Updated pbr breaks instancing

[riccardobl]

this line needs to be wPosition = TransformWorld(modelSpacePos).xyz; or it won't work with instancing

jmonkeyengine/jme3-core/src/main/resources/Common/MatDefs/Light/PBRLighting.vert

Line 79 in 2fa2101

wPosition = (g_WorldMatrix * vec4(inPosition, 1.0)).xyz;

also lPosition seems to be unusued

[yaRnMcDonuts]

Thanks for finding this,

I should've fixed this in this recent PR, where someone pointed out that calculating the world transform the wrong way for setting up fog was causing issues with instancing (946b3ea)

And I accidentally overlooked that the wPosition calculation needed updated like that too.

I'll submit a PR fixing this soon unless anyone else beats me to it.

also lPosition seems to be unusued

I left this there intentionally because it is often used by forks of PBRLighting, so I figured I should make it an easily accessible variable in the new PBRLightingUtils lib (similar to wPosition). That way users don't have to go and fork the whole .vert shader just to get accesss to the lPosition value.

[yaRnMcDonuts]

Also, just for learning purposes, I am curious how you determined that instancing is broken?

Did it cause a crash, or is it just no longer optimizing the FPS and you found out it was broken by observing a higher than usual FPS in instanced scenes? I did run the jme-examples related to instancing prior to testing each release for 3.8 and it didn't crash or have any apparent issues, so I likely wasn't looking for the right indicators.

I only ask because I haven't used instancing that much and not in a long time (I'm not at the point of optimizing my game yet, so this is a blind spot for me)

[riccardobl]

Hey, thanks.
Just by reading the code, Instancing.glsllib has this:

jmonkeyengine/jme3-core/src/main/resources/Common/ShaderLib/Instancing.glsllib

Line 78 in 2fa2101

#define g_WorldMatrix Use_the_instancing_functions_for_this

that should crash if g_WorldMatrix is being used, otherwise transforms would be all wrong.

[yaRnMcDonuts]

I submitted a PR to fix this and also added a test using pbr with instancing to jme-examples.

I'll try to have a 3.8.1 patch release out with this fix soon too since this is a fairly significant bug.

[riccardobl]

i'll try to submit some other patches today, i've found a possible issue in the way irradiance is calculated