On Dec 20, 2007, at 11:08 AM, Brian Budge wrote:

Hey Bigler -

How's it going?   Everything okay  in the ol' SLC? I've been looking at

Manta, mainly because Phillip Slussalek suggested it, and before that I

didn't know that the source was available :)

It seems like a super cool project.  I am trying to run a couple of

benchmarks on it, and maybe use it for a side experiment in my upcoming

paper.  I saw in the documents how to run it with the sceneviewer, but I

really just need to do batch rendering.  In fact, if I could use it to

benchmark the path tracer with X samples per pixel at AxB resolution using

the kd-tree, I'd be stoked to do that.  Is this possible?  I have specific

models in mind which are, unfortunately, proprietary (I'm sure you SCI folks

know how that is).

manta can run in "batch" mode.

bin/manta -res 1024x768 -imagedisplay "file(-type tga)" -bench 1 0

This will render a single frame and write it out to a file.

If you want to render using more than one sampler per pixel, you will need a different pixel sampler:

bin/manta -pixelsampler "jitter(-numberOfSamples 16)" ...

I'm not sure what the state of the path tracer is, but you can invoke it by changing the renderer:

bin/manta -renderer pathtracer ..

There are certain configurations of the pixel sampler and image traverser that can change the performance of the path tracer.  I'll let other chime in.

Also, just to see how crap my own ray tracer is, can I enable the kd-tree

for the viewer?  My own kd-tree ray tracer without sse is about the same

speed as tte DynBVH that you guys have, but I would guess that your kd-tree

is a good deal faster.

I assume you are using the triangleSceneViewer:

bin/manta -scene "lib/libscene_triangleSceneViewer(-KDTree)"

Just check out the .cc file for more options.

James