Brecht's easter egg surprise: Modernizing shading and rendering

I thaught I would bring this up again, as I know how the people in this thread like to try out the latest features in cycles. But what started as a problem with loseing the path to image textures, has now developed into not either being able to save blends, or reload them if you wish to revert. I think Ideasman42 is working on the problem, but be wary of any builds after 45069. I seems to be connected to the image texture node, but could be wider than that.

edit, sorry, that should have been 45070. I think it was 71 that broke it. It looks like it could be connected to this one.https://projects.blender.org/tracker/index.php?func=detail&aid=30618&group_id=9&atid=498

Brecht must be working hard to complete the features he mentioned in the Sunday Meeting, there hasn’t been any commits from him for two days now.

At the least, it seems by now they have Brecht working on Cycles more than anything else due to project Mango, though it would be nice to find new developers to help him with that work.

That is exactly right. It’s about fundamental intelligent optimaztions in the underlying algorithims that will prove to be the most solid foundation for later efficiency improvement/optimization.

It’s an important difference to note. Someone early mentioned that ‘premature optimzation is the root of all evil’. A cute reference! It’s far worse to spend ages making an O(2^N) algorithim more avoid wasting any possible minor cycle on a processor, than to instead spend 1000 times as long to come up with an O(log n) one to begin with! lol

Turning off “sample as lamp” on the emission shader, (as I would have thought :wink: causes the scene to turn entirely black. Turning it off on the two diffuse shaders makes no difference at all though.

I tried adding a glossy shader, expecting it to break, and it actually still works!

Now I’m even -more- confused as to why this ‘cheat’/optimization works so well.

And, sure, here’s the simple test .blend I made to demonstrate this

http://ever.mine.nu/blender/fast-emit-test2.blend

:wink:

this may be sort of “out there” but I was wondering, if Brecht is listening" is it feasible to move IoR to the transparency node??? for me it just makes more sense…would also make it work better with the glossy node…anyway just my 2cents…or my thoughts…

As far as I know, the transparent node works in a way that avoids some of the calculations done in the glass node, has its own depth setting in the integrator panel, and allows a shading trick for fast, transparent shadows.

As a result the user can use it to create shading shortcuts to make scenes with a lot of transparent surfaces and even some glass surfaces to render a bit faster than it would otherwise, it can also be used by itself as a shortcut for things like windows in cases where an actual refracting surface won’t make a major difference.

So I’m not sure of the feasibility of your request, especially if it is true that the code is a lot different than say, the same code as in the glass shader, but with fewer lines.

it is different, trust me…for instace using velvet with transparency would give an interesting result…just random thought…or glossy node with it would create leaded glass…
I know u can mix glass and glossy, but it would not be the same and less flexible.

Well I wouldn’t think this request would be as simple as copying the IOR code into the transparent node due to the differences I outlined if anyone was thinking that.

What kind of materials are you looking to make that brings about the request, anything you wouldn’t be able to do easily with the glass node or even a setup where the existing transparent node is involved?

Any GTX 680 benchmarks yet?

Early tests are showing a mixed bag of results for compute-performance of GTX 680 2GB. It’s is under-performing hugely in OpenCL, and it can’t even beat the old GTX 580 here, but it’s doing really good in DirectX GPU-compute simulations. The test is here:

http://www.anandtech.com/show/5699/nvidia-geforce-gtx-680-review/17

nVidia may optimizing their now all-important Kepler compiler and improve the compute-performance in the future. But it’s doing really good in some compute area already, and it’s really interesting to see how Cycles is doing now.

There are some used GTX 580 3GB for sale and they are really cheap now, but we can expect a GTX 680 4GB soon, too :slight_smile:

It’s really interesting to see a benchmark of Cycles on GTX 680 . Any?

I just want more control over things like glass, or water etc…they are not the same…using glass/glossy, and transparency with IoR would give much more flexibility to the artist while keeping things still relatively simple…
I like to think of a shader as a reflectance type, refraction index, diffuse, specular model, and a level of transparency…beyond that are other things like sss or a volume shader(could possibly combine a volume and sss into a complex shader I guess…).

AMD release OpenCL beta drivers yesterday. I just wanted to inform you. Maybe that’s usefull for people playing around with compiling OpenCL.

ok this is maybe a crazy question. but would it be possible to add options so that when we make group nodes we can make a new node out of that?

so that for example somone would create a metal group node. and it would have diffuse,reflections IOR.

it’s already like this. you can plug-back the pins you want to control dragging them to the left border. That’s the main feature for nodes as they are now. Also the property panel will reflect the changes, exposing only the controls that you setup. This is how a well-done material library won’t confuse a new user. Beware that nodegroups are going to changed though

either good and bad news here…
goods they are working on it
bad news we may end up getting opencl for 12.4 or beyond…
either way i cant wait

then there is no problem with nodes. the reason some dont want nodes is because they dont want everytime to create 6 nodes to get a normal shader.

i hope from now on everyone with cycles and node experience will tell new users that node groups will support the creation of custom nodes.

I looked into the recent Mac and Win builds to see if the new anisotropic shader is there and I did not see anything.
Are there currently no builds using this on GraphicAll.org?

The beta 12.4 OpenCL drivers are out… cycles does work with it but it has bugs… and you have to manually enable it… see the other AMD driver thread on how to do it.

thanks id already tested this…on an older driver 11.12…idont want to update driver until its officially released
pas maybe there could be an initiative to have all new renderer for opencl only, the back bone for this renderer could be cycles but this orientated around opencl thus freeing up time for brecht to work on cuda cycyles

45173: ColorRamp node
and fix alpha in it in 45176