LuxRender 1.4 OCL: SPPM - no GPU!??

Discussions related to GPU Acceleration in LuxRender

Moderators: Dade, jromang, tomb, coordinators

LuxRender 1.4 OCL: SPPM - no GPU!??

Postby nun1980 » Wed May 20, 2015 11:23 am

Hi!

Does SPPM support GPU (but not hybrid)?
CPU: Core i7-4790K@4.4GHz
RAM: 8GBx2 DDR3@2400MHz
Mboard: ASUS Z97-Pro Gamer
GPU: EVGA GF GTX 980 Ti SC+ ACX2.0+ NEW :D
Optical disc: LG Blu-Ray GGC-H20L
HDDs: WD 1TB + Toshiba 2TB-USB3
PSU: NOX 1kW
Monitor: CRT 19" Samsung SM 957MB
nun1980
 
Posts: 37
Joined: Tue Sep 28, 2010 4:09 pm
Location: Portugal

Re: LuxRender 1.4 OCL: SPPM - no GPU!??

Postby B.Y.O.B. » Wed May 20, 2015 2:03 pm

As far as I know neither hybrid nor pure GPU rendering are supported by SPPM.
User avatar
B.Y.O.B.
Developer
 
Posts: 5181
Joined: Wed Nov 10, 2010 4:10 pm
Location: Germany

Re: LuxRender 1.4 OCL: SPPM - no GPU!??

Postby moby1kenobi » Tue Jun 30, 2015 7:44 am

I don't think it supports either (yet ;) )
User avatar
moby1kenobi
 
Posts: 41
Joined: Sun Jun 16, 2013 7:39 am

Re: LuxRender 1.4 OCL: SPPM - no GPU!??

Postby burnin » Tue Apr 19, 2016 4:07 pm

Was just wondering if there's been any thoughts or development on the subject?
Any experiments elsewhere or demonstrations of SPPM on GPU?
Is it theoretically and practically possible?
This could be amazing...
burnin
 
Posts: 287
Joined: Mon May 03, 2010 8:04 pm

Re: LuxRender 1.4 OCL: SPPM - no GPU!??

Postby zsolnai » Wed Apr 20, 2016 1:28 am

It is indeed possible to implement SPPM on the GPU as Toshiya Hachisuka demonstrated here:
http://www.ci.i.u-tokyo.ac.jp/~hachisuka/gpusppm.zip

Our SPPM codebase is has not been updated in a while, and is implemented on the CPU. The current implementation also doesn't support participating media and subsurface scattering, so I think if there would be plans for improvements, these features would have priority. Unfortunately this is not the case, we simply don't have the resources to maintain so many algorithms. My humble opinion is that the best is if a system does one thing, and it does it well (one integrator, maybe two, fully-featured, GPU-supported, well maintained).
zsolnai
Developer
 
Posts: 519
Joined: Sun May 22, 2011 6:17 pm

Re: LuxRender 1.4 OCL: SPPM - no GPU!??

Postby sharlybg » Wed Apr 20, 2016 4:06 am

My humble opinion is that the best is if a system does one thing, and it does it well (one integrator, maybe two, fully-featured, GPU-supported, well maintained).


totally agree with You ! two robust rendering system for most of encountered situation.
Currently:

1/ Pathocl + Sobol ===>> extremly good result (speed quality) for outdoor scene (almost realtime on gpus)

2/ The current need is a second solution specially build for very fast indoor , caustics , and difficult lit scene !
i7 6700k + 32 Gb DDR4 + 2X R9 390 sapphir nitro.
User avatar
sharlybg
 
Posts: 730
Joined: Tue Nov 02, 2010 10:22 am
Location: Ivory coast

Re: LuxRender 1.4 OCL: SPPM - no GPU!??

Postby burnin » Wed Apr 20, 2016 7:26 am

Yup, also agree...

Note: Am no coder, just madly in love with science, math, physics, simulations... solving problems. Love to help... and ready to learn more.

Been experimenting a long time. Seeking a good engine for interior lighting scenarios that also has to be good with caustics! Working in Classic API environment on CPU, SPPM seemed to have always prevailed. (note: VRay and FinalRender are the only renderers that i have considered great with indoor scenes and fast caustics, just too tedious for user to set properly, conforming to the 3d scene).
Now, with amazing tech on our doorsteps (GPUs) and RedShift taking the flag, my thought was basically: "If this (SPPM) can be done with OCL, this is the chance for Lux, the second system. Yin Yang. As it could complement and fit right in for complex indoor scenes where shadows don't matter much (with lavishing decorations, lots of kitsch & exotic materials).
If things go well... old code could be removed.
For now am still investigating, researching... and as here are some very talented people doing amazing work, is why i asked. If possible and practical, will be glad to continue...
burnin
 
Posts: 287
Joined: Mon May 03, 2010 8:04 pm

Re: LuxRender 1.4 OCL: SPPM - no GPU!??

Postby sharlybg » Wed Apr 20, 2016 9:55 am

But the saddess part is that thing are not really simple as it seem !

1/ We are requesting features after features while new better rendering tech and papers are available making old one obsolete.this is inherent to science nature.

2/ Luxrender project Devs TEAM is too little to handle all at once (we are totally far from this).
Sometime I worry about luxrender project near Future ( how many active core Devs or Opencl Devs do we have actually ? ).
How many studios use lux as they main renderer ?
How strong is lux artists community ?
how lux gallery and showcase compare to cycles/vray/corona/redshift/octane ?
what is the politics to gather more donation and crowfounding do the current method work enought ?

My thought is That we do not have enougth Devs. do not forget that they are normal people with personnal problem in they own life like everyone.they are not superman (even superman have own life and own problem not only with batman of course).
I don't want Lux to turn like yafaray.
i7 6700k + 32 Gb DDR4 + 2X R9 390 sapphir nitro.
User avatar
sharlybg
 
Posts: 730
Joined: Tue Nov 02, 2010 10:22 am
Location: Ivory coast

Re: LuxRender 1.4 OCL: SPPM - no GPU!??

Postby burnin » Wed Apr 20, 2016 12:00 pm

It's anything but simple.
Was thinking more in long term (cc 2-5 years).
Since no GPU render engines i know of is quite capable of reasonable indoor+caustic renderings (except maybe redshift/closed) and being open i find this to be a nice candidate.
And would rather know now if striving is futile...
Who knows what future brings. One thing is for sure, we'll all die, tho for as long as am alive i will enjoy in things common ppl find it hard to even comprehend.
burnin
 
Posts: 287
Joined: Mon May 03, 2010 8:04 pm

Re: LuxRender 1.4 OCL: SPPM - no GPU!??

Postby nun1980 » Sat Jul 02, 2016 5:38 am

zsolnai wrote:It is indeed possible to implement SPPM on the GPU as Toshiya Hachisuka demonstrated here:
http://www.ci.i.u-tokyo.ac.jp/~hachisuka/gpusppm.zip

Our SPPM codebase is has not been updated in a while, and is implemented on the CPU. The current implementation also doesn't support participating media and subsurface scattering, so I think if there would be plans for improvements, these features would have priority. Unfortunately this is not the case, we simply don't have the resources to maintain so many algorithms. My humble opinion is that the best is if a system does one thing, and it does it well (one integrator, maybe two, fully-featured, GPU-supported, well maintained).
Glut32.dll missed... I tried add this file (year 2010) to folder but I got error message ("gpusppm.exe has stopped working")...
I've Win 10 x64 with GeForce 368.22WHQL driver (see my specs on bottom of my message).
CPU: Core i7-4790K@4.4GHz
RAM: 8GBx2 DDR3@2400MHz
Mboard: ASUS Z97-Pro Gamer
GPU: EVGA GF GTX 980 Ti SC+ ACX2.0+ NEW :D
Optical disc: LG Blu-Ray GGC-H20L
HDDs: WD 1TB + Toshiba 2TB-USB3
PSU: NOX 1kW
Monitor: CRT 19" Samsung SM 957MB
nun1980
 
Posts: 37
Joined: Tue Sep 28, 2010 4:09 pm
Location: Portugal

Next

Return to GPU Acceleration

Who is online

Users browsing this forum: No registered users and 1 guest