Shademap for stereoscopic - how to speed up rendertimes ?

[Pixelsonic]

Hi,

I´ve tried to figure out how to make use of the new shademap feature within the stereoscopic helper. I know it´s rendering faster when
directly rendering DOF and Motion Blur, but we are doing both effects in post. So I´ve setup my scene to render the first (LEFT) image with
saving a shademap per image, and then rendering the second image (RIGHT, with offset right, so I could use the first rendered image as LEFT).

The renderings took app. 20 min. per frame for the LEFT image (which saved the shademap file)
The renderings for the right image, which is then using the pre-rendered shademap, also took app. 20 min. per frame.

-> Total time per stereo image pair: app. 40 minutes

I don´t see any speed improvements when rendering this way, compared to rendering two separate images for LEFT and RIGHT without the
shademap. Rendering a stereo image instead of rendering two separate images for LEFT and RIGHT would not be an option, as it would increase
rendertime (pre-rendering shademap = 20 min., rendering stereo image = longer than 20. min. for second image).

Is there anything I´m overseeing ? Anything I´m doing wrong ? Or is the shademap primarily speeding up rendered DOF and Motion Blur ?

How is the shademap depending on the type of scene (with lots of small detail vs. low detail, many lights, less lights) ?

Any help would be appreciated.

Best,
Martin

[super gnu]

ive had this discussion with Vlado.. youd probably find that rendering the shade map once ( a full render) takes 20 mins, but rendering the stereo pair with the shade map takes about 5 mins. or at least it should do. like you im not 100% clear on how exactly the shade map works.

[vlado]

Did you check if the shademap files are actually written? Also, how did you render just the left image with shademaps (by default the helper will render the shademap from the center camera)?

Best regards,
Vlado

[Pixelsonic]

Thanks for the quick response.

First I´ve rendered the original camera (left) with interocular method: shift right, view: left. For the right image I switched view to "right" and
rendered the second image.

The shademap files were written, one for each frame, size was app. 33 mb per file on a 1998x1080 image size.

[super gnu]

i think your method might be the issue. it should only -write- a shade map -once. the second time it should use the saved map.

[Pixelsonic]

Hm, I don´t know what is wrong with the method. I´m talking about animation where the shademap needs to be rendered per frame and not once.

I just don´t see the speed difference in using the shademap feature:

Method WITH saving shademap:

1. render left image + save shademap : 20 min. per frame
2. render right image + USING shademap: 20 min. per frame
Total rendertime per stereoframe: 40 min.

Method WITHOUT using shademap:

1. render left image (without saving shademap): 20 min. per frame
2. render right image (without using shademap: 20 min. per frame
Total rendertime per stereoframe: 40 min.

What should I do differently ?