wgpu seems to ignore the wgpu::Limits config from RenderPlugin about bevy

[bugsweeper]

As described in wgpu::Device::limits

The limits which can be used on this device.

No better limits can be used, even if the underlying adapter can support them.

That is why you should ask your device what best limit you can use, and don't expect device would process unsupportable limits.
I think you could ask that limits by code like:

fn main() {
    let mut app = App::new();
    app.add_plugins(DefaultPlugins);
    if let Some(render_app) = app.get_sub_app_mut(RenderApp) {
        render_app.add_systems(ExtractSchedule, print_limits);
    };
    app.run();
}

fn print_limits(render_device: Res<RenderDevice>) {
    info!("{:?}", render_device.limits());
}

For example I got

Limits { ...
max_compute_workgroup_storage_size: 49152, max_compute_invocations_per_workgroup: 1024, 
max_compute_workgroup_size_x: 1024, max_compute_workgroup_size_y: 1024, 
max_compute_workgroup_size_z: 64, max_compute_workgroups_per_dimension: 65535
...}

For your case key value is max_compute_invocations_per_workgroup
Then you can dynamicaly use discovered limits in runtime with dispatch_workgroups

[bugsweeper]

1. why is it so low for this case?

I looked at source, looks like it is result from DirectX API call "D3D12CreateDevice" which fills limits. I can't debug it for you to figure out better because i'm linux user. I hope you will find out and write results of investigation here.

[00alia00]

Thanks for looking at this!

Maybe I am misunderstanding, but in every other engine (unity/unreal/various c# engines) on this hardware ive used 32,32,1 its been fine. I guess the question then becomes..

1. why is it so low for this case?
2. where are these limits from if not the underlying hardware?
3. whats the point of being able to specifiy the limits if its ignored?

I'll try printing it out when I get a few mins.