DP_Light Group memory leak ?


when using the Dp_Light Group node for exporting different
light passes it seems that memory is building up during rendering,
until it's over the limit.
After several frames rendering OK, the output of the layers stops.
Anyone had this problem also ??


I don't think you could have a memory leak
with Light Group nodes themselves,
but can't say more without your settings.

It's just a few lightgroup nodes that go into the global color outputs in PFNE.
But it looks like it's in combination with Multithread/Persp and
Raytrace/Global 3D shading buttons activated.
When I deactivate them memory doesn't go up...
There's also the SG_occl node but that doesn't seem to have an effect.


I found one in the Pixel Filter node editor,
I updated the x32 version of DP Filter 9.6,
let me know if that solves the problem on your side too.



My own testing was based on 5 Light Group nodes
in PFNE, raytrace + persp on, Perspective Camera,
exported/saved in 5 Global Buffers through a
Get Global Buffer node in PFNE, keeping an eye
on the RAM memory, no memory leak at frame 50.

Here's a scene to test. I replaced the original objects with a cube.
Just hit F10, the renders go into the extracted map.
On a single processor it's less visible then with multiple cores but
memory goes up.

BTW: I'm on Build 1539 LW9.6


  • DP_testscene.zip
    1 MB · Views: 280


Ok I got it, was in the Diffuse and Diffuse Shade LW Buffer
checked in the Pixel Filter node editor,
this has been fixed in the x32 version of DP Filter 9.6.

Thanks for report and sample scene,
Once again thanks for the update.
With one thread it seems to work now :), but...
when using multiple threads something strange happens now.
memory goes up like before but then it stops before reaching the limit.
I have 2 GIG RAM and it stops at 1.99 GIG.
So good thing is it doesn't go up from there anymore but the outputs
fail one by one. Every extra frame there's a buffer(s) stopping the output.


The last fix was not related to multithreading,
just for DP Filter PFNE internal buffer data.
on a single core there's no difference in memory
using one or (virtual)multi-threading in Render

If the Persp/AA option is enabled in PFNE,
Global and any other DP Filter Buffer are still
global, only one stored Buffer for one or more threads
it doesn't use more memory for MT.

In a Pixel Filter Lightwave handles threads for
Perspective or any other ACT2 Camera anti-aliasing,
this is an exclusive mechanism with the Pixel Filter API
class and strictly managed by LW.
This is not very well documented but I suspect
that it has a cost for multicore machine.
as I know a Pixel Filter plugin like PFNE can't interact
with this process.


Edit: of course I suppose that you checked the Persp/AA
option in PFNE, but does-it still happend with no LW Buffers
checked (Diffuse and Diffuse Shade in your sample scene)?
Last edited:


Depends of the reply to my question above,

do you still have this problem on MultiCore
MT enabled, Global buffers (your sample scene)
but no LW Buffers checked in PFNE?

Sorry, didn't see the edit in your post.
I did a test with the Diffuse and Diffuse Shade off.
MT/perspAA and Glob 3D shading on.
Still having the problem.
When I leave MT/PerspAA off no problem with memory.
And the same setting with Diffuse and Diffuse Shade on works also.
So looks like it's in the MT/Persp.AA part.
Top Bottom