Cummulus Clouds

[IcePowder]

Simple question two: Why Not?

IcePowder

[neon22]

I suggest you search for "GPU" in the forums. hopefuly all your questions wil be answered

[IcePowder]

Ok, I read the posts. Clouds are one of the last bastions that refuse suspension of disbelief. They are in a computational set that evades all claim to solution. Alvy Ray Smith said "we're going to have to learn how to cheat to solve the cloud problem" back in the '80's.  I had all but given up on seeing a believable solution in my lifetime a decade back.  Now, I'm not so sure that it can't happen.

The 13 Dwarf, parallel execution problem, is yielding promising preliminary results. Upcoming announcements will spotlight an industry change that foreshadows a second wave of applied computational solutions.

It's been quiet too long.

Again, why not?

IcePowder

[Tangled-Universe]

It's not like drawing pixels like in a game, but actually calculating volumes and how they interact with light.
It's complex math which is not always suitable for GPU rendering, since GPU's don't use the same logic and instruction set like CPU's and thus are actually very limited compared to CPU's.
GPU's are extremely fast, but "dumb".
Finding algorithms which are suitable for GPU coding + heavily multithreading (since that's mostly what GPU power is based on) is not an easy task.

If you still find this answer not satisfactory, which is perfectly fine with me, I suggest you continue this specific discussion in the appropriate thread.
Search for "cuda" in the "open discussion" section and you'll find the appropriate thread.

Cheers,
Martin

[IcePowder]

Hi Martin,

You're referring to scan-line rendering.  The next generation are distinct in that they are much more general purpose than the older scan-line based units.  I guess what I'm alluding to is that the next generation will be much more adept at double precision floating point arithmetic vs. the scan-line based polygon shaders that have been the norm.  I work with Nvidia so I'm completely familiar with what's in the pipeline. I'll bow out if that's what you really want...,

IcePowder

[Tangled-Universe]

What do you mean with next-generation?
Are the current GPU's considered next-generation already or is it something in the near/coming future (hence you mention "pipeline")?

We're basically made to believe that TG2 is unsuitable for GPU based calculations (see cuda thread).
If your professional(?) opinion is completely different then I'd like to strongly ecourage you to speak to Planetside and Matt explicitly.

I'm curious about your ideas, so I'm happy to hear from you in that cuda thread or PM
I have a couple dozen of PDF's at home from which I consider some extremely suitable for integration with TG2.

... I suggest you continue this specific discussion in the appropriate thread.
Search for "cuda" in the "open discussion" section and you'll find the appropriate thread.

Let's continue with cumulus clouds here

[TheBadger]

I'm So lost.

[Draigr]

What just transpired was a mild debate between two people who know what they're talking about but refused to go into detail in anything, knowing that the topic was large.


Cliffnotes is essentially:

Icepowder was wondering why GPU rendering wasn't being done, and then revealed that he works with Nvidia, and thus has hinted strongly that the next generation of GPU's might be suitable for Terragen rendering.

TangledUniverse was essentially telling him why Terragen doesn't do GPU rendering currently.

The pdf's and stuff they refer to are technical documents of features almost fully explained.

[Tangled-Universe]

Reduced the setup from 20+ nodes to just 6.

Here and there the shapes and details are very nice.
If you look (not so) carefully it is obvious that the top section has higher density than lower section of the clouds, even though I don't use a density modulator.
It's a known "feature" of the cloud system in Terragen, but I wonder if there's a way to:

1) get a more homogenous type of noise
2) get homogenous density from bottom to top

[rcallicotte]

Any answers about Martin's question about the "seed"?

[TheBadger]

Tangled-Universe,

You are doing just what I was asking about! Great work. I have no doubt that you will be able to workout the remaining difficulties you bring up in your last post. But I would like you to restate your answer to my questions about placement.

Suppose the clouds you are showing in this thread, turn out perfectly, could you also then control there look and placement? For example, look at the following luminist paintings. This is what I have been trying to do.

[Dune]

Great cloud already, Martin. If you PM me a tgd, I'll be happy to dive into it, see what I can come up with.

[Kevin F]

...... Martin. If you PM me a tgd, I'll be happy to dive into it, see what I can come up with.

wouldn't we all!

[Tangled-Universe]

Thanks guys

Tangled-Universe,

You are doing just what I was asking about! Great work. I have no doubt that you will be able to workout the remaining difficulties you bring up in your last post. But I would like you to restate your answer to my questions about placement.

Suppose the clouds you are showing in this thread, turn out perfectly, could you also then control there look and placement? For example, look at the following luminist paintings. This is what I have been trying to do.

Ghehe, I'm glad you have no doubts, but I certainly do

The clouds I created are 100% move-able. I only need to move the localized cloud and everything will move along

Controlling their look is a different matter, depending on what you want. For instance, less/more details, larger/smaller features?
Of course that's all possible, but it will have consequences for the settings which control other aspects for the noise.
So basically you shouldn't tinker too much with it. Changing seeds and moving the noise function within the localized clouds is a better and faster option in many cases.