SPPM renderer (CPU-only)

[Vutshi]

So is a wider or narrower curve better in this case? I've seen a few of these graphs around and don't really get what they are showing. How does the x-coordinate of a pixel affect the light intensity there? At least, how does it it affect it in such a way to produce a graph that looks kind of like a normal distribution?

The wider blue curve is perfect. This graph just shows the single line of pixels from the center of the image below:

This is how the plane with a single point like light source looks like.

The point is that for the setup there is an analytic solution and one can compare it to the output of the luxrender.

[Pilchard123]

Ah. I didn't realise it was based on a test image. I suppose I should have really.

Thanks.

[Vutshi]

Hi,

I repeated the experiment with the plane and point like light source (1m above the plane) for camera situated 0.1m above the plane. Path and bidir give correct result proportional to 1/R^3 and sppm gives something new. Now it is proportional to 1/R^2 in the central region, and it behaves strangely in the tails. It looks like sppm lacks something like a Cos(theta). Convergence is veeeery slow for sppm.

SPPM 500K photons per pass

UPDATE: sppm with 5K photons per pass behaves differently, it starts as ~R^(-4) and then continues to grow towards maybe the same state as 500K.

SPPM 5K and 500K photons per pass

[guibou]

I've pushed a few commits for sppm. Mainly a small performance improvment and the habilitie to use the probabilistics approach using a switch in sppmintegrator.

Please try to use the "useproba" setting of the sppm integrator. You should see no difference at all, and if this is the case, I plane to switch to this approach permanently in a near future (It may allows lot of interesting improvements in memory, speed, better behavior of initial radius, less issue for parallel computation and network rendering). Please, I really need feedbacks

[Chawl]

In anycase if someone needs some fresh binaries

Edit: See next message.

[Chawl]

Some SPPM changes are detected on the changelog, so it's time to build...

Edit: Updated to recent.

[guibou]

Some SPPM changes are detected on the changelog, so it's time to build...

Yes, fixed a small race condition inside the parrallelhashgrid accel.

Thank for your build.

[Chawl]

SPPM halttime fixed, so right time for another binary.

[Chawl]

SPPM is marvellous but glossy hitpoints implementation is still itchy. I know this is a known issue but reminding anyway. I've used the latest luxrender binary from previous post and got the blender scene from this thread.

Here are the shots.

SPPM, Glossy Threshold=100 (luxblend default)

SPPM, Glossy Threshold=0 (eye pass only)

SPPM, Glossy Threshold=15

SPPM, Glossy Threshold=5000

MLT, (unbiased preset)

[guibou]

SPPM is marvellous but glossy hitpoints implementation is still itchy. I know this is a known issue but reminding anyway. I've used the latest luxrender binary from previous post and got the blender scene from this thread.

Yes, it is mainly because glossy surfaces mixes between the glossy behavior and a diffuse one.

We have to split both behavior, or we have to find a better heuristic.