Performance issues when plotting userscript topics from live data

[rg-gravis]

Foxglove studio uses very large amounts of CPU when plotting userscript topics live. On a Lenovo P1 Gen 6 (13th Gen Intel Core i9-13900H) running Ubuntu 20.04, Foxglove studio uses 2.5 to 3.5 cores for local or cloud data, and 5 to 6.5 for live data, with the same data and layout. Memory usage stays consistent between sources, at about 1.4-1.6 GiB. When not plotting a userscript topic, the CPU usage is between 1 to 2.5 cores, for any data source.
The script is attached here, /joint_commands is published at 100 Hz and /joint_states is published at about 1000 Hz, and the length of both iterated arrays is 6.
I am on Foxglove Studio version 2.19.1.

[jtbandes]

Internal tracking issue: https://linear.app/foxglove/issue/FG-10047

[defunctzombie]

What are the types of /joint_commands and /joint_states topics? Can you share some recorded data that we can replay to reproduce?

[nthiad]

I couldn't replicate this slowdown on debian trixie with the foxglove studio 2.23.0 deb nor on web with live data. I got some typescript errors from the userscript on 2.19.1 deb. I'm also on a thinkpad (x1 carbon) with a 13th gen intel cpu.

Are you using the deb or the snap for studio? We've had some performance issues with the snap in particular.

What message encoding are you using? For my live test data I used protobuf but I wonder if the issue could show up in the other encodings.

If you have any other data such as a layout file or the data itself that could also help me to replicate this. Thanks!

[nthiad]

Friendly ping! @rg-gravis are you still encountering this performance bug? If so, more info would really help. Thanks.

[ghost]

@rg-gravis is no longer working with us, I'll take this over. I'll reproduce next time I have the opportunity and get back to you.