- Do you plan to add an option to export the profiled data into CSV for further custom analysis? I regularly keep track of the CPU timings of my projects in the Excel sheet to control the planned frame budget, and retyping the data is error-prone and takes time.
- Can you improve the data analysis so we can select between Update and FixedUpdate timings? If the frame time is longer, FixedUpdate executes more times, thus falsely indicating that methods there take longer. Could you add an option to analyze FixedUpdate methods separately?
- What are the general plans for Profile Analyzer? What new features and improvements can we expect?
Hi @ProceduralPixels - thank you for the questions!
-
Profile Analyzer already has an option to export the profiled data to CSV here:
Take a look and let us know how you get on! -
This is an interesting idea, thank you for suggesting it. We’re reviewing some different profiling workflows at the moment and will think about FixedUpdate as part of that.
-
We don’t currently have plans to add a lot of new features to Profile Analyzer, but we’re always open to suggestions! We are trying to better integrate it into the main profiler workflow. Personally I’d like to make it easier to open Profile Analyzer and diff captures there; are there any features or improvements you’d particularly like to see (other than the FixedUpdate analysis you’ve already suggested)?
Hi! Thanks for the response. I didn’t notice the Export button. For some reason, I expected that I would be able to change the format when using the “Save” button. My Bad!
I would love to see the option to perform analysis in the FixedUpdate context or in the label invocation context. Right now, the analysis is for the whole frame, so it is not possible to see how a single method invocation changes when it is not called regularly in a frame, and I would love to perform some analysis for a particular label that is not routinely called (or what impact does it have for FixedUpdate)
Right now, I use Total time and invocation count to calculate the average execution time of some labels (to not be bound to a frame context), but it is not enough. I would like to see the median, min, max, and standard deviation calculated for different contexts.
EDIT:
It was unintuitive to use the “depth slice” parameter at first, but after I got used to it, I didn’t think I could come up with a more intuitive idea. So maybe some information under the question mark icon should be enough.
Just understanding the concept that depth slice refers to the hierarchy depth in the CPU profiler was hard for me for some reason. But I think that this graphic illustrates this quite well.
Other than that, I think that this tool is great in its current state. I love it.
Thanks for the response!
Thank you for the feedback about the depth slice, we’ll have a think about that; maybe adding a depth column like in your screenshot might help to make that clearer?
I think this could be the easiest, and the best solution for this problem. I like it.