-
Notifications
You must be signed in to change notification settings - Fork 3
Performance
The actual runtime depends on the complexity of the SLI image stack. Especially the number of images in the stack and the number of image pixels can have a big influence. To test the performance, four different SLI image stacks from the coronal vervet brain section (containing 24 images with 2469x3272 pixels each) were analyzed by running the program (generation of all 12 parameter maps with high resolution: --optional --roisize 1), using different thread counts and averaging the number of pixels evaluated per second. In total, 32.314.632 line profiles were evaluated for this performance evaluation. All performance measurements were taken without times for reading and writing files.
Our testing system consists of an AMD Ryzen 3700X at 3.60-4.425 GHz paired with 16 GiB DDR4-3000 memory. Other system configurations might take longer or shorter to compute the parameter maps.
| Thread count | Average pixels per second | Time in minutes for this example (8.078.658 pixels) |
|---|---|---|
| 4 | 4416 | 29:41 |
| 6 | 6232 | 21:36 |
| 8 | 7654 | 17:34 |
| 10 | 8326 | 16:10 |
| 12 | 9384 | 13:54 |
| 16 | 11300 | 11:54 |