What you miss with q3map2's `-fast` switch
| Mode | time (secs) | time (percent) | Bounces | Image |
|---|---|---|---|---|
-dirty -fast |
137 | 38 |  |
|
-dirty –-fast-- |
480 | 53 |  |
|
-fast |
332 | 89 | |
|
---fast-- |
4694 | 492 | |
- Compile options
-light -samples 3 -bounce 9999 -bouncescale 3.255 -gamma 2 -compensate 4 -patchshadows -v -skyfix.- Compile machine
- P4 2.4GHz 800mhz fsb, 1GB RAM
Hrm, -dirty seems to have a heavy influence on bounces :o.
I tweaked the -bouncescale with a -dirty compile,
thus the non-dirty compiles are overblown a bit – probably raising the number
of bounces quite a lot. I will try to get a nice -bouncescale
value for non-dirty compiles when I find time.
Difference between the two -dirty images
Difference between the two –-dirty– images
Ignore the white lines at the edges of polygons, they're caused by the merge of non-pixel-perfect images. The noise is caused by JPEG artifacts in both the source and these images.
You can Download the Cornellbox, as well as the accompanying readme . It's built from the data available from the Cornell University with rounded integer-precision.