[bjorn.syse]

Hi!

I'm curious on what effect the number of polygons in my emitters have, on rendertime. Let's say I have a bunch of planes acting as rooflights. What difference would it make if they were 1 quad polygon, or two triangle polygons, or even if the polygon count was higher.

I assume the relationship polygoncount vs rendertime is not linear, no? So how does it look? Has it got more effect in the beginning, like if 10 polygons would cause a drastic difference compared to 1 polygon? Or is the difference also visible when you allready have a lot of polygons in you emitters. How would 200 polys compare to 400 let's

If there exists such a relationship between the rendertime and the polycount in the emitters, please share som insight on how to think here. a graph if possible

regards,

- björn

[Bubbaloo]

There were improvements to the way Maxwell handles high poly emitters for V2. In 1.7, there used to be problems with rendered output (strange patterns). BTW, since Maxwell triangulates meshes, 1 quad polygon = 2 triangles.

But I haven't done any comparison tests. It would be a good test to do.

[bjorn.syse]

BTW, since Maxwell triangulates meshes, 1 quad polygon = 2 triangles

aH, all that merging in vain

Yes, a few tests would come in handy of course. Should you or me do them? I'm on a macbook pro

[Bubbaloo]

Time is money, and I'm broke.
Let's both put it on our to-do lists.

[bjorn.syse]

Fair enough! Unless someone allready has some insight to share on this matter.

[kami]

I did some tests a while ago. I got no exact numbers, but I didn't make that much of a difference, but it is noticable, if there are way more emitting polygons.

[polynurb]

Time is money, and I'm broke.
Let's both put it on our to-do lists.

if you get into this.. i would be interested in how welding affects the lighting and speed of calculation.. if at all..

as time allows i'll be happy to contribute my own tests...

[brodie_geers]

Actually I posted the exact same question awhile back with similar answers. So I did a quick test

http://maxwellrender.com/forum/viewtopi ... 97&t=33919

In short I found that the # of polygons affects the render time significantly however the drop-off gets less as you increase emitters.

-Brodie

[bjorn.syse]

Ah, how silly. And I thought I made a search before I posted, but I must have missed it.

Ok, another question then. I know lighting with IBL is at least over 30 times faster to clear SSS than emitters. I tested this in this thread (http://www.maxwellrender.com/forum/view ... 97&t=32689) with these results:



But, what if an emitter has a HDR mapped to it, what category would that fall under?

- Björn

[tom]

But, what if an emitter has a HDR mapped to it, what category would that fall under?

It's still an emitter. Emitter and image based spherical lighting are two things apart. So, applying a HDR texture on an emitter can only make it slightly slower than a uniform emitter.

[bjorn.syse]

But, what if an emitter has a HDR mapped to it, what category would that fall under?

It's still an emitter. Emitter and image based spherical lighting are two things apart. So, applying a HDR texture on an emitter can only make it slightly slower than a uniform emitter.

Thanks, good to know!

[Fernando Tella]

Something to take into account is that when you mask an emitter it is not emitting with the total intensity set at the mxm but with an intensity proportional to the amount of white in the mask. The intensity does not distribute over the unmasked part but the whole mesh (despite it is masked or not).

[bjorn.syse]

Something to take into account is that when you mask an emitter it is not emitting with the total intensity set at the mxm but with an intensity proportional to the amount of white in the mask. The intensity does not distribute over the unmasked part but the whole mesh (despite it is masked or not).

Also good to know, thanks!

[kami]

Plane: SL:12 # of Triangles: 2 Time Passed: 23s Benchmark: 3228.69

Cube: SL:12 # of Triangles: 12 Time Passed: 31s Benchmark: 2511.20

Sphere: SL:12 # of Triangles: 5040 Time Passed: 52s Benchmark: 1451.43

It's not that bad ... 2500 times more polygons meaning it takes twice as long to render time.

[brodie_geers]

Well, twice as long CAN be life or death.

I think the most important thing to notice is that the dropoff is far from linear. The first 25% loss is just in going from 2 triangles to 12. So I'd say that if you've got 1 light in the scene it may be more valuable to optimize than if you've got, say 30 lights in the scene.

-Brodie