Graphics Reference
In-Depth Information
12
Postlin-naive
Postopen
Prelin-naive
Preopen
10
8
6
4
2
0
3
4
5
6
7
8
9
10
AvgDepth
Figure 1.6.
Frame time (ms) versus average depth complexity.
GeForce Titan, driver
320.49, load-factor: 0.5, 2.5M fragments.
70
60
50
40
30
Preopen
Preopen-ec
20
10
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Opacity
Figure 1.7.
Frame time versus opacity for Pre-Open with and without early culling.
Early culling (green) quickly improves performance when opacity increases.
GeForce
Titan, driver 320.49, load-factor: 0.5, 9.8M fragments
1.8 Conclusion
Our tests indicate that Pre-Lin has a significant advantage over other techniques,
while the Open-alloc cell allocation strategy falls behind. This is, however,
not a strong conclusion. Indeed, all of these methods, with the exception of
Post-Lin, are penalized by the emulation of the atomic max 64 bits. More
importantly, the implementation of the Open-alloc techniques currently suffers
from unnecessary atomic operations introduced to avoid race conditions.
The Lin-alloc cell allocation strategy strongly benefits from the dedicated
increment atomic counters. Our tests indicate that without these, the Build
