RC2 Flatpak Desktop installation - Sub standard Performance ... system HOG !

[aiamuzz]

Hello,

All excited I installed the RC2 Flatpak Desktop app and i found it to be very very poor in performance, it is a huge resource drain on my Core i5 16GB laptop, as i am writing with the desktop app open in the background(dormant) my CPU is at a 99%(85.7% for this app alone) hog and the memory for this app alone is at 17%.

I also have the browser version open on my chrome to run performance comparisons ... attaching the screen grab for the comparison video.

Record_select-area_20260126195421.mp4

1/3 is run on the browser version ... (highly responsive)
Alt & Tab
2/3 is run on the flatpak desktop version (very very poor response)
Alt & Tab
3/3 is again performance on the browser version (highly responsive)

I have only tried to move the 3 rectangles in a fast circular motion and tried zooming in and out as fast as i could.

the result is clearly visible besides the resource hog that it has been.

If there are any diagnostic commands to pipe out to better shed on the inner workings for identifying the issue, please let me know so i can run those commands and post back the outputs.

Cheers.

[0HyperCube]

Hello @aiamuzz,

I get <0.5% CPU for idle. I'd guess that software emulation is being used (instead of the GPU)?

If you launch the flatpak from the command line, you should see a log with the adapter:

Using WGPU adapter: AdapterInfo { name: "llvmpipe (LLVM 21.1.8, 256 bits)", vendor: 65541, device: 0, device_type: Cpu, driver: "llvmpipe", driver_info: "Mesa 25.3.3 (git-a9c6420d8c) (LLVM 21.1.8)", backend: Vulkan }

If it reports "llvmpipe" then it is using software emulation.

I guess it would be useful if you'd be able to paste the logs from running the application?

[aiamuzz]

Hello @aiamuzz,

I get <0.5% CPU for idle. I'd guess that software emulation is being used (instead of the GPU)?

If you launch the flatpak from the command line, you should see a log with the adapter:

Using WGPU adapter: AdapterInfo { name: "llvmpipe (LLVM 21.1.8, 256 bits)", vendor: 65541, device: 0, device_type: Cpu, driver: "llvmpipe", driver_info: "Mesa 25.3.3 (git-a9c6420d8c) (LLVM 21.1.8)", backend: Vulkan }

If it reports "llvmpipe" then it is using software emulation.

@0HyperCube ... you are right its reporting the exact same thing you've shared ...

MESA-INTEL: warning: Ivy Bridge Vulkan support is incomplete

Available WGPU adapters:
[0] Vulkan Cpu (Name: llvmpipe (LLVM 21.1.8, 256 bits), Driver: llvmpipe, Device: 0)

MESA-INTEL: warning: Ivy Bridge Vulkan support is incomplete
Using WGPU adapter: AdapterInfo { name: "llvmpipe (LLVM 21.1.8, 256 bits)", vendor: 65541, device: 0, device_type: Cpu, driver: "llvmpipe", driver_info: "Mesa 25.3.3 (git-a9c6420d8c) (LLVM 21.1.8)", backend: Vulkan }

I guess it would be useful if you'd be able to paste the logs from running the application?

how do i capture the logs ?
please share the path to the log files or the parameters i need to run this application with ... will post the logs

[0HyperCube]

Thanks for the information,

It looks like your vulkan CPU graphics acceleration is not compliant and so would be discarded by WGPU.

I presume you don't have any dedicated GPU?

I guess it should be able to use OpenGL on the CPU graphics? I guess it should be possible to figure out why this is not the case using the trace logs.

You can get trace logs with:

RUST_LOG=trace flatpak run art.graphite.Graphite &> /tmp/log_file.txt

[aiamuzz]

I presume you don't have any dedicated GPU?

glxinfo | grep "Device"
  Device: Mesa Intel(R) HD Graphics 4000 (IVB GT2) (0x166)

lspci -k | grep -iEA3 '3d|vga'
00:02.0 VGA compatible controller: Intel Corporation 3rd Gen Core processor Graphics Controller (rev 09)
        DeviceName: 64
        Subsystem: Hewlett-Packard Company Device 17df
        Kernel driver in use: i915

I guess it should be able to use OpenGL on the CPU graphics? I guess it should be possible to figure out why this is not the case using the trace logs.

You can get trace logs with:

RUST_LOG=trace flatpak run art.graphite.Graphite &> /tmp/log_file.txt

enclosing the log file, ...

In a matter of seconds the log file, a plain text file growing to 37MB ... whoa that's some serious logging

<!-- Failed to upload "graphite_log_file.txt" -->
looks like the file size has exceeded and the attachment has failed ...

@0HyperCube ... Is there some other way I can send you the file, on discord ?

[aiamuzz]

You can get trace logs with:

RUST_LOG=trace flatpak run art.graphite.Graphite &> /tmp/log_file.txt

you can access the message here ...

https://discordapp.com/channels/731730685944922173/874548734304874526/1465449642429911216

[0HyperCube]

Thanks. I was only really interested in the very start where it is finding the adapters. I guess some of the libraries do quite verbose logging.

It seems like the vulkan adapter is indeed being dropped for non-compliance:

gpu_hal::vulkan::adapter Adapter is not Vulkan compliant, hiding adapter: Intel(R) HD Graphics 4000 (IVB GT2)

It looks like there is a bug in wgpu where it fails to create OpenGl context gfx-rs/wgpu#8686 in some situations with BadMatch error:

Instance::new: failed to create Gl backend: InstanceError { message: "unable to create OpenGL or GLES 3.x context", source: Some(BadMatch) }

[aiamuzz]

It looks like there is a bug in wgpu where it fails to create OpenGl context gfx-rs/wgpu#8686 in some situations with BadMatch error:

@0HyperCube ... Is it sufficient information to tinker this bug off ?

[Keavon]

Pardon, what do you mean by "tinker this bug off"? I presume that was either a typo or I might be unfamiliar with that phrasing, but I figured I'd clarify so it's clear to everyone following along :)

[aiamuzz]

tinker this bug off ?

lolz ... if the 'bug' was a dent on the car, tinkering(flatenning) it off(fixing the bug) !!!

I might be unfamiliar with that phrasing,

Hahahah ... It just flowed off my finger to the keyboard on to this issue thread ...