DR and PPT?

[Micha_cg]

Hello,

today I try to use DR for a PPT rendering. From LC calculations I know, that all phases are merged at the end of the LC pass. But how works the PPT mode? The taskmanager dosn't show any continious network usage, maybe it's to small. So, I'm not sure, do the slaves add samples to the image? Doe's anybody know more about it?

Regards,
Micha

[Micha_cg]

Re: DR and PPT?

I tested it and the same image rendered 1min30s with and without DR looks a little bit different. So, it could be it work's.

[dalomar_cg]

Re: DR and PPT?

Interesting Micha...I haven't tested ppt with DR, so I can't say too much about it. For some reason though, I don't really think that it would be the best way to use DR. PPT is constantly refining the scene by solving the raycasting structure with a progressively smaller and smaller error. The thing about this is (AFAIK) is that in order to do this, the process requires access to the whole scene and the whole raycasting structure up until that point. DR works in a little bit of a different way, in that a "piece" of the scene is sent to the node to solve not the calculation that needs to be done. Once it solves that piece, it sends it back, gets another piece and so on. Since PPT isn't dealing in pieces I'm not sure what to expect. Maybe your PQMC method would be a better way to deal with a progressive image and DR.

I don't currently have anything set up to do any DR testing (I've just been moving a bunch of my stuff around), so once I get it all up and going I'll give it a shot.

[Micha_cg]

Re: DR and PPT?

Last I ask here, how much LC phases must be set to get the DR best working for all engines. My experience is, if I set 4 phases, all Quadcore slave engines run at full power. I suppose so, if I set 2 phases, at all engines will be used 2 CPUs. I will test it tomorrow.

[vlado]

Re: DR and PPT?

In DR mode, the light cache (either in single frame mode, or in ppt mode) is calculated independently by each machine, so currently it does not make much sense to use PPT mode with distributed rendering. This may change for future versions, although distributing PPT would be easier than distributing the light cache in general.

Best regards,
Vlado

[Micha_cg]

Re: DR and PPT?

Interesting. PPT per DR could be great. Some times PPT is a nice option. In my case it's a full screen shot of a diamond with dispersion (faked by three refraction layers). PPT is straight going through the calculation without I must think about a noise threshold for QMC, if the PPT looks fine, I stop it. PPT per DR could help to speed up the calcuation.

Thanks,
Micha