I do some tests and see when use preview size reduced speed is on my System lots faster as when i use 4 pass and a 10 megapixel image that is show complete.
I test with distortions bad trip its a big effect and peview size reduced need only 0.85 sec.
So my idea is, if FF can first scale and pan the image to the current zoom and pan values and put the visible result in a seperate buffer and calc the data first in this buffer is lots faster as use 4 pass or 3 pass, because maybe not so much cache miss happen.
with three pass need to show in best quality in zoom to fit + 1 zoom step to make image bigger in hand stop 1.6 -1.9 sec
with 4 pass need 2 passes to show best quality and need around 2.2 sec.
But when zoom to actual pixels then multi pass draw 3 passes and need 5 sec, to show best quality.
But when only the pixels calc in the visible part buffer, speed is the same 0.85 sec if you zoom large or not.
what do you think about that feature ?
I notice that the Option use all available CPU give same speed on my 4 Core System as use single CPU.
Come the 4 core support later ?
I think when only the visible buffer is render speed is good to do big effects and have a fast preview.
