Hi,
I've noticed significant improvement in update rate of BF+LC in Beta 2 (BF+BF with subsampling still remains sluggish unfortunately). It looks really fast. Mostly in simple scenes, view navigation seems to have 0 lag and almost same framerate as the viewport itself (60+FPS). That's really awesome, but I can see it only slightly bleeding through the old rendered pass because it shows up only in about ever 3-6th pixel.
So since times of standardized IPR are comming, since light cache prepass is intended to be a means of fast feedback for IPR, and since light cache is still crucial to most scenes (adaptive lights do not work without it for example), I was wondering, if we could get alternative visualization of light cache precalculation, that would be subsampled.
So instead of individual pixels slowly dissolving into old rendered pass, they would be subsampled based on on their density to always cover entire image area, and provide us with fast feedback. This, combined with how fast the update rate is, could really end up being very fast and responsive IPR solution. Right now, when I look at how LC is visualized, I see that it progressively covers each n'th pixel of the image, with growing rate. To me, that seems like ideal scenario for subsampling, where surrounding pixels would get color of the nearest rendered neighbour pixel
I've noticed significant improvement in update rate of BF+LC in Beta 2 (BF+BF with subsampling still remains sluggish unfortunately). It looks really fast. Mostly in simple scenes, view navigation seems to have 0 lag and almost same framerate as the viewport itself (60+FPS). That's really awesome, but I can see it only slightly bleeding through the old rendered pass because it shows up only in about ever 3-6th pixel.
So since times of standardized IPR are comming, since light cache prepass is intended to be a means of fast feedback for IPR, and since light cache is still crucial to most scenes (adaptive lights do not work without it for example), I was wondering, if we could get alternative visualization of light cache precalculation, that would be subsampled.
So instead of individual pixels slowly dissolving into old rendered pass, they would be subsampled based on on their density to always cover entire image area, and provide us with fast feedback. This, combined with how fast the update rate is, could really end up being very fast and responsive IPR solution. Right now, when I look at how LC is visualized, I see that it progressively covers each n'th pixel of the image, with growing rate. To me, that seems like ideal scenario for subsampling, where surrounding pixels would get color of the nearest rendered neighbour pixel
Comment