c42f31751d
slimmed down the shape interface
2010-10-17 00:08:55 +02:00
1be325172e
merged with Steve's miter hair changes
2010-10-16 16:06:23 +02:00
472eb9926a
tolerate more messages/sec
2010-10-16 16:05:14 +02:00
2004d65444
serialization bugfix
2010-10-16 15:43:34 +02:00
917c142312
MIPMap: be more resilient to infinities
2010-10-16 15:24:47 +02:00
df0600eccd
prevent the user interface from freezing when lots of messages are generated
2010-10-16 14:51:12 +02:00
fbfe3e395d
the octree-class is now completely lock-free (using atomic exchange operations)
2010-10-16 14:10:10 +02:00
2db91b43f9
scale ray differentials to increase texture sharpness when rendering with multiple samples (idea from PBRTv2)
2010-10-16 12:46:36 +02:00
b4f4bf55f7
be lenient when texturing objects without texture coordinates
2010-10-16 12:29:58 +02:00
781006c968
support face normals in the preview
2010-10-16 04:01:21 +02:00
4afd9afb25
more solid OBJ import
2010-10-16 02:15:35 +02:00
d6e0638ceb
bugfixes
2010-10-16 02:03:38 +02:00
d788010885
committed a number of accumulated architectural changes (see body for details)
...
- primitive clipping is now more paranoid about roundoff errors
- significantly improved the kd-tree benchmark "kdbench"
- fixed a parallelization-related bug in the kd-tree builder
- fixed some remaining floating point issues in the min-max binning code
- completely moved TriMesh over to a more compact SoA representation
- the OpenGL renderer now sends smaller packets of geometry to the GPU, which
helps keep the operating system more responsive (NVidia drivers didn't like
the previous behavior much).
2010-10-16 00:33:32 +02:00
24385b655d
nicer loading dialog
2010-10-14 02:53:20 +02:00
a1b8d2266d
copied over coherent RT code from the previous KDTree impl.
2010-10-14 01:00:42 +02:00
01aac828a9
preview bugfix
2010-10-14 00:15:58 +02:00
b2679e8adc
allow to disable the preview
2010-10-13 23:20:20 +02:00
0d085fc2dc
serious bugfixes to the builder
2010-10-13 22:35:15 +02:00
6a173cb9fd
cleanups
2010-10-13 20:56:13 +02:00
7f4d55bdbd
some bugfixes, PBRT-style traversal routine
2010-10-13 19:19:30 +02:00
d9225ae74f
nicer statistics
2010-10-13 12:37:17 +02:00
30d7e45479
rendering using the new tree code now works
2010-10-13 12:28:09 +02:00
3fa45c6cb2
integration of the new kd-tree implementation (no coherent RT yet)
2010-10-13 05:15:52 +02:00
c74924fb7a
cleanups
2010-10-13 03:38:44 +02:00
57c329d646
preparations to support the TriAccel4 format
2010-10-13 00:28:06 +02:00
5feb7753d8
collect some more statistics, code for fitting SAH cost constants to empirical measurements
2010-10-12 23:47:15 +02:00
efb75bbf7b
typo
2010-10-12 17:56:27 +02:00
13878de3fb
added a plain traversal routine, havran-style wins
2010-10-12 17:55:32 +02:00
a5736bd8ba
lots of bugfixes
2010-10-12 14:51:30 +02:00
5e7839b9b7
record SAH costs of the old implementation
2010-10-12 11:36:21 +02:00
82faa961a2
snapshot before searching for remaining traversal/building bug
2010-10-12 03:06:46 +02:00
76abaf1cb8
added a havran-style traversal routine
2010-10-12 02:36:05 +02:00
440656a875
initial generic ray intersection architecture
2010-10-12 01:10:28 +02:00
baa41a195a
cleanups
2010-10-11 21:04:05 +02:00
84cff419f8
build better trees
2010-10-11 14:20:20 -04:00
e0a1720581
compilation fixes
2010-10-11 14:00:14 -04:00
ac4fa10d5d
nicer statistics
2010-10-11 19:48:52 +02:00
d09655dbe6
a few more optimizations after profiling the builder
2010-10-11 19:40:40 +02:00
db444cd87f
parallelization is starting to work
2010-10-11 15:53:59 +02:00
2d5eff417c
better statistics, global indirection table
2010-10-11 14:05:58 +02:00
d36827be4e
do a better job at keeping track of node statistics
2010-10-11 10:09:39 +02:00
fa26a1a271
switched node allocator to a blocked vector
2010-10-11 09:00:14 +02:00
0df829eede
cleanups
2010-10-10 22:38:40 +02:00
bb83ad08cf
fixed the transition from MinMax binning to exact evaluation
2010-10-10 18:18:14 +02:00
a1ca92bda2
do a better job at keeping track of statistics
2010-10-10 18:10:14 +02:00
87b82cc5e1
added the empty space bonus
2010-10-10 13:23:04 +02:00
1058cfe7a7
cleanups
2010-10-10 13:06:42 +02:00
83958d8f91
clipping now also works
2010-10-10 13:06:16 +02:00
679cd18613
exact SAH-optimized kd-tree builder works for the first time (without primitive clipping)
2010-10-10 12:23:55 +02:00
577a732cd0
SAH costs seem more reasonable now
2010-10-10 12:05:32 +02:00
Wenzel Jakob