Feature: Diff compare captures

[cugone]

Feature Request

Background

A lot of times I find myself trying to figure out why something is not rendering when it should or rendering when some other unrelated state is active. In order to investigate the issue I take multiple captures and have to manually compare their event browser draw calls and/or renderer state. This takes time because of the loading and unloading of captures and when there are many events in the browser.

Feature

It would be nice to have a feature that could, at the very least, compare two captures' events or draw calls and list what is different between them.

[baldurk]

This has been a long-term wishlist feature for a long time but it would be a substantial undertaking. Given that and the fact that manual comparison is often good enough, I wouldn't expect this to happen until a long time from now as there are much more impactful features to implement sooner.

You can use the existing export features to help. For example on the event browser you can export the list of drawcalls to a txt file in both captures and diff the txt file externally. Or you can do the same with the pipeline state exported to html.

One other thing, you mention 'the loading and unloading of captures' - if you're doing comparison of a before/after kind of thing I would definitely recommend that you open up two copies of RenderDoc. If you haven't saved the captures yet you can still do that with 'open in new instance' on the connection window:

[cugone]

Thanks. Your suggestions to open the captures in different instances or export them will help in the interim. :)

Too bad there's no way to vote on issues, that way developers can see what's popular and potentially prioritize them, especially on mature code bases such as this.

[baldurk]

I'm not sure what you mean 'developers', there's just me :stuck_out_tongue: . I do keep an eye on what features are requested both on github and via other things like email, and it affects what I work on to some degree, but I don't think voting is generally a good idea especially on a pretty niche project. It means the loudest voices drown out others, and by human nature people will ask for everything, or whatever they need right now.

[null77]

+1 vote to this feature request. I'm trying to get a high level view about what changed in the trace from my experiment. Both export as XML and text were not getting me what I wanted. Lots of spurious diffs and the text view lacked other API call info. However there may be Vulkan layers that accomplish what I want so it might not be critical to get this into RDoc.

[T-rvw]

@null77 However there may be Vulkan layers that accomplish what I want so it might not be critical to get this into RDoc. Can you help to give more information about Vulkan layer and RenderDoc? I'm new to them.