 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
I'm rendering an animation of a somewhat complex landscape based on isosurfaces,
and rendertimes vary between 30 minutes and infinite.
I have taken the usual precautions to limit number of reflections, max_gradient
and so forth, but still a few individual frames appear to never finish, the
parallel povray processes chewing away at the last few tens of pixels.
My problem is now that I do not want to change the scene files since most of the
animation is already done and only 5 frames missing. An easy mitigation would be
to render those 5 as far as one can get, and use gimp or whatever to replace the
missing few pixels from the frames before and after - much like in standard
movie restoration.
The problem is - povray wouldn't let me have those. All it produces is a
pov-state file, and whenever I initiate any kind of abort, no image can be found
on disk (3.7, the version).
So my question is: Is there a way to abort the render such that I have the
partially (almost fully, in fact) rendered image on disk?
What would be even better still - is there a possibility to set a hard limit to
the render time of individual frames in an animation, after which povray would
give up on that frame, put the image to disk as it is, and continue with the
next one? That would be really useful...
m.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 25/07/2014 14:48, mfeldt wrote:
> I'm rendering an animation of a somewhat complex landscape based on isosurfaces,
> and rendertimes vary between 30 minutes and infinite.
>
> I have taken the usual precautions to limit number of reflections, max_gradient
> and so forth, but still a few individual frames appear to never finish, the
> parallel povray processes chewing away at the last few tens of pixels.
>
> My problem is now that I do not want to change the scene files since most of the
> animation is already done and only 5 frames missing. An easy mitigation would be
> to render those 5 as far as one can get, and use gimp or whatever to replace the
> missing few pixels from the frames before and after - much like in standard
> movie restoration.
>
> The problem is - povray wouldn't let me have those. All it produces is a
> pov-state file, and whenever I initiate any kind of abort, no image can be found
> on disk (3.7, the version).
>
>
> So my question is: Is there a way to abort the render such that I have the
> partially (almost fully, in fact) rendered image on disk?
No. But the pov-state file contains the picture you need... in internal
format. (and with a few missing squares, the ones that have not been
fully rendered)
Is the scene short-enough to be used in a bug report ?
> What would be even better still - is there a possibility to set a hard limit to
> the render time of individual frames in an animation, after which povray would
> give up on that frame, put the image to disk as it is, and continue with the
> next one? That would be really useful...
>
--
IQ of crossposters with FU: 100 / (number of groups)
IQ of crossposters without FU: 100 / (1 + number of groups)
IQ of multiposters: 100 / ( (number of groups) * (number of groups))
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 25/07/2014 16:26, Le_Forgeron wrote:
> On 25/07/2014 14:48, mfeldt wrote:
>>
>> So my question is: Is there a way to abort the render such that I have the
>> partially (almost fully, in fact) rendered image on disk?
>
> No. But the pov-state file contains the picture you need... in internal
> format. (and with a few missing squares, the ones that have not been
> fully rendered)
>
There is an easier format, in another file, if your frame is large
enough to trigger the file-buffered storage.
The setting is Max_Image_Buffer_Memory (MI), default 128 megabytes.
Minimal value to have a file is 1 megabyte (MI1), and
height*width*20/1048576 must then be strictly greather than the value
(so height*width*20 >= 2*1048576 , or h*w >= 104857 for -MI1)
The format is 5 floats per pixel, with the end of the file containing
the height, width and size of pixel. The five floats are the five RGBFT
channel, in that order.
--
IQ of crossposters with FU: 100 / (number of groups)
IQ of crossposters without FU: 100 / (1 + number of groups)
IQ of multiposters: 100 / ( (number of groups) * (number of groups))
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Le_Forgeron <jgr### [at] free fr> wrote:
> On 25/07/2014 14:48, mfeldt wrote:
> > I'm rendering an animation of a somewhat complex landscape based on isosurfaces,
> > and rendertimes vary between 30 minutes and infinite.
> >
> > I have taken the usual precautions to limit number of reflections, max_gradient
> > and so forth, but still a few individual frames appear to never finish, the
> > parallel povray processes chewing away at the last few tens of pixels.
> >
> > My problem is now that I do not want to change the scene files since most of the
> > animation is already done and only 5 frames missing. An easy mitigation would be
> > to render those 5 as far as one can get, and use gimp or whatever to replace the
> > missing few pixels from the frames before and after - much like in standard
> > movie restoration.
> >
> > The problem is - povray wouldn't let me have those. All it produces is a
> > pov-state file, and whenever I initiate any kind of abort, no image can be found
> > on disk (3.7, the version).
> >
> >
> > So my question is: Is there a way to abort the render such that I have the
> > partially (almost fully, in fact) rendered image on disk?
>
> No. But the pov-state file contains the picture you need... in internal
> format. (and with a few missing squares, the ones that have not been
> fully rendered)
>
> Is the scene short-enough to be used in a bug report ?
There is a simple solution: POV-Ray knows nothing about the content of the scene
file. So it is possible to use an empty scene file to continue a trace...
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 25/07/2014 20:51, Thorsten Froehlich wrote:
> There is a simple solution: POV-Ray knows nothing about the content of the scene
> file. So it is possible to use an empty scene file to continue a trace...
>
Intriguing! Could you expand on that, please?
When I get a scene that takes ages to finish rendering the last block. I
reduce the block size to the smallest so that I can get more cores
working on it.
Won't help if a pixel refuses to render but does speed up some files.
--
Regards
Stephen
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> No. But the pov-state file contains the picture you need... in internal
> format. (and with a few missing squares, the ones that have not been
> fully rendered)
>
> Is the scene short-enough to be used in a bug report ?
>
I'm afraid right now, I cannot really isolate the offending code, especially sic
it only happens in about 3 out of 300 frames.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
"Thorsten Froehlich" <nomail@nomail> wrote:
> Le_Forgeron <jgr### [at] freefr> wrote:
> > On 25/07/2014 14:48, mfeldt wrote:
> > > I'm rendering an animation of a somewhat complex landscape based on isosurfaces,
> > > and rendertimes vary between 30 minutes and infinite.
> > >
> > > I have taken the usual precautions to limit number of reflections, max_gradient
> > > and so forth, but still a few individual frames appear to never finish, the
> > > parallel povray processes chewing away at the last few tens of pixels.
> > >
> > > My problem is now that I do not want to change the scene files since most of the
> > > animation is already done and only 5 frames missing. An easy mitigation would be
> > > to render those 5 as far as one can get, and use gimp or whatever to replace the
> > > missing few pixels from the frames before and after - much like in standard
> > > movie restoration.
> > >
> > > The problem is - povray wouldn't let me have those. All it produces is a
> > > pov-state file, and whenever I initiate any kind of abort, no image can be found
> > > on disk (3.7, the version).
> > >
> > >
> > > So my question is: Is there a way to abort the render such that I have the
> > > partially (almost fully, in fact) rendered image on disk?
> >
> > No. But the pov-state file contains the picture you need... in internal
> > format. (and with a few missing squares, the ones that have not been
> > fully rendered)
> >
> > Is the scene short-enough to be used in a bug report ?
>
> There is a simple solution: POV-Ray knows nothing about the content of the scene
> file. So it is possible to use an empty scene file to continue a trace...
Interesting. I'll give that a try, even though it still means that the
following frames don't even get started until manually halting the renderer.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Stephen <mca### [at] aolcom> wrote:
> On 25/07/2014 20:51, Thorsten Froehlich wrote:
> > There is a simple solution: POV-Ray knows nothing about the content of the scene
> > file. So it is possible to use an empty scene file to continue a trace...
> >
>
> Intriguing! Could you expand on that, please?
>
> When I get a scene that takes ages to finish rendering the last block. I
> reduce the block size to the smallest so that I can get more cores
> working on it.
> Won't help if a pixel refuses to render but does speed up some files.
>
> --
>
> Regards
> Stephen
I do the same, but still - if in an animation of 350 frames the 264th hangs, a
lot of CPU power gets wasted.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
BTW, I am now running from an external script that provides a timeout and use
the reconstruction of pov-state files with empty scenes. Works at least.
I also found that getting stuck depends somewhat on the processor type - not all
frames get stuck on all machines. In my sequence of 275 I only have 2 that
always refuse to render completely, the others hang only on one of three
possible machines... weird!
m.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
"Thorsten Froehlich" <nomail@nomail> wrote:
> Le_Forgeron <jgr### [at] freefr> wrote:
> > On 25/07/2014 14:48, mfeldt wrote:
> > > I'm rendering an animation of a somewhat complex landscape based on isosurfaces,
> > > and rendertimes vary between 30 minutes and infinite.
> > >
> > > I have taken the usual precautions to limit number of reflections, max_gradient
> > > and so forth, but still a few individual frames appear to never finish, the
> > > parallel povray processes chewing away at the last few tens of pixels.
> > >
> > > My problem is now that I do not want to change the scene files since most of the
> > > animation is already done and only 5 frames missing. An easy mitigation would be
> > > to render those 5 as far as one can get, and use gimp or whatever to replace the
> > > missing few pixels from the frames before and after - much like in standard
> > > movie restoration.
> > >
> > > The problem is - povray wouldn't let me have those. All it produces is a
> > > pov-state file, and whenever I initiate any kind of abort, no image can be found
> > > on disk (3.7, the version).
> > >
> > >
> > > So my question is: Is there a way to abort the render such that I have the
> > > partially (almost fully, in fact) rendered image on disk?
> >
> > No. But the pov-state file contains the picture you need... in internal
> > format. (and with a few missing squares, the ones that have not been
> > fully rendered)
> >
> > Is the scene short-enough to be used in a bug report ?
>
> There is a simple solution: POV-Ray knows nothing about the content of the scene
> file. So it is possible to use an empty scene file to continue a trace...
BTW - using a completely empty scene file is *not* always the best solution, as
apparently pov-state does not contain *all* the information.
If e.g. display_gamma is used in the original scene file, that info does not get
transmitted and the result may actually look completely different!
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
|
|
