Page 1 of 81

LuxCore: issues, problems, regressions

PostPosted: Tue Oct 07, 2014 9:15 am
by SATtva
Guys, it seems bump is totally broken in LuxCoreRenderer (LuxBlend API 1.x) at the moment:

untitled.Scene.00001.jpg
Lux classic Path

untitled.Scene.00002.jpg
LuxCore Path (LuxCoreRenderer)

untitled.Scene.00003.jpg
LuxCore Path OpenCL (LuxCoreRenderer), CPU device only

Bump doesn't show up no matter how large is the height value.

Re: LuxCore: new features

PostPosted: Tue Oct 07, 2014 10:36 am
by Dade
SATtva wrote:Bump doesn't show up no matter how large is the height value.


SATtva, you have to define UV coordinates for the object or LuxCore will not apply bump/normal mapping. If you unwarp the sphere everything will work fine:

bump.jpg


This happen because LuxCore use the UV coordinates to define the orientation of the local frame, this has been added in the past in order to avoid some type of problem with anisotropic material (the problem was discovered by Tomb). I have to check what classic Lux does when no UV information are available and replicate the same behavior.

P.S. Pick an option:

[*]Sobol+Sun = sudo rm -rf /
[*]Sobol+Distant light = sync
[*]Metropolis+sun = sync ; sh back_up_system.sh

My eyes were bleeding for the fireflies :lol:

Re: LuxCore: new features

PostPosted: Wed Oct 08, 2014 2:41 am
by SATtva
Dade wrote:SATtva, you have to define UV coordinates for the object or LuxCore will not apply bump/normal mapping. If you unwarp the sphere everything will work fine:

Ah, that did the trick. But i must say this is very counter-intuitive. Usually one decides to use procedural textures just when he don't want to mess with UV maps, so you can easily assume there is no UV map at all. Could LuxCore generate UV map internally when it's needed?

Re: LuxCore: new features

PostPosted: Wed Oct 08, 2014 4:10 am
by Dade
SATtva wrote:
Dade wrote:SATtva, you have to define UV coordinates for the object or LuxCore will not apply bump/normal mapping. If you unwarp the sphere everything will work fine:

Ah, that did the trick. But i must say this is very counter-intuitive. Usually one decides to use procedural textures just when he don't want to mess with UV maps, so you can easily assume there is no UV map at all. Could LuxCore generate UV map internally when it's needed?


LuxCore should be now able to render bump mapping even when UV coordinates are not available, can you try it ?

Re: LuxCore: new features

PostPosted: Wed Oct 08, 2014 4:40 am
by jensverwiebe
Dade wrote:
SATtva wrote:
Dade wrote:SATtva, you have to define UV coordinates for the object or LuxCore will not apply bump/normal mapping. If you unwarp the sphere everything will work fine:

Ah, that did the trick. But i must say this is very counter-intuitive. Usually one decides to use procedural textures just when he don't want to mess with UV maps, so you can easily assume there is no UV map at all. Could LuxCore generate UV map internally when it's needed?


LuxCore should be now able to render bump mapping even when UV coordinates are not available, can you try it ?


Compiled the new feature, but get immidiately kernel compile error with SATtva's testscene now. Also on CLcpu !

RUNTIME ERROR: PathOCLBase kernel compilation error


UPDATE: without OCL is fine

Jens

Re: LuxCore: new features

PostPosted: Wed Oct 08, 2014 8:45 am
by SATtva
And here's another issue which definitely is a bug/regression compared to classic Lux: when calculating emission power it seems LuxCore treats every polygon as a separate object instead of calculating power over all polygons of an object using a given emissive material. Here's a simple test: each meshlight is a separate object, they both are using the same material; a left-hand meshlight is a simple plane (one quad) while a right-hand meshlight is split to 11 faces (11 quads). With Lux classic both meshlights produce the same emission power; with LuxCore emission is multiplied.

luxcore-emission-test.Scene.00001.jpg
Lux classic Path

luxcore-emission-test.Scene.00002.jpg
LuxCore Path

Re: LuxCore: new features

PostPosted: Thu Oct 09, 2014 2:55 am
by Dade
jensverwiebe wrote:Compiled the new feature, but get immidiately kernel compile error with SATtva's testscene now.


I should have fixed the problem.

Re: LuxCore: new features

PostPosted: Thu Oct 09, 2014 3:47 am
by jensverwiebe
Dade wrote:
jensverwiebe wrote:Compiled the new feature, but get immidiately kernel compile error with SATtva's testscene now.


I should have fixed the problem.


Jep, i verify you did :) cpu + gpu ....

Thx


Jens

Re: LuxCore: new features

PostPosted: Fri Oct 10, 2014 9:47 am
by Dade
SATtva wrote:And here's another issue which definitely is a bug/regression compared to classic Lux: when calculating emission power it seems LuxCore treats every polygon as a separate object instead of calculating power over all polygons of an object using a given emissive material. Here's a simple test: each meshlight is a separate object, they both are using the same material; a left-hand meshlight is a simple plane (one quad) while a right-hand meshlight is split to 11 faces (11 quads). With Lux classic both meshlights produce the same emission power; with LuxCore emission is multiplied.


I should have fixed the problem:

al.jpg

Re: LuxCore: new features

PostPosted: Sat Oct 11, 2014 12:54 am
by SATtva
Great, thank you!