide to describe a common scenario that occurs when the automatic gearbox is given a latency time that is shorter than the time taken to complete a gear change.
- A new section “The Vehicle Slows Down Unnaturally” has been added to the vehicle troubleshooting guide to describe the steps that can be taken to help the vehicle slow down more smoothly.
- The section “SI Units” in the vehicle guide has been updated to include the new functon PxVehicleWheelsSimData::setMinLongSlipDenominator.
Final release of the new PhysX SDK 3.3 is now available for public, after two rounds of Beta Testing.
Update: PhysX SDK 3.3.1 is available
If you are not familiar with SDK 3.3 branch yet, we recommend you to refer Beta-1 and Beta-2 previews for features description, and for “The Evolution of the PhysX SDK” article for the performance improvements (and there are plenty) data.
As compared with PhysX 3.3 Beta-2, final release is including a set of bug-fixes among with minor improvements, and also more detailed documentation.
Full SDK 3.3.0 Release Notes can be viewed here.
About a month ago, NVIDIA has revealed a new unified GPU accelerated physics framework – NVIDIA FLEX – at “The Way It’s Meant To Be Played” press event in Montreal.
Today, Miles Macklin, physics programmer at NVIDIA and lead-developer of the FLEX system, has joined us to share first-hand details about this exciting technology.
PhysXInfo.com: So what is the NVIDIA FLEX exactly ? What are the main features of FLEX ?
Miles Macklin: FLEX is a multi-physics solver for visual effects.
It grew out of the work I did on Position Based Fluids, which was later extended to support two-way coupling between liquids and different object types such as clothing and rigid bodies.
The feature set is largely inspired by tools like Maya’s nCloth and Softimage’s Lagoa. The goal is to bring the capabilities of these off-line applications to real-time games.
NVIDIA has released PhysX SDK 3.3 Beta 2 – first public Beta release of the new major version of the PhysX SDK 3.x
Update: PhysX SDK 3.3.0: Final Release
PhysX SDK 3.3 introduces a broad set of the new features and also presents a vast performance improvement.
|PhysX SDK 3.3 Beta 2 – Release Notes|
- Added PhysXDevice/64.dll to the PC packages. See Windows readme for more details.
- Added support for the NVIDIA Kepler GPU architecture.
- Added support for the Nintendo Wii U console.
- Added support for Windows 8 Modern UI applications (ARM and x86).
- Ported our SIMD library to the ARM NEON architecture.
- Multi Box Pruning (MBP) is offered as an alternative broad phase algorithm to Sweep And Prune (SAP). MBP shows improved performance when all objects are moving or when inserting large numbers of objects. Its generic performance might be inferior to SAP when many objects are sleeping.
- Significant performance and stability optimizations for rigid body solver.
- New function to compute the minimum translational distance and direction to separate two overlapping geometry objects.
- Improved Persisten Contact Manifold (PCM) contact generation mode is now often faster and more robust than the still available legacy path.
- Improved performance of scene queries and contact reports.
- Improved behavior and performance of Continuous Collision Detection (CCD).
- Reduced memory footprint of rigid body classes.
- Added support for sharing shapes among rigid bodies.
- Significantly improved cloth behavior and GPU performance.
- Added support for cloth colliding against itself, other cloth instances, and scene geometry.
- Improved useability of binary and xml serialization.
- Memory can be saved for objects that do not participate in the simulation and are used for scene queries only. For details see the new flag PxActorFlag::eDISABLE_SIMULATION.
Full PhysX SDK 3.3 Beta 2 Release Notes can be viewed here.