 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
tho### [at] trf de (Thorsten Froehlich) wrote in <395a47bc@news.povray.org>:
>In article <8F617F84Elavendersmith@204.213.191.228> ,
>lav### [at] hotmailcom (lavender) wrote:
>
>> But Povray in all of its power can not even
>> brake up the image into chunks and pass those to nodes to render and
>> return the part of the image. This IS a simple venture. I have built
>> an external system to do just this but it is clumsey. Even simpler is
>> to send frames out to different systems. This would help also.
>
>While this is easy to implement on _one_ platform, doling it on multiple
>platforms is an extreme amount of work. How many of the "professional"
>programs can do this on more than just a few Unix platforms? The
>answers is simple: None. Pixar added Windows NT support for RenderMan
>just a short time ago, and they have hundreds of millions of dollars
>(after the movies they made for Disney). There is no standard API for
>networking defined in either the C or C== standard, even Unix has
>Sockets and Streams.
>
would it be possible to use CORBA to do this?
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Chris Huff wrote:
>It is possible to render a portion of an image, but getting the images
>tiled without discontinuities is a problem. If you distribute parts of
>an image to different machines this way, features like radiosity,
>antialiasing, etc can produce a "tiles" effect, like a mosaiac
>photograph.
>
What if you make the portions a bit bigger than needed, so the "tiles"
overlap?
Ingo
--
Photography: http://members.home.nl/ingoogni/
Pov-Ray : http://members.home.nl/seed7/
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
In article <8F619D6A0dashixpressnet@204.213.191.228>, das### [at] x-pressnet
(daishi) wrote:
> but if you use the same settings on all the machines it shouldn't be an
> issue, should it?
No, because radiosity is view dependant, and also because it relies on
randomized sampling, caching data, as well as generating data on the
fly, each square will have shading which is a little bit different from
it's neighbors. Similar problems may exist for other features.
--
Christopher James Huff - Personal e-mail: chr### [at] maccom
TAG(Technical Assistance Group) e-mail: chr### [at] tagpovrayorg
Personal Web page: http://homepage.mac.com/chrishuff/
TAG Web page: http://tag.povray.org/
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On Wed, 28 Jun 2000 20:57:13 +0200, Thorsten Froehlich wrote:
>The difference between Win 9x and Win NT/2000 is the quality (and
>probability method) of the scheduling algorithms, but they are still both
>pre-emptive.
>In fact this is one of the major improvements in Win 9x since Win 3.x.
9x is still dependent on a lot of 16-bit code, which is to a certain extent
cooperative. 16-bit apps running on 9x cooperate with each other, and the
entire 16-bit subsystem (if you can call it that) is one process from the
point of view of the 32-bit scheduler.
NT has the same arrangement, except for the dependence on 16-bit code,
unless you set up your 16-bit apps to run in separate processes (this is
not the default.)
So 9x and NT are mostly preemptive, but they still have a cooperative
system inside for legacy apps.
--
Ron Parker http://www2.fwi.com/~parkerr/traces.html
My opinions. Mine. Not anyone else's.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
I've been following this thread for a while and have been playing with
raytracing for a few years.
I wonder if it would be possible for a single machine to generate the
radiosity map and then pass it around. This would solve the randomization
issues. just passing a seed value would only work if all the machines had
the same radomize function.
Alan
"Chris Huff" <chr### [at] maccom> wrote in message
news:chrishuff-CE60D9.14082128062000@news.povray.org...
> In article <8F619D6A0dashixpressnet@204.213.191.228>, das### [at] x-pressnet
> (daishi) wrote:
>
> > but if you use the same settings on all the machines it shouldn't be an
> > issue, should it?
>
> No, because radiosity is view dependant, and also because it relies on
> randomized sampling, caching data, as well as generating data on the
> fly, each square will have shading which is a little bit different from
> it's neighbors. Similar problems may exist for other features.
>
> --
> Christopher James Huff - Personal e-mail: chr### [at] maccom
> TAG(Technical Assistance Group) e-mail: chr### [at] tagpovrayorg
> Personal Web page: http://homepage.mac.com/chrishuff/
> TAG Web page: http://tag.povray.org/
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
tho### [at] trfde (Thorsten Froehlich) wrote in <395a4a62@news.povray.org>:
>In article <8F6177D69dashixpressnet@204.213.191.228> , das### [at] x-pressnet
>(daishi) wrote:
>
>> 'partially' which means win98 is cooporative......(just like mac OS,
>> not counting X of course)
>
>No, it is pre-emptive with a monolithic operating system. If you write
>a simple application that just executes a loop lets say 10 billion
>times, will the other applications still get a share of the processor?
>Yes, because the operating system will interrupt it after a few
>milliseconds and then switch to another application. This is
>pre-emptive multitasking.
>
>On a Mac with i.e. Mac OS 9 an application executing the same program
>will not share the processor with other applications until it calls the
>operating system and allows it. This is co-operative multitasking.
>
>The difference between Win 9x and Win NT/2000 is the quality (and
>probability method) of the scheduling algorithms, but they are still
>both pre-emptive.
>In fact this is one of the major improvements in Win 9x since Win 3.x.
>
>
> Thorsten
>
>
>PS: If you don't believe me, get any recent CS textbook on operating
>systems, and it will tell you Win 9x and NT both use pre-emptive
>multitasking :-)
>
>
>____________________________________________________
>Thorsten Froehlich, Duisburg, Germany
>e-mail: tho### [at] trfde
>
>Visit POV-Ray on the web: http://mac.povray.org
>
while I agree with what your saying. there was an article a while ago
discussing NT vs win98 that said win98 was cooporative. I can't seem to
find the article though...
win98s scheduling needs improvement then, cause compared to other multi-
tasking OSs it sucks..
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 28 Jun 2000 15:42:35 -0400, daishi wrote:
>while I agree with what your saying. there was an article a while ago
>discussing NT vs win98 that said win98 was cooporative. I can't seem to
>find the article though...
>
>win98s scheduling needs improvement then, cause compared to other multi-
>tasking OSs it sucks..
It's no more cooperative than NT. It sucks because it switches back and
forth from the (mostly non reentrant) 16-bit core a lot.
--
Ron Parker http://www2.fwi.com/~parkerr/traces.html
My opinions. Mine. Not anyone else's.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Thorsten Froehlich wrote:
>
> In article <8F6177D69dashixpressnet@204.213.191.228> , das### [at] x-pressnet
> (daishi) wrote:
>
> > 'partially' which means win98 is cooporative......(just like mac OS, not
> > counting X of course)
>
> No, it is pre-emptive with a monolithic operating system. If you write a
> ...
> In fact this is one of the major improvements in Win 9x since Win 3.x.
>
> Thorsten
But, under windows 3.1, separate MS-DOS boxes were preemptively
multitasked. Goes back farther that you might think.
Win 9x crashes are typically due to bad programs writing into memory
regions that are accessible to all running apps. And the reason for
those common regions are to retain backwards compatability...
Xander
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
chr### [at] maccom (Chris Huff) wrote in <chrishuff-
CE60D9.14082128062000@news.povray.org>:
>In article <8F619D6A0dashixpressnet@204.213.191.228>, das### [at] x-pressnet
>(daishi) wrote:
>
>> but if you use the same settings on all the machines it shouldn't be an
>> issue, should it?
>
>No, because radiosity is view dependant, and also because it relies on
>randomized sampling, caching data, as well as generating data on the
>fly, each square will have shading which is a little bit different from
>it's neighbors. Similar problems may exist for other features.
>
okay, so do it by line insted of by tile. send the first 10 lines to one
comp, next 10 to another, etc. only problem you might run into is
reassembling the images in the right order. pov-ray already supports this
with its resume feature, just no one has written a way to do it
distributedly
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 28 Jun 2000 18:40:05 -0400, daishi wrote:
>okay, so do it by line insted of by tile. send the first 10 lines to one
>comp, next 10 to another, etc. only problem you might run into is
>reassembling the images in the right order. pov-ray already supports this
>with its resume feature, just no one has written a way to do it
>distributedly
Same problem. Believe us, we've actually thought about this problem and
it's not as simple as you think. (Why do I find myself saying that more
and more these days?)
--
Ron Parker http://www2.fwi.com/~parkerr/traces.html
My opinions. Mine. Not anyone else's.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
