This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1. (odedgers.com) : http://bit.ly/cuda_blog
2. : - - - CPU -
3. ( ): Sp = 1 / (a + (1 a) / p) p a p = 2 p = 8 p = 1000 a =
0,9 1,05 1,1 1,11 a = 0,5 1,33 1,77 2,0 a = 0,1 1,81 4,71 9,91
7. - host - CPU - device - GPU - GPU - - (grid) - SM -
warp
8. GPU (RW) SM 2-4 GPU (RW) DRAM ~500 GPU (RW) SM 2-4 CPU (RW),
GPU (RW) DRAM ~500 CPU (RW) , GPU (Read-only) DRAM + SM ~500 DRAM
2-4 CPU (RW) , GPU (Read-only) DRAM + SM ~500 DRAM 2-4
9. B (0:0) B (0:1) B (0:2) B (0:3) B (1:0) B (1:1) B (1:2) B
(1:3) B (2:0) B (2:1) B (2:2) B (2:3) B (3:0) B (3:1) B (3:2) B
(3:3) B (4:0) B (4:1) B (4:2) B (4:3) Grid Block T (0:0:0) T
(0:1:0) T (0:2:0) T (1:0:0) T (1:1:0) T (1:2:0) T (2:0:0) T (2:1:0)
T (2:2:0) 20 (5x4), 18 (3x3x2) . 360 .
10. - void - - PCI-Express - warp
11. CUDA Device Host CUDA Driver CUDA Driver API CUDA Runtime
API Libraries Application