May 2014
Intermediate to advanced
522 pages
16h 1m
English
Content preview from GPU Pro 5
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
Start your free trial
5. Non-separable 2D, 3D, and 4D Filtering with CUDA 489
Megavoxels/Second
0
150
100
50
200
250
300
350
400
450
0 20406080100120
Volumes of Size 128 × 128 × 64
Shared
Shared unrolled
FFT
Megavoxels/Second
0
150
100
50
200
250
300
350
400
450
0 20 40 60 80 100 120 140
Volumes of Size 128 × 128 × 64
Shared
Shared unrolled
FFT
Figure 5.11. Performance, measured in megavoxels per second, for the different im-
plementations of 4D filtering and data sizes ranging from 128 × 128 × 64 × 16 to
128 × 128 × 64 × 128. The FFT-based approach clearly outperforms the convolution
approaches. The results for a 7 × 7 × 7 × 7 filter are shown in the upper plot and the
results for a 11 × 11 × 11 × 11 ...