LuxRender v1.2RC1

News & Announcements regarding releases, features, exporters and project coordination.

Moderators: Dade, coordinators

Re: LuxRender v1.2RC1

Postby Dade » Mon Feb 11, 2013 4:24 am

Rhys wrote:Just a note, testing Lord's new build viewtopic.php?p=92264#p92264, hybrid path remains misbehaving in the same way.


Have you tried the same test with a random sampler or to run the rendering for a very long time ? That kind of result may originate by the very large number of metropolis samplers used by hybrid compared to normal.

SATtva, the problem fixed in the last commit is different, it was about NULL material being not so NULL (i.e. it was visible).
User avatar
Dade
Developer
 
Posts: 8404
Joined: Sat Apr 19, 2008 6:04 pm
Location: Italy

Re: LuxRender v1.2RC1

Postby jeanphi » Mon Feb 11, 2013 4:52 am

Dade wrote:Have you tried the same test with a random sampler or to run the rendering for a very long time ? That kind of result may originate by the very large number of metropolis samplers used by hybrid compared to normal.

I think that's the case.

Dade wrote:SATtva, the problem fixed in the last commit is different, it was about NULL material being not so NULL (i.e. it was visible).

The void amplifier scene was using a glass material with IOR 1, so that's the same case than here. Null material was already fixed by my previous commit.

Jeanphi
jeanphi
Developer
 
Posts: 7943
Joined: Mon Jan 14, 2008 7:21 am

Re: LuxRender v1.2RC1

Postby Rhys » Mon Feb 11, 2013 7:02 am

Dade wrote:Have you tried the same test with a random sampler or to run the rendering for a very long time ?

It would take like 1million spp to have anything to compare too. :? But yeah, using Low discrepancy seems to behave similar to normal path.
Rhys
 
Posts: 776
Joined: Fri Mar 06, 2009 5:40 pm
Location: Australia

Re: LuxRender v1.2RC1

Postby jeanphi » Mon Feb 11, 2013 7:26 am

Rhys wrote:But yeah, using Low discrepancy seems to behave similar to normal path.

That confirms that the issue is the large number of concurrent metropolis samplers in the hybrid case. I think that to be practical in the hybrid case, only a population based metropolis sampler would do.

Jeanphi
jeanphi
Developer
 
Posts: 7943
Joined: Mon Jan 14, 2008 7:21 am

Re: LuxRender v1.2RC1

Postby Rhys » Mon Feb 11, 2013 7:52 am

With stupid ray buffer size of 64 it looks like this.
64.png

and GPU makes squealing sounds :twisted: .
Rhys
 
Posts: 776
Joined: Fri Mar 06, 2009 5:40 pm
Location: Australia

Re: LuxRender v1.2RC1

Postby Dade » Mon Feb 11, 2013 7:54 am

Rhys wrote:With stupid ray buffer size of 64 it looks like this.
64.png

and GPU makes squealing sounds :twisted: .


It should be comparable to running native C++ Lux with 64 threads.
User avatar
Dade
Developer
 
Posts: 8404
Joined: Sat Apr 19, 2008 6:04 pm
Location: Italy

Re: LuxRender v1.2RC1

Postby pciccone » Tue Feb 12, 2013 10:49 pm

Is it a known thing that the Windows 32-bit NON-OpenCL archive in the download page is actually the OpenCL version?
Just FYI.
User avatar
pciccone
Developer
 
Posts: 1579
Joined: Wed Jan 13, 2010 11:02 am
Location: South Carolina

Re: LuxRender v1.2RC1

Postby SATtva » Tue Feb 12, 2013 10:57 pm

Hah! This would explain some of the issues recently reported in the forum. :)
Linux builds packager
聞くのは一時の恥、聞かぬのは一生の恥
User avatar
SATtva
Developer
 
Posts: 7164
Joined: Tue Apr 07, 2009 12:19 pm
Location: from Siberia with love

Re: LuxRender v1.2RC1

Postby pciccone » Tue Feb 12, 2013 11:03 pm

Actually, I'm not totally sure. It does not crash if reading a scene with hybrid set and it does set that it falls back to non-OpenCL devices because my Windows box doesn't have an OpenCL GPU. It also contains SLG3, which usually is not included in the non-OpenCL distribution. On the other hand the OpenCL version does complain about the lack of the OpenCL dll...
User avatar
pciccone
Developer
 
Posts: 1579
Joined: Wed Jan 13, 2010 11:02 am
Location: South Carolina

Re: LuxRender v1.2RC1

Postby Lord Crc » Wed Feb 13, 2013 7:37 am

pciccone wrote:Actually, I'm not totally sure. It does not crash if reading a scene with hybrid set and it does set that it falls back to non-OpenCL devices because my Windows box doesn't have an OpenCL GPU. It also contains SLG3, which usually is not included in the non-OpenCL distribution. On the other hand the OpenCL version does complain about the lack of the OpenCL dll...


If it's not complaining about OpenCL.dll, then it's not OpenCL-enabled, simple as that :) This is because we don't do late loading of the DLL, so the OpenCL-enabled builds will fail during startup if the dll is not found.

I've implemented CPU fall-back in hybrid renderer some time ago, and SLG3 also supports some native (non-OpenCL) rendering methods, hence why it was included.

edit: I could still have made a mistake of course, I haven't had time to check it out.
May contain traces of nuts.
User avatar
Lord Crc
Developer
 
Posts: 5032
Joined: Sat Nov 17, 2007 2:10 pm

PreviousNext

Return to News & Announcements

Who is online

Users browsing this forum: Bing [Bot] and 1 guest