gpu process time is too large about generic_dbscan HOT 11 CLOSED

@stoneeejianbing
Did you run the following build procedure?

make gpu_apps -j`nproc`
./build/examples/test_pointcloud_clustering_gpu ./data/street_no_ground.pcd 0.7 3

I rerun just now and on my machine it is still giving me 30 ms.

Which GPU are you using btw?

from generic_dbscan.

my cuda version is 11.4
and the run result is this :

number of points: 11619
number of clusters: 271
processing time (cpu): 76.2[ms]

number of points: 11619
number of clusters: 204
processing time (gpu): 1074.5[ms]

my pc nvidia smi : NVIDIA-SMI 470.129.06 Driver Version: 470.129.06 CUDA Version: 11.4

think you

from generic_dbscan.

this is only the driver

NVIDIA-SMI 470.129.06 Driver 

Can you please tell me the hardware gpu you use? I was testing with a geforce titan x

from generic_dbscan.

from generic_dbscan.

this is my pc ,and i also have a xavier,may be i can test code in the xavier?

from generic_dbscan.

I think if you can, you should try this code with Jetson xavier.
The architecture of Jetson is different, so I am not sure if you can build the current code on Jetson though.

from generic_dbscan.

if you are familiar with cuda programming, you can also try tweaking number of threads using here though

https://github.com/xmba15/generic_dbscan/blob/master/cuda/src/DBSCAN.cu#L154-L155

from generic_dbscan.

@stoneeejianbing any update on xavier?

from generic_dbscan.

yes,i have run the code on xavier,and i do not any change on the code. This take 300 ms time use gpu on xavier, and i run the code follow the read me. may be i must set something in the code ?

from generic_dbscan.

@stoneeejianbing I do not have the hardware to test so I am afraid the above advice is all I can provide for now.

from generic_dbscan.

still thank you very much

from generic_dbscan.