(L) [2006/07/10] [greenhybrid] [slowly coming to the conclusion] Wayback!Some months ago I complained "oh that whole math and I don't have the time" (lyc, maybe you remember). I am not sure if my implementation is correct at the moment, but I think slowly I am reaching higher spaces [SMILEY Wink]
Next target: Metropolis Light Transport.
EDIT: Or, lets say, next target is some tone mapping^^
[IMG #1 ]
For some data I just copy+paste the newest content of my heimseite:
- 600x600 pixels
- 17,250 seconds / 4.79 hours rendertime
- 2,534,120,000 rays ( 2.534 billion )
- 200 samples per pixel
- maximum path length is 25
- objects (from left to right): glossy mirror, diffuse sculpture, diffuse box, glossy mirroring totem (and a surrounding diffuse box)
EDIT2: and yepp, I know. Rendertime is not that impressive... Without any funky monte-carlo stuff the scene takes 7 seconds. I don't have the models triangle counts in mind, something in the region of totally 0,2 - 0,3M
_________________
[LINK http://greenhybrid.net/ greenhybrid.net]
Real Men code Software Graphics.
[IMG #1]:Not scraped:
https://web.archive.org/web/20071012135106im_/http://www.root-engine.net/greenhybrid/images/glasspp_log_0050.jpg
(L) [2006/07/10] [greenhybrid] [slowly coming to the conclusion] Wayback!I've just calculated, I am at poor 0,15M Rays per second...at 1,5M RPS the rendertime would be just 28minutes [ouugghh], or when doubling the samples 56 minutes [ayygghhh]
It's really time to rewrite the kd-compiler... [SMILEY Sad]
Or to use phantom's or tbp's core ^^ (just kidding, but I'll have a look at them)
_________________
[LINK http://greenhybrid.net/ greenhybrid.net]
Real Men code Software Graphics.
(L) [2006/07/11] [lycium] [slowly coming to the conclusion] Wayback!hmm, 7 seconds for 200spp with 600^2 res? suggests to me you're spending far too much time deep down in the ray hierarchy with the gi stuff. before you attempt mlt you might like to read kajiya's 1986 siggy paper on path tracing - i doubt the former will be as forgiving to imperfect implementations as the latter, especially in matters of path weighting, probabilities etc.
btw, in that rendered image, it looks like the diffuse reflection isn't so... diffuse ;) it seems to "lean to the right", and generally the indirect light doesn't look too cosine-distributed. this may or may not be related to the glossy scene elements (you'll get a major noise reduction if you make those surfaces diffuse btw) but in general getting that correct before plugging in a complex sampling engine is paramount - you'll drive yourself insane looking for "bugs" in the metropolis code while the problem lies with the path sampling, and vice versa.
(L) [2006/07/11] [greenhybrid] [slowly coming to the conclusion] Wayback!I did not find the sig-paper. lyc, do you have a copy of the kayija-paper?
google just said me that cite-seer is down or buggy and that Kayija is working at M$.
_________________
[LINK http://greenhybrid.net/ greenhybrid.net]
Real Men code Software Graphics.
