GPU+CPU support for Bi-directional mode

[Googog Mogog]

Googog,

With all due respect to your observations, please educate yourself before posting!

There are people that would somewhat disagree with what you observe as the real world behavior of light. Such are for example Richard Feynman who has received a nobel price for his work in the area of Quantum Electro Dynamics. You mentioned Maxwell earlier. Has it ever dawned on you why this software is called Maxwell or are you really not familiar with the work of James Clerk Maxwell ?

There are a good number of people that have worked for decades of their academic careers on lighting simulation algorithms an have compared these with real world measurements. Phong, Blinn, Ward .... sounds familiar ?

If you dive a little deeper into Indigo's abilities and read through the forums you will find that Inigo is indeed quite accurate in simulating the behavior of light.

I said "this is my observation"

I am not saying that i am great scientist of light physics or someone better then any other i just write my observation which means through which i simulate light in my scene.
And u need to educate yourself first i am not talking about visual psychology of light inside renderers. I just said they are not "Real light simulators" means they are only based on codes and mathematical calculations. Not real light itself!

In real world light is light not a visual representation of codes and values and calculations. Got it!

However unbiased renderers simulate code generated Light as accurate as possible but it does not mean that there is real light inside 3d scene.

[Googog Mogog]

As for "true light simulation", if we restrict ourselves to Classical Geometric Optics (no quantum phenomena, no interference etc.), then Indigo is pretty much correct.

Thanks for the link i dont read that before. This classical Observation is nearly same as my own observation.

However there is one question. If u simulate reflection with an HDR Image they look awesome. But in an Unbiased renderer even with the correct computer generated light simulation these reflections somewhat look computer generated not like real chrome ball. But some images generated with indigo look so good and as close as real photo.

[pixie]

As for "true light simulation", if we restrict ourselves to Classical Geometric Optics (no quantum phenomena, no interference etc.), then Indigo is pretty much correct.

Thanks for the link i dont read that before. This classical Observation is nearly same as my own observation.

However there is one question. If u simulate reflection with an HDR Image they look awesome. But in an Unbiased renderer even with the correct computer generated light simulation these reflections somewhat look computer generated not like real chrome ball. But some images generated with indigo look so good and as close as real photo.

Because there's more to real photo then it meets the eye...

[galinette]

Unbiased has a very specific technical definition, it's not a matter of opinion It's about the correctness (on average) of a statistical estimator, which is what Monte Carlo renderers fundamentally create.

Please see: http://www.cgafaq.info/wiki/Bias_in_rendering

As for "true light simulation", if we restrict ourselves to Classical Geometric Optics (no quantum phenomena, no interference etc.), then Indigo is pretty much correct.

One exception to this : light polarization, afaik Indigo (and every gfx oriented renderer, by opposition to light simulators meant to simulate and not to graph) is based on scalar light model. Some errors may arise, especially when rendering glass

Etienne

[lycium]

Ahh good point Etienne! As expected from someone who wrote his own polarisation-correct ray tracer

As you know, but might be worth mentioning here, in the before-time, Indigo used to support polarisation, and could successfully reproduce the result of Brewster's experiment.

[KuroiNeko]

[OFF-TOPIC]
I believe in science as I believe in math... one can't exist without the other... calculations for simulations on the light behavior can be accurate... It is not and never will be Light it Self... but even light has its logical part and if there is logic in light it can be calculated. I'm not here to fight anyone but I believe and have Faith in Indigo and it's developers... been with Indigo since it was a free software and I just Love it...

now... back on topic... I can't wait for a release of Indigo that can make Hybrid bidir rendering... and I'd like to ask If there is any possibility of a MTL or similar system with CUDA acceleration... Path Only just ain't doing that trick to me with archviz... meshlight is really a pain in the @$$ path trace only...

[lycium]

Single-directional path tracing with Glass Acceleration is actually great for archviz in my experience (sometimes even beating bidir, e.g. when the scene is very large).

Nevertheless, of course we'd love to have all rendering modes working perfectly with GPU support. It's just a massive combinatorial workload (all features, all rendering modes, all CPU/GPU platforms), something we'll be looking into after 3.2 Stable.