1
1
Version History
2
2
---------------
3
3
4
+ ### New Features in Embree 2.16.3
5
+ - Increased accuracy for handling subdivision surfaces. This fixes
6
+ cracks when using displacement mapping but reduces performance
7
+ at irregular vertices.
8
+ - Fixed a bug where subdivision geometry was not properly updated
9
+ when modifying only the tesselation rate and vertex array.
10
+
4
11
### New Features in Embree 2.16.2
5
12
- Fixed bug that caused NULL intersection context in intersection
6
13
filter when instancing was used.
@@ -25,17 +32,21 @@ Version History
25
32
### New Features in Embree 2.16.0
26
33
- Improved multi-segment motion blur support for scenes with
27
34
different number of time steps per mesh.
28
- - New top level BVH builder that improves build times and BVH quality of two-level BVHs.
35
+ - New top level BVH builder that improves build times and BVH quality
36
+ of two-level BVHs.
29
37
- Added support to enable only a single ISA. Previously code was
30
38
always compiled for SSE2.
31
- - Improved single ray tracing performance for incoherent rays on AVX512 architectures by 5-10%.
32
- - Improved packet/hybrid ray tracing performance for incoherent rays on AVX512 architectures by 10-30%.
33
- - Improved stream ray tracing performance for coherent rays in structure-of-pointers layout by 40-70%.
39
+ - Improved single ray tracing performance for incoherent rays on
40
+ AVX-512 architectures by 5-10%.
41
+ - Improved packet/hybrid ray tracing performance for incoherent rays
42
+ on AVX-512 architectures by 10-30%.
43
+ - Improved stream ray tracing performance for coherent rays in
44
+ structure-of-pointers layout by 40-70%.
34
45
- BVH builder for compact scenes of triangles and quads needs
35
46
essentially no temporary memory anymore. This doubles the
36
47
maximal scene size that can be rendered in compact mode.
37
- - Triangles no longer store the geometry normal in fast/default mode which reduces
38
- memory consumption by up to 20%.
48
+ - Triangles no longer store the geometry normal in fast/default mode
49
+ which reduces memory consumption by up to 20%.
39
50
- Compact mode uses BVH4 now consistently which reduces memory
40
51
consumption by up to 10%.
41
52
- Reduced memory consumption for small scenes (of 10k-100k primitives)
@@ -47,17 +58,19 @@ Version History
47
58
Embree and the application is simplified.
48
59
- Fixed a bug that would have caused a failure of the BVH builder
49
60
for dynamic scenes when run on a machine with more then 1000 threads.
50
- - Fixed a bug that could have been triggered when reaching the maximal number of
51
- mappings under Linux (` vm.max_map_count ` ). This could have happened when creating a
52
- large number of small static scenes.
61
+ - Fixed a bug that could have been triggered when reaching the maximal
62
+ number of mappings under Linux (` vm.max_map_count ` ). This could have
63
+ happened when creating a large number of small static scenes.
53
64
- Added huge page support for Windows and MacOSX (experimental).
54
65
- Added support for Visual Studio 2017.
55
66
- Removed support for Visual Studio 2012.
56
- - Precompiled binaries now require a CPU supporting at least the SSE4.2 ISA.
67
+ - Precompiled binaries now require a CPU supporting at least the
68
+ SSE4.2 ISA.
57
69
- We no longer provide precompiled binaries for 32 bit on Windows.
58
- - Under Windows one now has to use the platform toolset option in CMake to switch
59
- to Clang or the Intel® Compiler.
60
- - Fixed a bug for subdivision meshes when using the incoherent scene flag.
70
+ - Under Windows one now has to use the platform toolset option in
71
+ CMake to switch to Clang or the Intel® Compiler.
72
+ - Fixed a bug for subdivision meshes when using the incoherent scene
73
+ flag.
61
74
- Fixed a bug in the line geometry intersection, that caused reporting
62
75
an invalid line segment intersection with primID -1.
63
76
- Buffer stride for vertex buffers of different time steps of triangle
@@ -114,14 +127,14 @@ Version History
114
127
- Reduced memory consumption when using lots of small dynamic objects.
115
128
- Fixed bug for subdivision surfaces using low tessellation rates.
116
129
- Hair geometry now uses a new ribbon intersector that intersects with
117
- ray-facing quads. The new intersector also returns the
118
- v-coordinate of the hair intersection, and fixes artefacts at junction
119
- points between segments, at the cost of a small performance hit.
130
+ ray-facing quads. The new intersector also returns the v-coordinate
131
+ of the hair intersection, and fixes artefacts at junction points
132
+ between segments, at the cost of a small performance hit.
120
133
- Added ` rtcSetBuffer2 ` function, that additionally gets the number of
121
134
elements of a buffer. In dynamic scenes, this function allows to
122
- quickly change buffer sizes, making it possible to change the number of
123
- primitives of a mesh or the number of crease features for subdivision
124
- surfaces.
135
+ quickly change buffer sizes, making it possible to change the number
136
+ of primitives of a mesh or the number of crease features for
137
+ subdivision surfaces.
125
138
- Added simple 'viewer_anim' tutorial for rendering key
126
139
frame animations and 'buildbench' for measuring BVH (re-)build
127
140
performance for static and dynamic scenes.
@@ -135,8 +148,8 @@ Version History
135
148
- Up to 20% faster BVH build performance on the second generation
136
149
Intel® Xeon Phi™ processor codenamed Knights Landing.
137
150
- Improved quality of the spatial split builder.
138
- - Improved performance for coherent streams of ray packets (SOA layout),
139
- e.g. for fast primary visibility.
151
+ - Improved performance for coherent streams of ray packets (SOA
152
+ layout), e.g. for fast primary visibility.
140
153
- Various bug fixes in tessellation cache, quad-based spatial
141
154
split builder, etc.
142
155
@@ -171,8 +184,8 @@ Version History
171
184
172
185
- Improved performance for streams of coherent (single) rays flagged
173
186
with ` RTC_INTERSECT_COHERENT ` . For such coherent ray streams, e.g.
174
- primary rays, the performance typically improves by 1.3–2× .
175
- - New spatial split BVH builder for triangles, which is 2–6× faster
187
+ primary rays, the performance typically improves by 1.3-2x .
188
+ - New spatial split BVH builder for triangles, which is 2-6x faster
176
189
than the previous version and more memory conservative.
177
190
- Improved performance and scalability of all standard BVH builders on
178
191
systems with large core counts.
@@ -189,12 +202,12 @@ Version History
189
202
- Geometry types can get disabled at compile time.
190
203
- Modified and extended the ray stream API.
191
204
- Added new callback mechanism for the ray stream API.
192
- - Improved ray stream performance (up to 5– 10%).
205
+ - Improved ray stream performance (up to 5- 10%).
193
206
- Up to 20% faster morton builder on machines with large core counts.
194
207
- Lots of optimizations for the second generation Intel® Xeon Phi™
195
208
processor codenamed Knights Landing.
196
209
- Added experimental support for compressed BVH nodes (reduces node
197
- size to 56– 62% of uncompressed size). Compression introduces a
210
+ size to 56- 62% of uncompressed size). Compression introduces a
198
211
typical performance overhead of ~ 10%.
199
212
- Bugfix in backface culling mode. We do now properly cull the
200
213
backfaces and not the frontfaces.
@@ -205,8 +218,8 @@ Version History
205
218
206
219
### New Features in Embree 2.9.0
207
220
208
- - Improved shadow ray performance (10– 100% depending on the scene).
209
- - Added initial support for ray streams (10– 30% higher performance
221
+ - Improved shadow ray performance (10- 100% depending on the scene).
222
+ - Added initial support for ray streams (10- 30% higher performance
210
223
depending on ray coherence in the stream).
211
224
- Added support to calculate second order derivatives using the
212
225
` rtcInterpolate2 ` function.
@@ -227,26 +240,29 @@ Version History
227
240
- Added support for quad geometry (replaces triangle-pairs feature).
228
241
- Added support for linear motion blur of user geometries.
229
242
- Improved performance through AVX-512 optimizations.
230
- - Improved performance of lazy scene build (when using TBB 4.4 update 2).
243
+ - Improved performance of lazy scene build (when using TBB 4.4 update
244
+ 2).
231
245
- Improved performance through huge page support under linux.
232
246
233
247
### New Features in Embree 2.7.1
234
248
235
249
- Internal tasking system supports cancellation of build operations.
236
250
- ISPC mode for robust and compact scenes got significantly faster
237
- (implemented hybrid traversal for bvh4.triangle4v and bvh4.triangle4i).
251
+ (implemented hybrid traversal for bvh4.triangle4v and
252
+ bvh4.triangle4i).
238
253
- Hair rendering got faster as we fixed some issues with the SAH
239
254
heuristic cost factors.
240
255
- BVH8 got slight faster for single ray traversal (improved sorting
241
256
when hitting more than 4 boxes).
242
257
- BVH build performance got up to 30% faster on CPUs with high core
243
258
counts (improved parallel partition code).
244
- - High quality build mode again working properly (spatial splits had been
245
- deactivated in v2.7.0 due to some bug).
259
+ - High quality build mode again working properly (spatial splits had
260
+ been deactivated in v2.7.0 due to some bug).
246
261
- Support for merging two adjacent triangles sharing a common edge
247
262
into a triangle-pair primitive (can reduce memory consumption and
248
263
BVH build times by up to 50% for mostly quad-based input meshes).
249
- - Internal cleanups (reduced number of traversal kernels by more templating)
264
+ - Internal cleanups (reduced number of traversal kernels by more
265
+ templating).
250
266
- Reduced stack size requirements of BVH builders.
251
267
- Fixed crash for dynamic scenes, triggered by deleting all
252
268
geometries from the scene.
@@ -416,29 +432,6 @@ Version History
416
432
- New variant of the SAH-based builder using triangle pre-splits
417
433
(Xeon Phi)
418
434
419
- ### Example Performance Numbers for Embree 2.1
420
-
421
- BVH rebuild performance (including triangle accel generation, excluding
422
- memory allocation) for scenes with 2–12 million triangles:
423
-
424
- - Intel® Core™ i7 (Haswell-based CPU, 4 cores @ 3.0 GHz)
425
- - 7–8 million triangles/s for the SAH-based BVH builder
426
- - 30–36 million triangles/s for the Morton code-based BVH builder
427
- - Intel® Xeon Phi™ 7120
428
- - 37–40 million triangles/s for the SAH-based BVH builder
429
- - 140–160 million triangles/s for the Morton code-based BVH
430
- builder
431
-
432
- Rendering of the Crown model (` crown.ecs ` ) with 4 samples per pixel
433
- (` -spp 4 ` ):
434
-
435
- - Intel® Core™ i7 (Haswell-based CPU, 4 cores CPU @ 3.0 GHz)
436
- - 1024×1024 resolution: 7.8 million rays per sec
437
- - 1920×1080 resolution: 9.9 million rays per sec
438
- - Intel® Xeon Phi™ 7120
439
- - 1024×1024 resolution: 47.1 million rays per sec
440
- - 1920×1080 resolution: 61.1 million rays per sec
441
-
442
435
### New Features in Embree 2.0
443
436
444
437
- Support for the Intel® Xeon Phi™ coprocessor platform
0 commit comments