LuxCoreRender Forums

Is there a way to reduce or completely avoid the compiling time everytime we do changes on materials, lights and so on when in viewport mode or everytime we launch the final render?

If a recompiling during the final render it's not a big issue, during the rendered viewport preview it's quite bothering....and as soon the scene grow in complexity the compiling time become really a waste of time.

Looking to other GPU render engine like Cycles or Vray we don't have a recompiling everytime. Is it not possible to have in LuxCore?
What kind of problems shall arise?

Vray manual said this about OpenCL
https://docs.chaosgroup.com/display/VRA ... ompilation

What will be developed to improve this situation?

LuxCore uses a kernel cache so a specific version of the code is compiled only one time (i.e. the first time is used). The option to compile
many types of code fragments at once is also available (it is in LuxCoreUI).

If you run the pre-view 2 times in a row, you should notice that there is no compilation the second time.

However LuxCore dynamically generate OpenCL code to render same type of recursive materials/textures: for any material/texture referencing other materials/textures (i.e. mix material, add texture, etc.). This is code is cached too but as soon as you edit the material/texture node connections, LuxCore will need to generate and compile new version of the code.

This is also the reason why the default for view port rendering is the CPU and not the GPU.

This problem can be addressed in the frame work of v2.3 and all materials related stuff however it is not a small amount of work.

This problem can be address in the frame work of v2.3 and all materials related stuff however it is not a simply problem to solve.

So there is hope

like it.

Dade wrote: ↑Wed Apr 03, 2019 2:38 pm This problem can be address in the frame work of v2.3 and all materials related stuff however it is not a small amount of work.

Ok...so we'll postpone this matter for the v2.3 time.