💻 CUDA, C++, CMake
-
CPU and GPU (Cuda) implementation of binary opening/closing
-
Library Needed: Opencv
-
Run CPU: src/ Cmake ./gpgpu path_src
-
Run GPU: cuda/ Makefile ./main path_src name_dst
-
Test images: build/
Result: (1000x1000) image with 5x5 kernel
Benchmark:
Kernel size: 5x5 - i7-8750H 6 coeurs/12 threads - Fréquence 2,20 GHz - 4,10 GHz et un GPU Nvidia 1050Ti 4Go 768 CUDA Cores
| Center-aligned | Center-aligned | Center-aligned | Center-aligned |
|---|---|---|---|
| 128x128 | 32 ms | 0.06 ms | 0.056 ms |
| 256x256 | 37 ms | 0.099 ms | 0.1 ms |
| 600x600 | 83 ms | 0.359 ms | 0.386 ms |
| 800x800 | 140 ms | 1.074 ms | 0.989 ms |
| 1000x1000 | 184 ms | 1.383 ms | 1.11 ms |
| 4k | 1278 ms | 32.705 ms | 17.26 ms |
| 8k | 4848 ms | 123.230 ms | 62.111 ms |