*In the other thread you were talking about building blocks your team builds and stuff TDs can build. Does that implies that what is built in the core is more eficient then the scripted code built by TDs? Say if someone writes a little fluid simulator or raytracer ontop of Fabric, through scripting, will it not be as efficient as it could?
No - the Fabric Core is a multi-threaded execution engine, it just takes KL code and makes it very fast. All functionality beyond that - scenegraph, geometry etc is written with Python and KL. For example, the subdivision surfaces work (currently around 35X faster than subDs in Maya, and we'll do more work on this later in the year - expect another jump in performance) is done entirely in KL (some work is done using vertex shaders to get the GPU doing some of the work).
The important element to understand is this: KL works like a scripted language for authoring purposes, but it's a dynamically compiled language. So anything written in KL will run as compiled code - it's just that we're handling the low-level parts of writing high-performance code.
To give you an idea of where we're heading with this - here's some early work on compiling KL code to run on the GPU.
http://fabricengine.com/2012/07/gpu_com ... y_preview/. This is extremely powerful - KL code compiling direct to run on the GPU - no need to write CUDA or OpenCL, it's exactly the same code being compiled on CPU or GPU.
Our goal from day 1 - high-performance coupled with a framework that a Python guy can use. Obviously, as you step away from our existing modules and extensions, and start building your own stuff, you need a more traditional CS background - but right now a TD with zero C++ knowledge can build something extremely powerful using Creation.
*In the beggining you guys were talking about having about Fabric everywhere, through browsers. Considering the target audience, the current focus on QT does look very appealing, but I dont see you guys talking about HTML anymore. Is that still a goal?
Not right now - we were using a browser plug-in and it was clear that vanilla technologies like WebCL and WebGL were going to become viable alternatives. Cloud-based streaming applications are also becoming feasible options for professional graphics - I think we're going to see a wave of this kind of stuff at Siggraph. We're focused on desktop performance right now, but still have the Fabric Core running in the cloud using Python or Node.js. We're interested in getting Creation apps running as streamed applications - the framework was designed for that kind of scenario. When you look at things like virtualised GPU and cloud-based workstations, it's an exciting area. We just figure we should get people building tools first before worrying about alternate platforms. Watch this space.
**Edit: Sorry, I went off on a tangent and didn't answer your question properly. HTML is interesting for UI, but we find Qt just does a better job - and it's the de facto standard within the markets we're interested in. Right now we're 100% focused on Python and Qt
