search

src-d / kmcuda

Large scale K-means and K-nn implementation on NVIDIA GPU / CUDA
Other
791 stars 145 forks source link

Using more than 2 GPUs #10

Closed igriffer closed 7 years ago

igriffer commented 7 years ago

Hello! I have more than two GPUs and I want to run kmcuda on 3 and 4 devices. How can I do it?

vmarkovtsev commented 7 years ago

The defaults activate all the GPUs. Are there any problems?

igriffer commented 7 years ago

On devices 0 and 1 very heavy processes are started. How does the code behave in this case?

vmarkovtsev commented 7 years ago

device argument gives you the fine control, please see the corresponding docs in README.md

igriffer commented 7 years ago

device CUDA device OR-ed indices - usually 1. For example, 1 means using first device, 2 means second device, 3 means first and second device (2x speedup). Special value 0 enables all available devices. Can I use only 3 or only 4 or only 3 and 4 devices?

vmarkovtsev commented 7 years ago

only 3

0100 = 0x8

only 4

1000 = 0x10

only 3 and 4

1100 = 0x18

igriffer commented 7 years ago

Ok, thank you!