Textual CPU usage

[nschloe]

When running tiptop and updating every 2 seconds, it eats 7% of CPU:

It's not tiptop's own code that is causing this (which is essentially just a bunch of psutil calls every 2 seconds), but I'm not sure either what in Textual uses so much CPU. Perhaps the constant redraws?

[willmcgugan]

It's a combination of the rendering within the Python code and the work the terminal itself is doing.

Tiptop must be doing some of that work though. Those beautiful braille graph widgets must require some cpu.

The tables are expensive as they were built to be extremely flexible. There's a lot of work done to calculate optimal columns and cell sizes. In the future, Textual will have a dynamic datatable widget which won't have quite so many options, but renders much quicker. Future versions will also get faster as hotspots are optimized.

[nschloe]

Those beautiful braille graph widgets must require some cpu.

It's only a multiline string, and every two seconds, the last column is updated. If anything, it's the string redraws, but I'd say even those are cheap.

The tables are expensive as they were built to be extremely flexible. There's a lot of work done to calculate optimal columns and cell sizes.

I understand. Since the tiptop layout is static, it'd probably help to have updatable table content, instead of having to recreate the table layout on every render.

[willmcgugan]

There are some experimental optimizations in Rich 10.15.0a0 which may make a difference.

[tusharsadhwani]

@willmcgugan try out https://pypi.org/p/yelp-gprof2dot

[willmcgugan]

I would expect a graph visualisation would be way to large to be helpful!

[nschloe]

I wrote tuna for cProfile visualization, but that's not the problem here. It's basically impossible to profile tiptop with

python3 -mcProfile -o program.prof ​/home/nschloe/.local/bin/tiptop

Even after running for one second, it won't exit gracefully. killing is the only option, and that doesn't leave a profile. What's more, after killing every mouse movement produces those dreaded

patterns.

[willmcgugan]

Hmm. Can't imagine why profiling breaks things. I'm afraid I don't have time to investigate at the moment.

[nschloe]

cProfileing on a simpler version shows that most of the runtime is spent in run() -- understandable, that's what it's meant to do.

Not sure how else to profile Textual code.

[muideen1990]

Nico Schlömer cProfileing on a simpler version shows that most of the runtime is spent in run() -- understandable, that's what it's meant to do. Not sure how else to profile Textual code. It look nice [Nishat Aktar - Chat @ Spike](https://spikenow.com/r/a/?ref=spike-organic-signature&_ts=1b3a3s) [1b3a3s] On November 24, 2021 at 22:32 GMT, Nico Schlömer ***@***.***> wrote: cProfileing on a simpler version shows that most of the runtime is spent in run() -- understandable, that's what it's meant to do. [screenshot](https://user-images.githubusercontent.com/181628/143320882-d1178a64-9295-4c63-ac07-67d77440c995.png) Not sure how else to profile Textual code. — You are receiving this because you are subscribed to this thread. Reply to this email directly, [view it on GitHub](#172 (reply in thread)), or [unsubscribe](https://github.com/notifications/unsubscribe-auth/AVFXZBKAFSARSPAJMCHE6VDUNVRYPANCNFSM5IQP4UKQ).