I'd like to ask a question to understand better how to optimize Starling code.
I worked recently on a way to reimplement iOS UI features using Starling or ND2D (like a scrollview or coverflow). My main issue was to handle the big delays that were caused by textures uploading.
For example, I have a coverflow of 50 pictures thumbnails, stored in an atlas. When I click on one of them, I then want to replace the thumbnail with a high res version of the picture, and zoom into this picture. But on an iPhone 3GS or iPod 4 for example, uploading the high res version causes a significant delay. And in fact, this delay can be felt everywhere, everytime I initialize my screens.
So I decided to give the native GPU mode a go, and to my great suprise, I could achieve the same results with silky smooth movements and almost zero delays when adding elements to the display list (a very few hiccups, but almost unnoticable).
So I was really wondering how the GPU mode was behaving to upload the graphic content so quickly, compared to Starling, for example. Is there some trick in Starling that would allow me to get the same results as with GPU mode?
Any information would be useful. Thank you! 🙂