Question about the cpu consume

[chengwei0427]

Hi,Dr Xu Thank you for the amazing work，I have a problem about the program cpu. *my computer is Ubuntu16.04 : intel core-i7-10700F@2.9GHz6,GeForce GTX 1660Ti/PCle/SSE2 I have a vlp16, and test the data I recorded use the mapping_velodyne.launch with the default parameters, I find that, the data processing frequency is lower than 10Hz[Not real time**] in the first few minutes, and the CPU usage is also high. Is this normal?

thanks.

[XW-HKU]

That's really abnormal, could you share me a piece of data?

I just test HDL-32E (32 line) datasets on my pc (intel i9-10980HK CPU @ 2.40GHz, 16G RAM), the average processing time is 21ms with Rviz opened. Actually I tested on a Intel i7-8550U 1.9 GHz, 8G RAM before with HDL-32E, even with Rviz open, the average processing time is around 30ms. So I think the over 100ms in VLP-16 dataset is pretty abnormal, maybe something wrong happened.

[XW-HKU]

What happens if you set scan_publish_en in velodyne.yaml to 0?

And you can try the avia datas in the readme to check the processing time for a avia scan. Its real-time performance should be guaranteed if everything normal.

[chengwei0427]

Hi,Dr Xu Thanks for your reply. I test a new data set today, the cpu is still very high(about 160%). What happens if you set scan_publish_en in velodyne.yaml to 0? The cpu is about 150% after set scan_publish_en and dense_publish_enable to 0. And you can try the avia datas in the readme to check the processing time for a avia scan. I can't access Google because of the gfw.

thanks!

[XW-HKU]

Hi,Dr Xu Thanks for your reply. I test a new data set today, the cpu is still very high(about 160%). What happens if you set scan_publish_en in velodyne.yaml to 0? The cpu is about 150% after set scan_publish_en and dense_publish_enable to 0. And you can try the avia datas in the readme to check the processing time for a avia scan. I can't access Google because of the gfw. Here is my test dataset. 链接：https://pan.baidu.com/s/1epdYQ7_PmyQiVeD4Iw9iew 提取码：b4qt

thanks!

I am downloading the bag.

Maybe the cpu consume rate shown in top commend is not important since 100% stands for fully using only one core and you have 6 cpu core. You can use sudo htop to check each CPU core's usages. The more important thing is the real-time performence, the easist way to varify it is stopping the bag play too see if the fast-lio update is stop in the same time.

[XW-HKU]

Hi,Dr Xu Thanks for your reply. I test a new data set today, the cpu is still very high(about 160%). What happens if you set scan_publish_en in velodyne.yaml to 0? The cpu is about 150% after set scan_publish_en and dense_publish_enable to 0. And you can try the avia datas in the readme to check the processing time for a avia scan. I can't access Google because of the gfw.

thanks!

I already downloaded it.

And seems it consumes smaller than 10 ms in my computer (intel i9-10980HK CPU @ 2.40GHz, 16G RAM). The ave total stands for the average processing time for a LiDAR scan.

Could you show me the output after the commend uname -p in your computer?

[chengwei0427]

Hi,Dr Xu Thanks for your reply. It's x86_64 after the commend uname -p; the dependency is Eigen:3.2.92,pcl:1.9 Maybe something wrong with my dependency?

[XW-HKU]

Hi,Dr Xu Thanks for your reply. It's x86_64 after the commend uname -p; the dependency is Eigen:3.2.92,pcl:1.9 Maybe something wrong with my dependency?

May be you can try to upgrade eigen to latest. looks like the processing time increase is relative to the matrix computation.

[chengwei0427]

Hi,Dr Xu The cpu consume is about 40% after update Eigen version[Eigen 3.3.6], The cpu range is probably 35% to 52%. Is this normal? Thank you very much.

[XW-HKU]

That looks very normal, and how about the processing time?

[chengwei0427]

The average processing time is around 10ms, it's amazing!

[XW-HKU]

Thanks, that's great

[chengwei0427]

Thank you for your patient reply, I close the issue as the questions are solved.

Thanks again