32 bit version:
http://www.indigorenderer.com/joomla/in ... &Itemid=62
64 bit version:
http://www.indigorenderer.com/joomla/in ... &Itemid=62
This release pretty much focused on performance and correctness. I spent a lot of time refactoring and improving the bidir code, both in terms of structure and performance.
Cornell box test now renders at approximately 38Ksamples/second (single thread), up from about 28Ksamples/second in 0.9 test 6.
Changelog:
0.9 test 7
* fixed bug with containing media for lights.
* raise error on re-used mesh name
* changed specular BSDF *again*
* fixed diffuse transmitter to avoid noise.
* changed reinhard tonemapping for better contrast on LDR images
* speedups / refactor for bidir code.
* more efficient sampling for participating media / SSS
* some correctness fixes for participating media / SSS
As per usual, let me know of any bugs you spot!
EDIT: updated 32 bit link to version where OpenEXR support is enabled.
Indigo 0.9 test 7
Indigo 0.9 test 7
Last edited by OnoSendai on Fri Oct 05, 2007 6:42 am, edited 4 times in total.
-
Lord of the Rings Junkie
- Posts: 222
- Joined: Fri Feb 23, 2007 4:38 pm
wooooo!
big performance improvements in my new pool render tests
24000 samples and 45+ micro-seconds / sample
now
31000 samples/second 31+ micro-seconds / sample
I have never hit over 25,000 samples before with Bidirectional MLT even in simple scenes
---
BTW, Ono have you fixed the sun position change when using Bidirectional MLT compared to MLT?
big performance improvements in my new pool render tests
24000 samples and 45+ micro-seconds / sample
now
31000 samples/second 31+ micro-seconds / sample
I have never hit over 25,000 samples before with Bidirectional MLT even in simple scenes
---
BTW, Ono have you fixed the sun position change when using Bidirectional MLT compared to MLT?
-
Heavily Tessellated
- Posts: 108
- Joined: Thu Aug 10, 2006 4:20 pm
- Location: Huh?
well, from the wine front, it's a mixed bag. The defaults like ior_test.igs runs 20% faster or more, but on the two scenes I'm currently working on (both heavy glass and refl use) it's about the same or even slower (one is 250us/sample on test6, to 270us/sample on test7)
Also a weird change, did you defer the text writes as part of the speed improvements? The console output comes up a character at a time! It's reminiscent of chatting on a 9600 baud modem. This is the exact same wine setup, but I'd like to hear other wine users...
Also a weird change, did you defer the text writes as part of the speed improvements? The console output comes up a character at a time! It's reminiscent of chatting on a 9600 baud modem. This is the exact same wine setup, but I'd like to hear other wine users...
thanks Ono!! 
i also get strange speed results (winXP 32bit):
a very basic scene renders faster (90000 to 135000 samples/sec), but a scene with .exr lighting, a glass and a liquid gives this result:
09 test 6: 16000+ samples/sec
09 test 7.2: starts with 9000, then goes down to 8000..
will test it further..
Ciao!!
Marco
i also get strange speed results (winXP 32bit):
a very basic scene renders faster (90000 to 135000 samples/sec), but a scene with .exr lighting, a glass and a liquid gives this result:
09 test 6: 16000+ samples/sec
09 test 7.2: starts with 9000, then goes down to 8000..
will test it further..
Ciao!!
Marco
Thanks Master Ono for this great SpeedUp release!!!
I did 2 comparison rendering between 0.9t6 & 0.9t7 each for 5 min:
I did 2 comparison rendering between 0.9t6 & 0.9t7 each for 5 min:
Code: Select all
dof_test.igs +31% samples / second
empty.igs +46% samples / secondpolygonmanufaktur.de
Who is online
Users browsing this forum: No registered users and 383 guests