macOS problems

[Vutshi]

Describe the bug
I have two issues on macOS:

1. Purple stripes in fullscreen (in window mode everything is fine):

2. More complicated issue is that martyPC randomly freezes my macOS after approximately half an hour of work. First the system freezes such that only mouse can be moved, after that macOS (I guess) automatically reboots itself. Most likely it is an Apple's problem. Nevertheless, do you have any idea what can go wrong in this case?

Environment (please complete the following information):

• macOS 13.6.7 (22G720)
• 2,3 GHz Quad-Core Intel Core i7
• Intel Iris Plus Graphics
• Rust 1.78.0

Build info

• Built from source fdf1231

[dbalsom]

Thanks for the report. I am not sure why you'd have issues in fullscreen considering I request a borderless window instead of exclusive mode - but maybe that works differently in MacOS. The grey color is drawn by using the GPU's "clear color" - in fullscreen maybe there's an assumption that the entire screen will be drawn so this is turned off? If you toggle between scaling modes in fullscreen, are any artifacts of the previous display size left behind?

The reboot issue is very concerning. Obviously I don't want to damamge anyone's computer! In theory nothing a program could possibly do should be able to cause the operating system to unintentionally reboot, but there are always new bugs to find.

Are you running MartyPC as a normal user or an administrator?

The colors + rebooting makes me concerned a bit about potential GPU issues. Does your system get very hot? Do you have a way to monitor GPU/CPU temperatures?

Do you have a video game or other 3d application you can test to make sure that using the GPU itself isn't an issue?

Could you open Activity Monitor while MartyPC is running and examine the values for CPU % , GPU % and memory usage? Does memory usage continue to climb while MartyPC is running?

[Vutshi]

If you toggle between scaling modes in fullscreen, are any artifacts of the previous display size left behind?

Toggling the scaling modes does not affect the background stripes. Curiously, when switching to fullscreen mode, I see the expected plain gray background for a brief moment before the stripes appear. Maybe it is a bug in the GPU driver...

Are you running MartyPC as a normal user or an administrator?

Normal user.

The colors + rebooting makes me concerned a bit about potential GPU issues. Does your system get very hot? Do you have a way to monitor GPU/CPU temperatures?

It does get warm such that fan kicks in mildly. One cpu core is at 100%. Here is some monitoring data:

I find it peculiar that the cpu load does not drop at all if I pause the emulation in martyPC.

In general, my MacBook is stable enough to run at full CPU load for days (I do some serious computations sometimes). However, it is not completely trouble-free. Occasionally, Mathematica or Firefox seem to be able to crash it, but this happens rarely, maybe a couple of times a year. In contrast, martyPC crashes macOS very consistently. It probably found a weak spot in this last Intel-based MacBook Pro, which Apple doesn't seem to support much anymore…

Do you have a video game or other 3d application you can test to make sure that using the GPU itself isn't an issue?

I can run 3D Switch games in Ryujinx.

Could you open Activity Monitor while MartyPC is running and examine the values for CPU % , GPU % and memory usage? Does memory usage continue to climb while MartyPC is running?

Memory consumption is around 100Mb it grows and drops back from time to time.

Best

[dbalsom]

Can you try building the egui example from here:
https://github.com/parasyte/pixels/tree/main/examples/minimal-egui

and see if you have any of the same issues?

Can you go back to the 0.1.3 release and see if it does the same?

[Vutshi]

Can you try building the egui example from here: https://github.com/parasyte/pixels/tree/main/examples/minimal-egui

and see if you have any of the same issues?

Yes, stripes are there, just red:

I noticed that this simple example of a flying square loads one core to 100%. Is this normal?

Can you go back to the 0.1.3 release and see if it does the same?

Yes, it does the same.

[dbalsom]

If you run the Pixels example for 30+ minutes does it also reboot your system?

[Vutshi]

If you run the Pixels example for 30+ minutes does it also reboot your system?

No, it does not. I ran Pixels all day without issues. The only abnormalities were the CPU load: ~30% when Pixels is in focus and ~120% when minimized to the tray. I'll report this on their GitHub.

Regarding macOS instability, it seems to occur when debug features (CPU state, instruction history, disassembly, etc.) are activated. I'll try playing a DOS game for an hour without these features enabled.

[dbalsom]

can you show me the Performance window in MartyPC?

also, can you see if you have stripes in the wgpu examples (upstream of pixels)
https://github.com/gfx-rs/wgpu

[Vutshi]

can you show me the Performance window in MartyPC?

Here it is:

also, can you see if you have stripes in the wgpu examples (upstream of pixels)

Wgpu shows me the same strips but in a new color :). Should I report the bug to them as well, or is the Pixels bug report sufficient?

[dbalsom]

If the issue is upstream of Pixels i'd say it is more appropriate to open it on wgpu. Pixels maintainer can't fix it

[Vutshi]

Hi @dbalsom,
how can I apply this workaround to martyPC?

[dbalsom]

I suppose I can add a macos_stripe_fix flag to the config and then when we resize for fullscreen subtract one from the resize dimensions.

[Vutshi]

That would be awesome.

[Vutshi]

While testing the fix for #120, I encountered the freeze and reboot again. I've sent another report to Apple :)

[dbalsom]

I bumped the wgpu version in the current branch. Can you check if you still have the same problems?

if you're still seeing the stripes, try the macos_stripe_fix setting.

[Vutshi]

if you're still seeing the stripes, try the macos_stripe_fix setting.

This workaround works!!! Thank you.

[Vutshi]

Hi @dbalsom,

Do you know what MartyPC is doing when emulation is paused? One CPU core is constantly at 100% utilization.

[dbalsom]

that's odd. it shouldn't really do anything different, but obviously it will spend less time per frame emulating, but it should just do updates and yield.

can you show me the performance window when paused?