PATHGPU with OpenCL and CUDA support

Discussion related to the LuxCore functionality, implementations and API.
User avatar
Dade
Developer
Posts: 4684
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: PATHGPU with OpenCL and CUDA support

Post by Dade » Wed Apr 29, 2020 2:53 pm

marcatore wrote:
Wed Apr 29, 2020 12:36 pm
I've problem with bump and HSV nodes.

I've opened this on BlendLuxCore but actually I don't know if it's something related more to LuxCore itself or Blender addon.
https://github.com/LuxCoreRender/BlendL ... issues/436

Anyway I report this.
It is likely to be a bug due to the compile-only-once change, it should be easy to fix but I need to find the time.
Support LuxCoreRender project with salts and bounties

User avatar
Dade
Developer
Posts: 4684
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: PATHGPU with OpenCL and CUDA support

Post by Dade » Wed Apr 29, 2020 3:02 pm

Odilkhan Yakubov wrote:
Wed Apr 29, 2020 11:29 am
Why kernels always compiling on every new scene or when re-open Blender itself? It goesaround 5-10 minutes for compiling everytime :(
Are you using a Russian name as Windows login ? The problem may be there: the kernel cache is stored in your Windows user directory and I have the feeling it may fail (and recompile) if you use a user name with no Latin (i.e. Cyrillic) characters.

Can you try to create an account on your PC with only latin letters ? something like "abc" and check if login with "abc" you still get the problem ?
Support LuxCoreRender project with salts and bounties

User avatar
B.Y.O.B.
Developer
Posts: 3768
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: PATHGPU with OpenCL and CUDA support

Post by B.Y.O.B. » Wed Apr 29, 2020 4:20 pm

This special character problem is something that should really be resolved.

User avatar
Odilkhan Yakubov
Posts: 127
Joined: Fri Jan 26, 2018 10:07 pm
Location: Tashkent, Uzbekistan

Re: PATHGPU with OpenCL and CUDA support

Post by Odilkhan Yakubov » Wed Apr 29, 2020 5:27 pm

Dade wrote:
Wed Apr 29, 2020 3:02 pm
Odilkhan Yakubov wrote:
Wed Apr 29, 2020 11:29 am
Why kernels always compiling on every new scene or when re-open Blender itself? It goesaround 5-10 minutes for compiling everytime :(
Are you using a Russian name as Windows login ? The problem may be there: the kernel cache is stored in your Windows user directory and I have the feeling it may fail (and recompile) if you use a user name with no Latin (i.e. Cyrillic) characters.

Can you try to create an account on your PC with only latin letters ? something like "abc" and check if login with "abc" you still get the problem ?
I'm using Russian for translate overall Windows words, but Username is just "User" and in "Local disk C" has not any Cyrillic words. Strange.
Image

User avatar
Odilkhan Yakubov
Posts: 127
Joined: Fri Jan 26, 2018 10:07 pm
Location: Tashkent, Uzbekistan

Re: PATHGPU with OpenCL and CUDA support

Post by Odilkhan Yakubov » Wed Apr 29, 2020 5:32 pm

I'm using E-Cycles too by Matieu and here CUDA Boost option doesnt working too on me. It shows up "Display Driver has been stopped and recovered" and then close the Blender. I've tried update my vga driver but it shows me that latest drivers isnt installed on my pc. That OIDN requiered something from PC Specs?
Image

User avatar
Dade
Developer
Posts: 4684
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: PATHGPU with OpenCL and CUDA support

Post by Dade » Wed Apr 29, 2020 6:23 pm

Odilkhan Yakubov wrote:
Wed Apr 29, 2020 5:32 pm
That OIDN requiered something from PC Specs?
OIDN can be accelerated by some AVX/AVX2 instruction but the code paths should be optional and work with any relatively recent CPU.
Support LuxCoreRender project with salts and bounties

User avatar
Odilkhan Yakubov
Posts: 127
Joined: Fri Jan 26, 2018 10:07 pm
Location: Tashkent, Uzbekistan

Re: PATHGPU with OpenCL and CUDA support

Post by Odilkhan Yakubov » Wed Apr 29, 2020 9:57 pm

Dade wrote:
Wed Apr 29, 2020 6:23 pm
Odilkhan Yakubov wrote:
Wed Apr 29, 2020 5:32 pm
That OIDN requiered something from PC Specs?
OIDN can be accelerated by some AVX/AVX2 instruction but the code paths should be optional and work with any relatively recent CPU.
Oh. Are you mean is it worked with C2Q Q6600 too?
Image

User avatar
Dade
Developer
Posts: 4684
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: PATHGPU with OpenCL and CUDA support

Post by Dade » Thu Apr 30, 2020 12:58 am

Odilkhan Yakubov wrote:
Wed Apr 29, 2020 9:57 pm
Dade wrote:
Wed Apr 29, 2020 6:23 pm
Odilkhan Yakubov wrote:
Wed Apr 29, 2020 5:32 pm
That OIDN requiered something from PC Specs?
OIDN can be accelerated by some AVX/AVX2 instruction but the code paths should be optional and work with any relatively recent CPU.
Oh. Are you mean is it worked with C2Q Q6600 too?
The Q6600 starts to be really old. intel OIDN requires SSE4.1 support but Q6600 may not support that instruction set (I'm not sure, the information I have found are not very clear).
Support LuxCoreRender project with salts and bounties

User avatar
Odilkhan Yakubov
Posts: 127
Joined: Fri Jan 26, 2018 10:07 pm
Location: Tashkent, Uzbekistan

Re: PATHGPU with OpenCL and CUDA support

Post by Odilkhan Yakubov » Thu Apr 30, 2020 3:39 am

Dade wrote:
Thu Apr 30, 2020 12:58 am
Odilkhan Yakubov wrote:
Wed Apr 29, 2020 9:57 pm
Dade wrote:
Wed Apr 29, 2020 6:23 pm


OIDN can be accelerated by some AVX/AVX2 instruction but the code paths should be optional and work with any relatively recent CPU.
Oh. Are you mean is it worked with C2Q Q6600 too?
The Q6600 starts to be really old. intel OIDN requires SSE4.1 support but Q6600 may not support that instruction set (I'm not sure, the information I have found are not very clear).
:( it so sad that C2Q Q6600 @2.4ghz is set SEE3 instuction table. OIDN is supported at least SEE4.1 instruction set.
Here closed issue:
https://github.com/LuxCoreRender/LuxCore/issues/352
Here closed issue
Image

Post Reply