Comments on: Depth of Field demo updated http://dword.dk/blog/2006/07/depth-of-field-demo-updated/ Software development, 3d graphics, open source, and other stuff... Mon, 06 Feb 2012 04:34:56 +0000 http://wordpress.org/?v=2.6 By: DWORD http://dword.dk/blog/2006/07/depth-of-field-demo-updated/#comment-12999 DWORD Tue, 03 Jun 2008 12:24:23 +0000 http://dword.dk/blog/2006/07/depth-of-field-demo-updated/#comment-12999 Hi Neil, I can't remember exactly what goes on in the scripts, but that line is for setting a parameter named 'source' of type 'int' to zero for a shader. You can probably find an explanation in the Ogre maunal. Check out the part on <a href="http://www.ogre3d.org/docs/manual/manual_14.html" target="_blank" rel="nofollow">Material Scripts</a> (and shaders). param_named is described here: <a href="http://www.ogre3d.org/docs/manual/manual_23.html#SEC110" target="_blank" rel="nofollow">http://www.ogre3d.org/docs/manual/manual_23.html#SEC110</a> Hope this helps, Christian Hi Neil,

I can’t remember exactly what goes on in the scripts, but that line is for setting a parameter named ’source’ of type ‘int’ to zero for a shader. You can probably find an explanation in the Ogre maunal. Check out the part on Material Scripts (and shaders). param_named is described here: http://www.ogre3d.org/docs/manual/manual_23.html#SEC110

Hope this helps,
Christian

]]> By: Neil http://dword.dk/blog/2006/07/depth-of-field-demo-updated/#comment-12994 Neil Fri, 30 May 2008 20:12:17 +0000 http://dword.dk/blog/2006/07/depth-of-field-demo-updated/#comment-12994 Hi, hopefully you're still around to answer this (I know this code is old). I'm trying to figure out this shader (By creating a .compositor file for it), but I keep running into a problem with the line: param_named source int 0 What is "source"? I can't seem to find any reference to it anywhere else. Hi, hopefully you’re still around to answer this (I know this code is old).

I’m trying to figure out this shader (By creating a .compositor file for it), but I keep running into a problem with the line:

param_named source int 0

What is “source”? I can’t seem to find any reference to it anywhere else.

]]> By: DWORD http://dword.dk/blog/2006/07/depth-of-field-demo-updated/#comment-12 DWORD Mon, 07 Aug 2006 16:56:12 +0000 http://dword.dk/blog/2006/07/depth-of-field-demo-updated/#comment-12 Hi Michail, Thanks for the update. To be honest I don't know where the NUM_TAPS=12 comes from; I just took it from the paper. And I don't know how to calculate the poisson distribution either, I found the offsets by taking a screenshot of the sample distribution in the paper and checking pixel positions of the dots. Ugly but it worked. ;) But it does appear that using 10 taps would be better, if that allows more GPUs to run it. Unfortunately I don't know how to calculate such poisson distributions. If you know, please enlighten me. I'm too lazy to search for it myself at the moment. Hi Michail,

Thanks for the update. To be honest I don’t know where the NUM_TAPS=12 comes from; I just took it from the paper. And I don’t know how to calculate the poisson distribution either, I found the offsets by taking a screenshot of the sample distribution in the paper and checking pixel positions of the dots. Ugly but it worked. ;)

But it does appear that using 10 taps would be better, if that allows more GPUs to run it. Unfortunately I don’t know how to calculate such poisson distributions. If you know, please enlighten me. I’m too lazy to search for it myself at the moment.

]]> By: Michail Glukhov http://dword.dk/blog/2006/07/depth-of-field-demo-updated/#comment-11 Michail Glukhov Wed, 02 Aug 2006 10:16:38 +0000 http://dword.dk/blog/2006/07/depth-of-field-demo-updated/#comment-11 Something strange happened with most important part of my message, so I repeat it: Some statistic on videocards: less 1 fps: FX 5900, FX5500, FX5200, Radeon 9600 (hangs, no picture) ~200 fps: X700 Pro, 6600GT, 6800LE As you can see, cards got divided into generations (similar situation in your thread in Ogre forum). I've changed back to Radeon, and started finding out source of this problem, and I found it. Changing NUM_TAPS in last combining shader to 10 (with commenting last two array elements assigning part, of course) made my poor Radeon run this demo at reasonable FPS. Same with your original demo - setting NUM_TAPS more, then 10 causes crash. Something strange happened with most important part of my message, so I repeat it:

Some statistic on videocards:
less 1 fps: FX 5900, FX5500, FX5200, Radeon 9600 (hangs, no picture)
~200 fps: X700 Pro, 6600GT, 6800LE

As you can see, cards got divided into generations (similar situation in your thread in Ogre forum).

I’ve changed back to Radeon, and started finding out source of this problem, and I found it. Changing NUM_TAPS in last combining shader to 10 (with commenting last two array elements assigning part, of course) made my poor Radeon run this demo at reasonable FPS. Same with your original demo - setting NUM_TAPS more, then 10 causes crash.

]]> By: Michail Glukhov http://dword.dk/blog/2006/07/depth-of-field-demo-updated/#comment-10 Michail Glukhov Wed, 02 Aug 2006 10:03:20 +0000 http://dword.dk/blog/2006/07/depth-of-field-demo-updated/#comment-10 <p>Hi, Christian!<br /> First of all, thanks for great demo!</p> <p>Being not too familiar with Ogre engine, we ported your demo to GLScene (with only GLSL shaders, GLScene doesn't support DirectX).</p> <p>I should mention, that original demo didn't work on my Radeon 9600 (it crashed with standard windows error "The program has performed an illegal operation"), neither worked our GLScene version. Changing videocard to GF 6600GT made both demos to run flawless.</p> <p>We posted demo in glscene.general newsgroup, and got very different feedback.<br /> Some statistic on videocards:<br /> 10 causes crash.</p> <p>Seems, that there is some kind of hardware limitation of shader operations in older cards, that causes this.</p> <p>So, I have some questions<br /> - Why you've taken 12 points for Poisson distribution (In ATI paper there are 12, but where from this number comes)?<br /> - How did you calculated values for Poisson array?</p> <p>Maybe, it would be good solution to reduce number of points to 10 (with recalculating distribution)? This can widen list of videocards, suitable for running your demo.</p> <p>Michail</p> <p>P.S. You can contact me via e-mail or ICQ [snip].<br /> P.P.S. Sorry for my english :) </p> Hi, Christian!
First of all, thanks for great demo!

Being not too familiar with Ogre engine, we ported your demo to GLScene (with only GLSL shaders, GLScene doesn’t support DirectX).

I should mention, that original demo didn’t work on my Radeon 9600 (it crashed with standard windows error “The program has performed an illegal operation”), neither worked our GLScene version. Changing videocard to GF 6600GT made both demos to run flawless.

We posted demo in glscene.general newsgroup, and got very different feedback.
Some statistic on videocards:
10 causes crash.

Seems, that there is some kind of hardware limitation of shader operations in older cards, that causes this.

So, I have some questions
- Why you’ve taken 12 points for Poisson distribution (In ATI paper there are 12, but where from this number comes)?
- How did you calculated values for Poisson array?

Maybe, it would be good solution to reduce number of points to 10 (with recalculating distribution)? This can widen list of videocards, suitable for running your demo.

Michail

P.S. You can contact me via e-mail or ICQ [snip].
P.P.S. Sorry for my english :)

]]>