Need for GPU Compute support (CUDA or OpenCL)

Hi FilterForge team and good evening.

I'm a FilterForge user since version 5 and
many times it has been very helpful for
my 3D texturing projects.

But, especially during the last two years,
I really missed one fundamental feature that now on 2017, can't be postponed anymore.

I'm referring to GPU calculation support, via Cuda
or OpenCL.

I own both a Quadro and a bvery powerful Gtx 1080 Ti
but all that computing power it's totally ignored by FilterForge.

If FilterForge will not support GPU calculation
I think that, for practical reasons, it can't be anymore
one of my toolkit into my software pipeline.

FilterForge it's a very good piece of software,
but I hope that you will seriously put all your efforts
in writing a new version that can fully support GPU computing.

Regards.

Marco.

I don't speak for FF nor am I anybody with specific knowledge. So time and time again this request comes up and gets shot down. The gist of it is that you have algorithms written for how the CPU calculates stuff where GPU would have a very different method of calculating certain things. Hence, this is no happy endeavor.

But the fact remains that GPU is the big star that steals the show. It's incredible seeing development so I can only hope that we'll see something one day. But I doubt it at the rate we're going at. Again, not speaking from internal knowledge or anything, this is just my assertion from past threads.

But hey. The recent version of FF really makes rendering faster in some cases, and it's frankly quite a speed boost.

Generally speaking, there are many algorithms actually better on the CPU than on the GPU. I wouldn't say that GPU is the magic bullet, it's just part of a possible solution to rendering in general. You're not alone wanting the whole thing but I really don't think you understand the large looming developmental side of the story.

This whole discussion also gets propped up on every forum I go to. The Foundry forums gets riddled with this whole GPU thing because we all think it's the future. I don't think anybody's wrong about it, I just straight up think that it's really only part of the bigger solution.

One of the key factors I notice about it is that I think a lot of folks don't seem to really understand the scope of the whole thing, they understand it at an individual level but don't seem to get the idea of the program working on thousands of computers across the world with different specs. I notice with GPUs some people will have a great experience running the program where some folks totally get screwed because they lack the capacity to render.

Oh yeah I use Octane Renderer myself. I think it's stunning. Thea is also very stunning too. I strongly think this is the future of many things.

What I'm really trying to say here is that you're not wrong nor anything. I'm just mentioning the fact that this discussion happens every once in a while on the forums so I thought I'd give some unofficial "long time FF user" perspective of the stuff that's been said so far. So you're totally in the right to throw the request in. I'm just placing in my understanding of the whole deal. Maybe in years time we'll see more development. Who knows.

I absolutely agree with the first post fr om mkdm, and if FF could be much faster thanks to any kind of optimization, either with GPU acceleration, or whatever way it cold be they shurely would not need to offer such high discounts and would earn much more money as many people would be really interested to use FF, specially for professional and business, where time is golden and means money, and they can not spend so much time waiting for the slow preview of FF or the mostly slow render.

The bad news is that it seems that when FF was made ten years ago, was not thought from start to make it possible to be compatible in the future with GPU, and now seems that they can not do nothing "easy" to add this option, and they could convert this for 64 bit, but regrettably I think that there will not be a faster FF version, considering that FF 7 will not have any hardware acceleration, as beta 1 is already about this

I really love FF and would wish all the best for Vladimir and FF company and they could give us a faster FF, specially and mainly on e preview part wh ere you have to customize and test values of the filters, but after 9 years waiting, I do not think that this would be possible, although of course that I wish I could be wrong

Did you see the post in Vlad's Corner "The Next Big Thing"? On March 17, 2017 he writes:

Quote

2. I think I decided on the overall GPU approach. We have separate subsets of nodes / components / functions that will serves as non-Turing-complete "domain specific sub-languages" for handling the GPU. For example, you could have a set of components for CUDA and a separate se t of components for Vulkan. The GPU components won't be Turing-complete, but they will be embedded in a Turing-complete "host filter / program".

Maybe it won't be 'Filter Forge' per se, but a spin off that will have GPU capability it sounds like.

NVIDIA OPENCL IS 1.x AMD is 2.X.

My be multi Threading would best in way of improvement.