[Bf-cycles] New ideas list on developer.blender.org

John Tech johnnytech at live.com
Tue Apr 5 01:48:56 CEST 2016


What do you think?

1. What about donation based feature requests?
A: You post a feature request, it requires a donation just to post it. The advantage is it guarantees someone with authority to add it to the TODO list will actually look at it.(Perhaps if it is KILLED with "Denied", person should get a refund, or wait before accepting fund.)
B. Everyone, and/or perhaps with higher-weight the $donors get to Upvote certain features etc.
C. Everyone can DONATE more towards a specific feature which also is a way to Upvote that feature. If a feature reaches certain funding amount and/or upvotes it is then added to the official TODO list and developers assigned.
D. Consider allowing NEW volunteer developers to start working for and get some sort of payment for good/acceptable work/patches, a kind of purse/bounty type system rewarding code contributions. Someone can see the "Donated" money on a particular feature and decide to work to implement it and send a pull request. Also, I expect that every feature request would keep a percent of the money donated for the foundation/Core developers etc and not entire amount is paid to external/volunteer as the bounty$ for a feature request etc.

I myself would be willing to donate to a system whereby my own needs/wants have a much higher chance of being addressed versus just blindly giving to someone who may implement requests from their other buddies instead on my nickel etc. There is an aversion in open source against "some" financially guided development it seems, and I think its unwarranted. There should be some 'regulation' by some "Core/leadership" but at the same time there should be some "free-market" dynamics allowed also. 

Some things that interest me would be support for both AMD and ATI GPU rendering on the same computer to render even one frame, and ability to use both OpenCL on one card and CUDA on another for example, likely you'd use CUDA on Nvidia and OpenCL on ATI. Especially for complex scenes it would be nice to be able to in addition use network rendering to easily (without time consuming non-crossplatform hacks) render even one frame across multiple hybrid-GPU (each PC with both Nvidia/AMD/ATI gpus) computers on the network. This could perhaps be done based on Tiling or instead based on "seeding" methods or perhaps "render passes" so that different PCs and/or different GPUs render different "passes", z-depth etc...and hand in hand with all this would be:

Adaptive AI(Artificial Intelligence/Machine Learning) Rendering (AAIR)

The Rendering system would learn the best way to partition rendering across its available renderfarm and by subsequent renders it would auto-optimize itself using perhaps a Hybrid-CPU+GPU method to render even on a per frame basis and/or animation frames.
Initially one would perhaps 'help' the algorithms by constraining them to the partitioning scheme you choose such as splitting render-passes or seeding or disabling one-frame support and letting it learn based on simpler previous frames rendering times and optimizing the future frames by giving it to the CPUs if warranted. As the AI/MachineLearning becomes smarter (longer code :() it could then be allowed more autonomy and such constraints would only be optional. I guess a "start" in this direction would be supporting HYBRID CPU+GPU rendering.

The other "BIG"/"HUGE" thing I would say which could become a very "profitable" yet open source initiative would be some sort of realtime PHYSICS/GAME engine that could compete in a sense with the rage of Amazon's Lumberyard/Crytek-Engine, SourceFilmMaker/Source2-Engine, Unity, Cocos2d-x, etc. This should probably be something that an external team very familiar with BLENDER3D may attempt, as it seems to me that Blender3D's original realtime Blender Engine is a dead project more or less, please correct me if I'm wrong. The licensing of an engine should be similar to Cycles, Apache 2, or MIT like Cocos2d-x, definitely not GPL.

In this day and age of the Microsoft Kinect, everyone wants to create their own realtime "movies" and I think realtime engines are the future and we will see more and more flexible movies produced in this manner. We already have decent "Pre-Viz" in various open source packages, the missing ingredient is well integrated Physics forces etc and easy rigging and control to create ones own RPG universe. Create the good quality realtime-Previz and the Physics and I think the rest of the game logic will be the easy part with many contributors.

Anyway, I realize the engine stuff is off-topic here but I thought perhaps the thoughts may give some of the developers here some ideas etc, I know sometimes its hard to have time to contemplate the bigger picture when you're too busy coding.
Best wishes!
----- Original Message -----
From: "Lukas Stockner" <lukas.stockner at freenet.de>
To: "Discussion list to assist Cycles render engine developers" <bf-cycles at blender.org>
Sent: Monday, April 4, 2016 10:25:49 AM
Subject: Re: [Bf-cycles] New ideas list on developer.blender.org


[1] is already close to what I meant, but both ([2] even more so) are
already specific plans about what will be added.
What I mean is more like a general "Hey, we could add this at some
point" (like the Ideas and ToDo in the Wiki) - while [1] is planning for
what is supposed to be added to 2.8x and [2] is only for stuff that is
already WIP.
Of course, it would probably need moderation since lists like these are
bound to attract feature requests from users...

Am 04.04.2016 um 19:18 schrieb Sergey Sharybin:
> Hi,
> There is already task for this i think [1].
> Additionally, thanks to Brecht, we now have tasks nicely organized in
> workboard [2], so think it makes sense to avoid uber-task and have
> individual tasks per official feature we're gonna to work and have those
> tasks listed under "Features and Plans" workboard.
> However, it's still quite technical and it's still nice to work on a
> more higher-level roadmap which would be more clear for users.
> [1] https://developer.blender.org/T46258
> [2] https://developer.blender.org/project/view/26/
> On Mon, Apr 4, 2016 at 7:10 PM, Lukas Stockner
> <lukas.stockner at freenet.de <mailto:lukas.stockner at freenet.de>> wrote:
>     Hi everybody!
>     Since creating a Cycles roadmap came up once again, I'd like to suggest
>     creating an up-to-date ideas list as a first step.
>     There is a ToDo-list on the Wiki (under
>     https://wiki.blender.org/index.php/Dev:Source/Render/Cycles), but it
>     seems to be quite outdated and unmaintained currently.
>     Therefore, my suggestion is to create a "Cycles Ideas" task on
>     developer.blender.org <http://developer.blender.org> where ideas can
>     be added by developers. Once
>     something gets implemented, the corresponding task/revision/commit can
>     then be linked from there. Also, it gives a quick overview on possible
>     and reasonable changes which is a good basis for creating a roadmap (and
>     also could serve as a kind of Cycles-specific Quick Hacks list for new
>     devs by marking some ideas as "easy")
>     So, what do you guys think about it?
>     Best regards,
>     Lukas
>     _______________________________________________
>     Bf-cycles mailing list
>     Bf-cycles at blender.org <mailto:Bf-cycles at blender.org>
>     https://lists.blender.org/mailman/listinfo/bf-cycles
> -- 
> With best regards, Sergey Sharybin
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> https://lists.blender.org/mailman/listinfo/bf-cycles

Bf-cycles mailing list
Bf-cycles at blender.org

More information about the Bf-cycles mailing list