New RTPATHOCL synchronization code

Discussion related to the LuxCore functionality, implementations and API.
User avatar
Sharlybg
Donor
Donor
Posts: 3101
Joined: Mon Dec 04, 2017 10:11 pm
Location: Ivory Coast

Re: New RTPATHOCL synchronization code

Post by Sharlybg »

BTW, is there any interest in adding CPU (i.e. hybrid => CPU+GPU) support for RTPATHOCL ? It may be helpful on high end CPUs where the CPU is about as fast as the GPU (i.e. RTPATHOCL is currently GPU-only rendering).
More power ? yes of course :D
Support LuxCoreRender project with salts and bounties

Portfolio : https://www.behance.net/DRAVIA
juangea
Donor
Donor
Posts: 332
Joined: Thu Jan 02, 2020 6:23 pm

Re: New RTPATHOCL synchronization code

Post by juangea »

Yes of course!

Any Ryzen/Threadripper can take leverage on this :)
User avatar
B.Y.O.B.
Developer
Developer
Posts: 4146
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: New RTPATHOCL synchronization code

Post by B.Y.O.B. »

I gave it a spin and it seems to work quite well so far. The Blender interface is much more fluid, even in heavy scenes.

The only oddity I observe is that after viewport denoising, a certain ghosting shows up. It seems like the film is not properly cleared after scene edits, leaving some samples of the last frame(s), which are picked up by the denoiser?
To reproduce, open the danish mood scene, start the viewport render and move the camera around (viewport denoising should already be enabled). Then stop moving and wait for 10 seconds, until the denoiser is started.
Attachments
ghosting.PNG
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: New RTPATHOCL synchronization code

Post by Dade »

B.Y.O.B. wrote: Sun Jun 21, 2020 2:02 pm The only oddity I observe is that after viewport denoising, a certain ghosting shows up. It seems like the film is not properly cleared after scene edits, leaving some samples of the last frame(s), which are picked up by the denoiser?
Yes, the frame is over-writing the older one ... my be the problem is in AOVs, I will check.
Support LuxCoreRender project with salts and bounties
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: New RTPATHOCL synchronization code

Post by Dade »

Dade wrote: Sun Jun 21, 2020 3:32 pm Yes, the frame is over-writing the older one ... my be the problem is in AOVs, I will check.
There was a problem when resizing the view port causing ghosting even without OIDN. I fixed this issue.

At the very beginning, there still some ghosting-like effect but I'm not sure if there is still something or it is just the normal OIDN behavior when there aren't enough samples :?:
Support LuxCoreRender project with salts and bounties
User avatar
TAO
Developer
Developer
Posts: 851
Joined: Sun Mar 24, 2019 4:49 pm
Location: France
Contact:

Re: New RTPATHOCL synchronization code

Post by TAO »

Great news.
I'm going to try it and see if everythings works as it should. Great job.
User avatar
B.Y.O.B.
Developer
Developer
Posts: 4146
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: New RTPATHOCL synchronization code

Post by B.Y.O.B. »

Dade wrote: Mon Jun 22, 2020 8:50 am There was a problem when resizing the view port causing ghosting even without OIDN. I fixed this issue.
I am still getting the same ghosting after moving the camera.

Also, RTPathOCL sometimes gets stuck completely when resizing the viewport, or when ending the viewport render, requiring to kill the Blender process via task manager. I got hangs like this both with OpenCL and with CUDA backend, only tested on Windows 10 so far.
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: New RTPATHOCL synchronization code

Post by Dade »

B.Y.O.B. wrote: Mon Jun 22, 2020 5:06 pm
Dade wrote: Mon Jun 22, 2020 8:50 am There was a problem when resizing the view port causing ghosting even without OIDN. I fixed this issue.
I am still getting the same ghosting after moving the camera.
I should have fixed this problem.
Support LuxCoreRender project with salts and bounties
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: New RTPATHOCL synchronization code

Post by Dade »

B.Y.O.B. wrote: Mon Jun 22, 2020 5:06 pm Also, RTPathOCL sometimes gets stuck completely when resizing the viewport, or when ending the viewport render, requiring to kill the Blender process via task manager. I got hangs like this both with OpenCL and with CUDA backend, only tested on Windows 10 so far.
I'm unable to reproduce this problem but I fixed something that could be related: https://github.com/LuxCoreRender/BlendL ... issues/485
Support LuxCoreRender project with salts and bounties
kintuX
Posts: 810
Joined: Wed Jan 10, 2018 2:37 am

Re: New RTPATHOCL synchronization code

Post by kintuX »

Dade wrote: Tue Jun 23, 2020 9:43 am
B.Y.O.B. wrote: Mon Jun 22, 2020 5:06 pm Also, RTPathOCL sometimes gets stuck completely when resizing the viewport, or when ending the viewport render, requiring to kill the Blender process via task manager. I got hangs like this both with OpenCL and with CUDA backend, only tested on Windows 10 so far.
I'm unable to reproduce this problem but I fixed something that could be related: https://github.com/LuxCoreRender/BlendL ... issues/485
Confirming, build against commit fec47c31b works OK - doesn't crash nor freeze 8-)
Post Reply