-
Somehow this topic always brings out the heavy debates, would branching where the conditional is the same across the entire drawcall be more performant than a lerp? my understanding was that If-s are mainly bad if the conditional evaluates differently within a 2x2 quadrant
Yes you're right. Dynamic branching incurs a performance cost if the conditional diverges across the pixel group (because the GPU processes multiple pixels in parallel). So if you use it essentially as a switch, it'll be fine.
-
-
Alright, glad that stuck with me then
It seems to be a pretty finicky topic based on previous posts I've seen (Maybe cause the severity differs among engines?), it gets hard to gauge which one's correct sometimes -
It's very situational. For example if you wanted to switch between some additional feature you added it's probs cool. You don't want to use a lot of instructions checking the conditional. If you do want to use a divergent branch, then that's the tricky part
Koniec rozmowy
Nowa rozmowa -
-
-
If I understand this right (which I probably don't), this isn't a good use case? This is how I use it a lot, to have a texture mapped over an gradientpic.twitter.com/ioRiDPeDPt
-
Those nodes should compile to roughly this. UE's HLSL assembler will decide whether to flatten the branch or not usually so you should generally trust it - in this case I'm pretty sure it'll be flattened. This will also evaluate to essentially the same as step(x, y)
pic.twitter.com/4xBfakAlhA
Koniec rozmowy
Nowa rozmowa -
Wydaje się, że ładowanie zajmuje dużo czasu.
Twitter jest przeciążony lub wystąpił chwilowy problem. Spróbuj ponownie lub sprawdź status Twittera, aby uzyskać więcej informacji.