-
Notifications
You must be signed in to change notification settings - Fork 100
/
release_notes.txt
198 lines (175 loc) · 12.5 KB
/
release_notes.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
1.2.0
------
- Improved collision against moving shapes, solver now correctly interpolates shape transforms for substeps
- Improved Linux support, removed GLUT/GLEW dependencies
- New API for adding soft-joints (dynamic shape-matching constraints) via. the extensions API, see NvFlexExtCreateSoftJoint()
- New API to retrieve particle neighbor information, see NvFlexGetNeighbors()
- New API to support shape collision filtering, collision channels can be assigned to particles and shapes to specify which pairs collide, see NvFlexPhase
- New API to support per-shape plastic deformation, it is now possible to specify plastic creep and threshold coefficient on a per-shape basis (previous global settings), see NvFlexSetRigids()
- New API to selectively disable solver features, this can lead to improved performance, see NvFlexFeatureMode, replaces global FlexParams::fluid option
- New API to allow sub-range particle buffer updates, see NvFlexCopyDesc
- New API for asynchronous compute, see NvFlexInitDesc
- New API change, NvFlexUpdateTriangleMesh() now takes vertices with 16-byte stride to remove CPU synchronization step in D3D implementations
- Fix for NvFlexParams::anisotropyScale, this change improves the fitting of ellipsoids to fluid particle distributions, it should now typically be left at 1.0
- Fix for NvFlexGetTimers() reporting incorrect values on D3D12
- Fix for vsync in D3D12
- Fix for crash when using Flex with NVIDIA driver 384.76
- Fix for unnecessary synchronization when calling NvFlexMap() on a device buffer
- Fix numerical precision for shape-matching constraints with large coordinates
- Fix for uninitialized contact indices returned from NvFlexGetContacts() if NvFlexUpdateSolver() has not been called
- Fix for SDF collision in D3D12
- Fix for free after delete in CUDA when using SDFs
- Fix for D3D warning when using NvFlexSetRigids()
- Add D3D12 rendering support to the demo
- Add support for device NvFlexBuffers on D3D
- Enable NvToolkit profiling zones so to provide annotations in NSight profiles by default, this requires nvToolsExt64_1.dll be present
1.1.0
-----
- New API style, for consistency with other products the API has now an NvFlex prefix and follows a naming convention similar to PhysX
- Add support for DirectX, in addition to CUDA there is now a cross platform DirectX 11 and 12 version of the Flex libraries that Windows applications can link against
- Add support for max acceleration clamping, see NvFlexParams::maxAcceleration, can be useful to reduce popping with fast moving kinematic shapes and large interpenetration
- Add support to querying compute device, see NvFlexGetDeviceName()
- Add support for flushing compute queue, see NvFlexFlush()
- Add support for multiple library instances, NvFlexInit() now returns a library which is bound to a single compute device
- Add support for local space particle simulation, see NvFlexExtMovingFrameInit() and two new local space fluid and cloth demos
- Add support for CUDA 8.0.44
- Remove NvFlexError enum, errors will be reported through the NvFlexErrorCallback passed to NvFlexInit()
- Remove NvFlexMemory enum, users should use the new NvFlexBufferType
- Fix bug in inflatable constraint solver that could cause some shapes to generate NaNs
- Fix for SDF contact generation when the particle lies completely outside volume, fixes a problem with ghost contacts
- Fix for friction being incorrectly applied on trigger shapes
- Fix for multi-phase fluid cohesion not working correctly, re-enable Rayleigh-Taylor instability test
- Fix bug with public projects that referenced internal CUDA paths
- Fix for calling NvFlexSetInflatables() with a count of zero
- Fix for buoyancy parameter
- Fix for bug in BVH builder that could cause missed collisions
- New optional NvFlexDevice library for handling GPU selection and initialization in CUDA
- New buffer based API, all data must now be passed to Flex through FlexBuffers created through NvFlexAllocBuffer()
- New stable shape matching code for more robust and efficient soft bodies, removes need for NvFlexParams::inertiaBias
- New collision shape API supports instancing of convex meshes and a much simplified API, see NvFlexSetShapes()
- Improvements to collision detection for moving shapes, Flex will now perform a CCD check also taking into account the shape trajectory using the prev/current transforms of each shape
1.0.0
-----
- Added support for reporting collision shape ids, and trigger volume shapes, see flexGetContacts()
- Optimizations to host code performance
- Fix for potential memory leak in SDF object destruction
- Fix for potentially missed collisions during convex shape CCD
- Fix for incorrect bounds computation during flexSetShapes() (if not specified by user)
- Fix for initial shape translations being incorrect when using a transform with flexExtCreateInstance()
- Move flexExt.h header to the /include folder
0.9.5
-----
- Fix for NaNs being generated during shape matching with plastic deformation enabled
- Fix for bug that could lead to translational drifting of shape matched rigid bodies
- Fix for incorrect interaction of restitution and friction when using the collision margin parameters
- New geometry instancing API for specifying collision shapes, adds support for transformed triangle meshes and signed distance fields, see flexSetShapes()
- New BVH based collision detection improves performance for scenes with large triangle meshes
- New mesh tearing API, see flexExtTearClothMesh()
- New CCD collision detection for particle versus convex shape, helps reduce tunneling with thin convexes
- Performance improvements for soft body cooking times in flexExtCreateSoftFromMesh()
- Exposed intertial bias for shape matching constraints, see FlexParams::mInertiaBias
- Exposed parameter to control the lifetime of diffuse particles, see FlexParams::mDiffuseLifetime
0.9.0
-----
- Added support for soft bodies, particles can now belong to multiple shape-matching clusters used to drive linear blend skinning
- Added support for particle rest positions that can be used to filter collisions between particles that touch in their rest pose, see flexSetRestPositions()
- Added support for spheres and capsules as collision shapes, renamed flexSetConvexes() to flexSetShapes()
- Added support for user callbacks, these can be used to implement custom constraints, force fields, etc. See flexRegisterSolverCallback()
- Added support for the NVIDIA control panel, Flex will now use the selected PhysX device if desired, see flexInit() for details
- Added support for Android platforms
- Added flexGetParams() method to retrieve default values and set defaults to sensible values
- Improved shape-matching algorithm with better inversion handling
- Optimizations for flexSetTriangles(), users can now pass the triangle bounds to avoid having them re-calculated explicitly
- Rigid constraints now use quaternions as their orientation representation, see flexSetRigids() and flexGetRigidTransforms()
- Exposed anisotropy min/max clamping for better control of particle fluid rendering, see FlexParams::mAnisotropyMin and FlexParams::mAnisotropyMax
- Fix for smooth particles not being generated if anisotropy was not enabled
- Fix for demo mouse picking
- Fix for flexGetContacts() returning incorrect values, added contact visualization to the demo app
- Fix for bug in flexGetBounds() which meant it could return bounds with one frame delay, users should now synchronize the flexGetBounds() call like other query methods
- Fix whitespace on flex.h to play nice with Github's code view
- Fix for triangle mesh collision when particle moves in the plane of the triangle incorrectly reporting contact
- Fix leak when destroying solver
- Fix for incomplete initialization of FlexExtAssets in the extension creation methods
- Rename FlexParams::mMaxVelocity -> FlexParams::mMaxSpeed
0.8.0
-----
- Upgrade to CUDA 7.0
- Linux binaries are now compiled with -fPIC
- Fix for extensions library not writing velocities correctly
- Fix for memory leak when using moving convex shapes
- Optimizations for inflatable constraints to reduce number of kernel launches
- Added option to control the max number of neighbors stored per-particle, this can reduce memory usage, see flexCreateSolver()
0.2.5
-----
- Fix for vertex normal calculation when using sparse active particle set
- Fix for division by zero in rest density calculation when fluid rest distance was 0.0
- Fix for infinite mass particles being moved by kinematic objects
- Added CUDA force fields to extensions library, see flexExtSetForceFields()
- Added particle contact reporting API, see flexGetContacts()
- Added early out to avoid Laplacian smoothing and anisotropy calculation if not using them
- Added global damping parameter to model viscous drag on particles (not just cloth), see FlexParams::mDamping
- Added user manual in pdf and chm formats
- Made enum names consistent, eFlexHost -> eFlexMemoryHost
- Upgraded to CUB 1.3.2
- Extended particle phase parameters to allow fine grained control over self-collision and fluid properties, see flexMakePhase()
0.2.3
-----
- Switched to bindless textures, 50% reduction in CPU usage, because of this change Flex now requires an SM3.0 or above GPU
- Optimized convex grid rasterization by assigning 4 warps/shape, reduces time from 1ms->0.25ms for large shapes
- Added error mesasge callback, see flexInit()
- Added flexSetFence()/flexWaitFence() for more fine grained synchronization with Flex GPU work, flexSynchronize() has been removed
- Added static/dynamic flags for convex shapes to allow prioritising static contacts over dynamic, can be useful to prevent tunnelling through thin shapes
- Added local/global relaxation to improve convergence for some scenes (see FlexParams::eRelaxationMode)
- Removed flexGetVorticities(), allows optimizations for vorticity confinement calculation
- Fix for flexSetSprings(), flexSetTriangles(), and flexSetConvexes() reallocing more often than necessary
- Fix for flexGetDensities(), and flexGetAnisotropy() not returning the full particle array if using sparse active set
- Fix for memory leak when instance creation fails in flexExtCreateInstance()
- Fix for memory leak when setting rigid bodies multiple times
- Fix for potential GPU crash when using plastic deformation on rigid bodies
- Smooth vertex normals are now computed for particles belonging to dynamic triangles (cloth), they can be retrieved via flexGetNormals()
- Optimized CompactObjects in flexExtPushToDevice()
- Made headers compile with /W4 on MSVC
- Added debug capture functionality
0.2.2
-----
- Fix for infinite mass particles gaining velocity during collisions
- Fix bug with planar triangle mesh rasterization
- Fix for zero-length distance constraints
- Fix cloth drag and lift causing instability at high velocities
- Fix for convex grid perf. when there is a single large convex body
- Improved behaviour when particles are overconstrained (constraint averaging), can reduce jitter for pinned cloth
- Improved adhesion behaviour for fluids against solids
- Improved restitution behavior
- Improved fluid rendering performance by using a smaller bilateral filter kernel
- Change particle friction to be an absolute parameter instead of a multiplier on dynamic friction
- Change sort to use async device-device memcpys to avoid stalling host
- Exposed new parameter to control collision margin against shapes see FlexParams::mShapeCollisionMargin
- Warning fixes for Linux and Windows
- Fix z-precision rendering issues in demo
- Added "Ext" prefix to extensions types
- Clamp particle velocity to maximum user velocity before prediction step, improves behaviour when interacting with force fields
0.2.1
-----
- Add support for lower dimensional particle rigid bodies (rods, planes)
- Allow specifying rigid rotations through the solver API
- Allow separate control of fluid and solid rest distances
- Add GUI to demo
0.2.0
-----
- Fix for rest density calculation when number of neighbors > than max
- Add support for async memory copies back to pinned host memory
- Add functions to allocate pinned host memory, see flexAlloc()
- Add rigid body support to FlexExtensions
- Add rigid body SDF collision support
- Add position level friction
- Static triangle grid now created on the GPU, see flexSetTriangles
- Unify triangle collision with rest of the pipline
- Exposed a new parameter to run a second collision pass against triangle meshes, see FlexParam::mEnableCCD
- Optimized rigid body solve using CUB parallel reductions
- Fix for solve springs when some particles have infinite mass
- Reduce allocations in flexSetConvexes
- Fix plastic deformation for rigids
- Switch to /MT for FlexExtensions library
0.1.0
-----
Initial release