Toon Shade GC 2a

Toon Shade GC 2a

Description

re-test
various fixes from the first toon shade

Message boards : Comments and discussion : 718

Author Message
Profile Janus
Volunteer moderator
Project administrator
Avatar
Send message
Joined: 16 Jun 04
Posts: 4483
Credit: 2,094,806
RAC: 0
Message 7212 - Posted: 28 Dec 2007, 23:42:43 UTC
Last modified: 28 Dec 2007, 23:46:22 UTC

This session is bitwise identical to the rejected 716.

A general note:
Please, if you use unsupported features but would like to render the session on BURP anyways: state clearly in the session description that you don\'t care about BURP making a best-effort attempt at rendering your scene. This way we can avoid having to double-upload sessions like this one simply because the first got rejected due to (for instance) using subsurface scattering in a multiple-parts-per-frame session.

yogibear7
Send message
Joined: 31 Aug 06
Posts: 14
Credit: 56,847
RAC: 0
Message 7213 - Posted: 29 Dec 2007, 0:46:29 UTC

I attempted to disable SSS

macgyver444
Project donor
Send message
Joined: 15 Nov 07
Posts: 1
Credit: 67,697
RAC: 0
Message 7216 - Posted: 29 Dec 2007, 4:05:00 UTC

28-Dec-2007 22:42:49 [BURP] Scheduler request succeeded: got 0 new tasks
28-Dec-2007 22:42:49 [BURP] Message from server: No work sent
28-Dec-2007 22:42:49 [BURP] Message from server: Your computer has 511 MB of memory, and 512 MB is needed


511 megs of ram? Odd, I was pretty sure I had 512 installed in this comp. Any chance BURP could cut me a little slack?

AC
Project donor
Avatar
Send message
Joined: 30 Sep 07
Posts: 121
Credit: 143,874
RAC: 0
Message 7218 - Posted: 29 Dec 2007, 5:03:04 UTC - in response to Message 7216.
Last modified: 29 Dec 2007, 5:07:23 UTC

511 megs of ram? Odd, I was pretty sure I had 512 installed in this comp. Any chance BURP could cut me a little slack?


That\'s a problem that pops up in the message boards frequently. This is a good example of one of those times when I wish it would cut a little slack though -- there are plenty (38k!) of WUs to go around and no other jobs to process.

kolban
Send message
Joined: 9 Dec 07
Posts: 12
Credit: 7,339
RAC: 0
Message 7223 - Posted: 29 Dec 2007, 18:18:28 UTC - in response to Message 7218.

I notice that this session is using MBLUR with OSA of 16. My understanding says that the image will be rendered 16 times for each part before returning a result. I was surprised to see 80+ minutes of CPU time consumed on my machine for what looks like 1/32 of a frame but if my understanding is correct then without motion blur the time would be 1/16 which would be 5 minutes per part. From what I read, there is an alternative to Motion Blur called Vector Blur that claims to be much faster ... I\'m wondering (out loud) if we should be wrorried about the use of this very expensive function?

Ref on Vector Blur: http://www.davidjarvis.ca/blender/tutorial-09.shtml

yogibear7
Send message
Joined: 31 Aug 06
Posts: 14
Credit: 56,847
RAC: 0
Message 7224 - Posted: 29 Dec 2007, 20:50:54 UTC

vector blur is a good solution. I\'ve also gone ahead and gotten rid of all sub-surfacr scattering. Is it still possible for this session to be cancelled and I re-submit it with the new settings?

Profile Janus
Volunteer moderator
Project administrator
Avatar
Send message
Joined: 16 Jun 04
Posts: 4483
Credit: 2,094,806
RAC: 0
Message 7225 - Posted: 29 Dec 2007, 21:44:40 UTC - in response to Message 7224.
Last modified: 29 Dec 2007, 21:45:38 UTC

vector blur is a good solution. I\'ve also gone ahead and gotten rid of all sub-surfacr scattering. Is it still possible for this session to be cancelled and I re-submit it with the new settings?

You should be aware that vector blur is a full-frame effect (as far as I know, let me know if this is wrong) so just as subsurface scattering it doesn\'t make sense to use it with frames that are split into multiple parts when rendered.

There is a video guide in the tutorial subforum on how to apply it after having rendered the normal scene with BURP.

It will be interesting to be able to compare the visual differences between the two versions of the scene mentioned here (a future one with vector blur and no SSS versus this one with normal motionblur and active SSS), so I\'ll let the current session run to completion for this reason.

At some point we should compile a really nice guide on all the things to think about before uploading stuff to BURP. Such a guide could neatly include links to old sessions in order to show exactly what happens if one fails to follow the suggestions given in the guide.

Rollo
Project donor
Send message
Joined: 11 Mar 06
Posts: 20
Credit: 497,175
RAC: 337
Message 7226 - Posted: 29 Dec 2007, 22:11:36 UTC

Is there a problem with motion blur and rendering a frame in parts? For example in frame 31 one can clearly see each part of the frame.

Profile Janus
Volunteer moderator
Project administrator
Avatar
Send message
Joined: 16 Jun 04
Posts: 4483
Credit: 2,094,806
RAC: 0
Message 7227 - Posted: 29 Dec 2007, 22:34:13 UTC - in response to Message 7226.
Last modified: 29 Dec 2007, 22:35:35 UTC

Is there a problem with motion blur and rendering a frame in parts? For example in frame 31 one can clearly see each part of the frame.

No, \"normal\" motionblur simply renders each pixel #OSA times, each time shifting the underlying moving geometry by (1/#OSA)*#mblur of a full frame\'s movement. In other words it renders the same frame several times and progresses the time counter in between each render by something like 0.1 frame. This effect creates the visual appearence of motionblur when the rendered frames are composited back on top of each other. This effect does not require knowledge of any adjacent pixels and hence causes no problems with distributed rendering.

The effect you see in the mentioned frame is that of subsurface scattering. Since the preprocessing is only done for the part being rendered and not the entire image each part will have insufficient information to generate the correct pixels close to the edge of that part (all a single computer is seeing outside its own part is pure black since it cannot communicate with the other computers rendering that frame). Since SSS influences how lit the pixels will be, the final image (when rendered in parts) may have wildly varying light intensities in each part.

nick
Send message
Joined: 23 Aug 06
Posts: 23
Credit: 251,238
RAC: 0
Message 7228 - Posted: 29 Dec 2007, 22:36:35 UTC

hey can we bump down the ram requirements on this one? on all my machines that have the 512 or more. there topping out at about 150MB, think we could push it down to something like 384? to be safe and it would speed up the processing.
thanks
____________

Profile Janus
Volunteer moderator
Project administrator
Avatar
Send message
Joined: 16 Jun 04
Posts: 4483
Credit: 2,094,806
RAC: 0
Message 7229 - Posted: 29 Dec 2007, 22:44:54 UTC - in response to Message 7228.
Last modified: 29 Dec 2007, 22:46:06 UTC

hey can we bump down the ram requirements on this one? on all my machines that have the 512 or more. there topping out at about 150MB, think we could push it down to something like 384? to be safe and it would speed up the processing.
thanks

I\'m sorry. Once the session has been started it is stuck at the same RAM requirement setting.

On the bright side: One thing I noticed is that RAM requirements are considerably lower when running a session in multipart mode compared to rendering the entire frame. This is quite interesting and something that will have to go into the design of the session autodetection/rejection system.

For future sessions this will be taken into consideration when checking the memory requirements indicated by the author.

This is good news for people with low memory machines.

nick
Send message
Joined: 23 Aug 06
Posts: 23
Credit: 251,238
RAC: 0
Message 7230 - Posted: 29 Dec 2007, 23:13:44 UTC

oh hey thats ok, just a suggestion, i only have one with less that a Gig right now.
____________

kolban
Send message
Joined: 9 Dec 07
Posts: 12
Credit: 7,339
RAC: 0
Message 7231 - Posted: 30 Dec 2007, 1:11:40 UTC - in response to Message 7230.

There was a question asked by the Session originator ... can Session 718 be cancelled?

Profile noderaser
Project donor
Avatar
Send message
Joined: 28 Mar 06
Posts: 507
Credit: 1,551,222
RAC: 210
Message 7232 - Posted: 30 Dec 2007, 7:33:33 UTC

Janus is going to let the session run, to test the differences between vector blur & no SSS and motionblur and SSS. See http://burp.boinc.dk/forum_thread.php?id=1024&nowrap=true#7225
____________

kolban
Send message
Joined: 9 Dec 07
Posts: 12
Credit: 7,339
RAC: 0
Message 7233 - Posted: 30 Dec 2007, 17:15:32 UTC - in response to Message 7232.

Aha ... thanks ... sorry ... I\'d missed that.

Neil

loki
Send message
Joined: 7 Feb 07
Posts: 97
Credit: 158,266
RAC: 0
Message 7236 - Posted: 30 Dec 2007, 21:01:58 UTC

What is the difference between this one and the completed 715?
That one has got some bad stuff, something...


Post to thread

Message boards : Comments and discussion : 718