Global image processing for preview ? #5371
Labels
scope: performance
Performance issues and improvements
type: enhancement
Something could be better than it currently is
type: question
Something to discuss
By reading issue #4527, I understand that it's quite complex to have a good preview of the processing, because of the dedicated pipeline that only compute the viewed part of the image at the preview's scale.
It has the pros of being faster to compute, but the cons of duplicating the pipeline, having approximate result when some complex tools are used, and make some tools quite difficult to develop, especially if you need the global image or non displayed part (like for Spot Removal tool) to compute the displayed part.
rtengine
has been designed around 2005 I guess, but in 2019, GPU can crunch numbers at outstanding speed using eg. OpenCL (see #1678). If you add clever caching and tiling, you shouldn't be far away of today's performance I guess. Thenrtengine
would be way simpler to refactor with a single pipeline (or at least less than 4 !) and to maintain, with the bonus of guaranteed rendering.So I opened this issue for discussion (only).
@agriggio I'd like to have your insight too.
The text was updated successfully, but these errors were encountered: