BlendLuxCore Development

Discussion related to the Engine functionality, implementations and API.
User avatar
Dade
Developer
Posts: 1533
Joined: Mon Dec 04, 2017 8:36 pm

Re: BlendLuxCore Development

Post by Dade » Sun Nov 25, 2018 9:03 pm

I have the feeling you are not aware AA is samples-per-pixel^2: if you set AA to 3, there will be 9 (3x3) samples per pixel.
Support LuxCoreRender project with salts and bounties

kintuX
Posts: 263
Joined: Wed Jan 10, 2018 2:37 am

Re: BlendLuxCore Development

Post by kintuX » Sun Nov 25, 2018 9:30 pm

Dade wrote:
Sun Nov 25, 2018 9:03 pm
I have the feeling you are not aware AA is samples-per-pixel^2: if you set AA to 3, there will be 9 (3x3) samples per pixel.
Am aware of the exponential ratio (quadratic - it's a surface, cubic - spatial...).
This are just objective observations - resulting values were taken from 'Statistics' report.

Why i'm confused - is there a bug or mistake or... would love to dive in the code, but i'm still nowhere near understanding how program parts, modules are interconnected. (I understand art, math, mechanics, architecture, concepts, law, principles, physics, chemistry, literature, music... but never did got any experience with the code or programming language - well, occasionally i did, but it was pure intuition and logic, trial & error)

User avatar
Dade
Developer
Posts: 1533
Joined: Mon Dec 04, 2017 8:36 pm

Re: BlendLuxCore Development

Post by Dade » Sun Nov 25, 2018 11:14 pm

kintuX wrote:
Sun Nov 25, 2018 9:30 pm
Dade wrote:
Sun Nov 25, 2018 9:03 pm
I have the feeling you are not aware AA is samples-per-pixel^2: if you set AA to 3, there will be 9 (3x3) samples per pixel.
Am aware of the exponential ratio (quadratic - it's a surface, cubic - spatial...).
This are just objective observations - resulting values were taken from 'Statistics' report.
I ... still have the same feeling, you can not have 2 samples per pixel because you can have 1x1 or 2x2 (i.e. 4) samples per pixel so your table must be wrong.

The rows should be:
[Samples/Pixel][....etc.]
1 ....
4 ....
9 ....
16 ....
25 ....
etc.
Support LuxCoreRender project with salts and bounties

kintuX
Posts: 263
Joined: Wed Jan 10, 2018 2:37 am

Re: BlendLuxCore Development

Post by kintuX » Sun Nov 25, 2018 11:36 pm

Dade wrote:
Sun Nov 25, 2018 11:14 pm
kintuX wrote:
Sun Nov 25, 2018 9:30 pm
Dade wrote:
Sun Nov 25, 2018 9:03 pm
I have the feeling you are not aware AA is samples-per-pixel^2: if you set AA to 3, there will be 9 (3x3) samples per pixel.
Am aware of the exponential ratio (quadratic - it's a surface, cubic - spatial...).
This are just objective observations - resulting values were taken from 'Statistics' report.
I ... still have the same feeling, you can not have 2 samples per pixel because you can have 1x1 or 2x2 (i.e. 4) samples per pixel so your table must be wrong.

The rows should be:
[Samples/Pixel][....etc.]
1 ....
4 ....
9 ....
16 ....
25 ....
etc.
I know :)
Sample values are copied from Statistics in Blender (my bad for not specifying that detail - all non-Bold characters are Values copied from there)
// note to self - use colors to differentiate better + add full description and detailed info
statistics.jpg
Strange (wrong) values start as SPP is set to 4+ (which should result in 16 and so on) - i don't know what the cause could be or where to look next... I suspect this 'bad' calculation (Samples Halt condition) to be the cause of Black Tiles in Tilepath further on.

User avatar
Dade
Developer
Posts: 1533
Joined: Mon Dec 04, 2017 8:36 pm

Re: BlendLuxCore Development

Post by Dade » Mon Nov 26, 2018 8:27 am

kintuX wrote:
Sun Nov 25, 2018 11:36 pm
Sample values are copied from Statistics in Blender (my bad for not specifying that detail - all non-Bold characters are Values copied from there)
// note to self - use colors to differentiate better + add full description and detailed info
But still, if you set always the halt condition to 100 SPP, it will stop half way for most of the tested AA. For instance, with an AA of 8, it makes sense to stop only at 64, 128, 192, etc. It is the reason for unfinished rendering in your original post.
Support LuxCoreRender project with salts and bounties

kintuX
Posts: 263
Joined: Wed Jan 10, 2018 2:37 am

Re: BlendLuxCore Development

Post by kintuX » Mon Nov 26, 2018 7:31 pm

Dade wrote:
Mon Nov 26, 2018 8:27 am
kintuX wrote:
Sun Nov 25, 2018 11:36 pm
Sample values are copied from Statistics in Blender (my bad for not specifying that detail - all non-Bold characters are Values copied from there)
// note to self - use colors to differentiate better + add full description and detailed info
But still, if you set always the halt condition to 100 SPP, it will stop half way for most of the tested AA. For instance, with an AA of 8, it makes sense to stop only at 64, 128, 192, etc. It is the reason for unfinished rendering in your original post.
Oh, Halt Samples & Tilepath Samples per Pass must have common factor. Ok, it's clear now, i see it was really dumb of me :) so thank you again for everything.

About suggested clamp value for CPU, i found it works perfectly with BiDir engine.
Guess all issues are solved now.

User avatar
B.Y.O.B.
Developer
Posts: 1814
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: BlendLuxCore Development

Post by B.Y.O.B. » Mon Nov 26, 2018 8:05 pm

I have added some warning messages to the halt condition panel that should help users choose the correct halt conditions if they use tiled rendering.
Attachments
2018-11-26_21-03-28.png
Support LuxCoreRender project with salts and bounties

kintuX
Posts: 263
Joined: Wed Jan 10, 2018 2:37 am

Re: BlendLuxCore Development

Post by kintuX » Mon Nov 26, 2018 8:08 pm

B.Y.O.B. wrote:
Mon Nov 26, 2018 8:05 pm
I have added some warning messages to the halt condition panel that should help users choose the correct halt conditions if they use tiled rendering.
Thank you very much :D

PS
I was also thinking that with Tilepath it could be better to define number of passes which would then determine the number of Halt Samples.
ie.
User defines:
  • AA Samples = x
  • Halt Passes = y
then gets automatically
  • Halt Samples = z = x^2 * y

User avatar
lacilaci
Posts: 458
Joined: Fri May 04, 2018 5:16 am

Re: BlendLuxCore Development

Post by lacilaci » Mon Nov 26, 2018 8:22 pm

B.Y.O.B. wrote:
Mon Nov 26, 2018 8:05 pm
I have added some warning messages to the halt condition panel that should help users choose the correct halt conditions if they use tiled rendering.
Can't you lock the slider input in steps? Or maybe set a meaningful default and hide the setting.

User avatar
B.Y.O.B.
Developer
Posts: 1814
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: BlendLuxCore Development

Post by B.Y.O.B. » Mon Nov 26, 2018 8:46 pm

kintuX wrote:
Mon Nov 26, 2018 8:08 pm
PS
I was also thinking that with Tilepath it could be better to define number of passes which would then determine the number of Halt Samples.
ie.
User defines:
  • AA Samples = x
  • Halt Passes = y
then gets automatically
  • Halt Samples = z = x^2 * y
lacilaci wrote:
Mon Nov 26, 2018 8:22 pm
Or maybe set a meaningful default and hide the setting.
This would be possible, but I'm not sure if it's worth spending time implementing this - tile path is more a niche feature in my opinion.
Also, it's one more abstraction to maintain (in the future).
But if someone wants to do it, I'll be happy to review the code.
lacilaci wrote:
Mon Nov 26, 2018 8:22 pm
Can't you lock the slider input in steps?
The Blender API doesn't really let you do this.
Support LuxCoreRender project with salts and bounties

Post Reply