I just wanna say Doc, as a gamer and enthusiast of graphic processing I am really enjoying your lectures. Thanks so much for posting and sharing information with the world. You’re awesome.

Professor: "shadows are hard" Unreal engine 4+: "hold my buffer"

Lecture start 4:42.

@think2086

4 жыл бұрын

@3:54 for true beginning, which includes the intro/review of where we are at as a setup.

"It's a really efficient algorithm, but it can take forever" - 37:36

16:45 starts the part on depth buffers

If we do frustrum clipping in clip space, we just need to compare x with w to know if point is IN or OUT of Right plane. Same rules applied to top/bottom(y vs +/-w), near/far( z vs +/-w) and left plane(x vs -w). And everything in clip space is still linear.

10 Years later we have realtime raytracing with rtx :0 38:30

@robrick9361

3 жыл бұрын

*realtime raytracing for simple geometry

"sometimes people like to roll their own code, which is typically not how we do things anymore" dated 2014 AHAHAHA

Tony Soprano would be curious about the fact there's a lot of clipping happening in the image space :)

shout out to nvidia rtx

"image" space? isn't it called a NDC space? (i could've called it "clip" space instead, if he hadn't specified -1 and +1 limits)

@webgpu

3 жыл бұрын

and this is at 8:50 (also, the previous "truncated pyramid" goes by the name "Frustum" :-) 👍

Hey what paper is he talking about?

@stefanotrettel6410

7 жыл бұрын

I think he is referring to "A characterization of ten hidden-surface algorithms" (Sutherland, Sproull, Schumacker)

Shit goes down 24:36

can somebody explain the divide through w=0 thing?

@Hanokei

3 жыл бұрын

I am not sure how it works exactly but what I thought is: if you consider the 2D visualization of xyzw where xyz represents horizontal axis and w represents vertical axis, you can easily say that any point we can take with w=0 will lay on the xyz plane and will create a vector from origin(v1). I created a vector from v2=(0, 0, 0, 1) which is a normal vector. dot(v1,v2) should be 0, therefore dot((0,0,0,1), (x,y,z,0)) = 0 confirms it. But I am not sure how you can get a different value and you can find that w is negative therefore you can produce a negative result from the dot product. Probably you need to do another z divide from a point that lays on the positive z axis. It would be great if someone could enlighten us.

@joepeters8746

3 жыл бұрын

@@Hanokei What I know: In homogenious Space we need an additional 1 dimension to represent points from Euklidian space. Also the Vector (1,0,0) ist the same as (2,0,0), bec all points that lay on the same line are equal in HS. So when we would devide through w=0, in Euklidian Coordinate system it would go to conceptually infinity, but we cant represent an infinite point, exept through a vector, thats why vectors are represented with w=0 in HS. So we want to eliminate all points that have a negativ w, bc they would get proected filpped.

datees this to 2009

stop saying okay compulsively like a twelve year old girl

@Shivang4

6 жыл бұрын

Too late man he is retired already.

@worlddj1364

6 жыл бұрын

Ken is not teaching anymore?!

@yb801

6 жыл бұрын

You can choose to not watch this

@DrPastah

2 жыл бұрын

okay