Add more GPU tracing spans

[Lampan-git]

Objective

• The RenderQueue in my tracy profiles looked a little empty

Solution

• Add either time_span(like smaa) or pass_span(like main_opaque_pass_3d_node), which I copied to the missing places.
• Because of point lights having 6 shadow views and spotlight having 1, so to not scale it per light, add the "auxiliary_views"-span before and after looping through all the RootView::Auxiliary, where shared_shadow_pass is "called" with run_schedule.

Testing

• Ran my game with tracy with this branch
• Ran the examples fullscreen_material, many_lights, volumetric_fog with tracy

Showcase

volumentric_fog example's new profile:

Click to view showcase

AI Disclosure

No code was written with AI. I used Claude Opus to find where spans were missing, reviewing and some explaining.

[Lampan-git]

I want to point out that auxiliary_views hides uniform_allocation, bin_unpacking and early_mesh_preprocessing, but if you zoom in you can see em.

[pcwalton]

Works for me as long as things continue to work on WebGL 2/WebGPU. Thank you!

[JMS55]

Can you explain auxiliary_views more? I don't think I understand.

[Lampan-git]

Can you explain auxiliary_views more? I don't think I understand.

In camera_driver it loops through all views. Since there are 6 view per point lights and 1 per spotlight for shadows this gets scaled with the amount of lights we have. In camera_driver it first goes through the Auxiliary views and then the Camera views are chained(for Cameras there is the per_view_shadow_pass, which then scales per Camera instead). This loops then runs the corresponding schedule per view. For Auxiliary this includes shared_shadow_pass. So to not create a span per shared_shadow_pass, I then put a time span before the loop, and the end it when reaching the Camera view part of the loop.

[JMS55]

Gotcha. I think auxiliary_views might be a bit confusing for users, but I'm willing to merge this as-is. People can always use e.g. nsight for more detailed info - timestamp queries are mainly meant for in-game overlays (I do this in Solari), and for quick eyeballing in Tracy when looking at CPU vs GPU bottlenecks.