Intel Labs released Embree to the public.
You can download the source at the page.
Embree is a collection of high-performance ray tracing kernels, developed at Intel Labs. The kernels are optimized for photo-realistic rendering on the latest Intel® processors
with support for the SSE and AVX instruction sets. In addition to the ray tracing kernels, Embree provides an example photo-realistic rendering engine. Embree is designed for
Monte Carlo ray tracing algorithms, where the vast majority of rays are incoherent. The specific single-ray traversal kernels in Embree provide the best performance in this scenario and they are very easy to integrate into existing applications.
Here´s a Windows 32b binary. I hope it works, does on my Q9550, Crashes on my Phenom2 X4. Might be Intel only but I couldn´t see it yet.
freeglut.dll is included, if something else is missing shout here
Your AV might alarm you due to the codebase of a raytracer, the file is scanned and I provided a MD5 so you know you got my compiled, scanned and packed version.
Use one of the .bat I included to test either the cornell box, a glas ssphere or a carpaint sphere.
Controls for the render viewport:
LMB: Rotate around center of interest
MMB: Pan
RMB: Dolly (move camera closer or away from center of interest)
Strg+LMB: Pick center of interest
Strg+Shift+LMB: Pick focal distance
Alt+LMB: Roll camera around view direction
L: Decrease lens radius by one world space unit
Shift+L: Increase lens radius by one world space unit
I think it is exciting to have optimized, professionaly made raytracing kernels on display.
I am very tempted, but the task really is non trivial and hardly a one man job and a lot of learning to do. However I am tending to ignore MLT. My goal would be to write an OpenCL accellerated (if possible) biased raytracer for animation. Something like a mix of BI and VRay to serve my own egoistic needs.
With the advance in computation power, the trend goes to more and more realistic raytracers. In the end the quality improves 1% at a time, while the rendertimes remain the same.
No one seems to jump at the chance to maintain a “yesterdays” quality, but make use of the improved speed.
In a few weeks I meet a few old friends again, theoretical mathematicians and technical computer scientists, I´ll inqure there.
Bao2´s efforts coding a raytracer where very promising already, but I guess by now he came to the same conclusion. Doing it without fundamental knowledge is rather futile. You need to know a lot about algorithms, parallelization, multithreading, 64b coding, GPGPU, datastructures, code optimazition.
So, yeh, it´s a dream of mine to code a biased renderer for quite some while now (1 year+) but all it is at the moment is reasearch and learning, not a single line of code written as of yet. And Cycles discouraged me quite a bit. It seems to the outside Brecht just shook the code out his pockets on the flight back home from New Zealand while I fail to grasp some of the advanced concepts of raytracing
I think that´s what a lot of freelancing animators would want. To be able to render animations fast in a decent quality, without having to rent a truckload of GHz/h at some farm, with a decently priced renderer.
I lack the expertise to produce a production ready product in a timely fashion, I can just tinker around, but I don´t understand why no one jumps on it.
The closest thing is Vray, or Thea with their biased modules and kernels. Maybe MachStudio as well but that´s more of an rasterization renderer than a raytracer.
I still think that BI would make a great renderer for animations with some love from a dev. But reading into existing code is often harder that to rewrite it.
