 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Hello!
I tried to work on a new version of "PvMegaPOV".
So I changed the code of Megapov and added the functions of PVMPOV to it.
Now I have a working version of MegaPOV with the PVM extension. It's working
but I'm wondering about one thing:
One of the new features of MegaPOV is presentated in "cloth.pov".
I tried to render it and figured out that the parsing and building the
".dat" 's are the most CPU-power-intensive things of the hole process.
Anyone have some idea/solution how to distribute it on different nodes? Is
there a possibility to do so? Rendering the "cloth.pov" would be really
faster if this would be possible.
MfG
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On Wed, 21 May 2003 17:52:00 +0200, "Daniel Genrich" <dan### [at] inventivesoft com>
wrote:
> Anyone have some idea/solution how to distribute it on different nodes? Is
> there a possibility to do so? Rendering the "cloth.pov" would be really
> faster if this would be possible.
Christoph will be more accurate on this but AFAIK this kind of simulation can't
be parallelized by rendering different frames on different machines. Simulation
is a highly sequential process, the calculations for each frame depend on those
from the frame before.
ABX
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Daniel Genrich wrote:
>
> One of the new features of MegaPOV is presentated in "cloth.pov".
> I tried to render it and figured out that the parsing and building the
> ".dat" 's are the most CPU-power-intensive things of the hole process.
> Anyone have some idea/solution how to distribute it on different nodes? Is
> there a possibility to do so? Rendering the "cloth.pov" would be really
> faster if this would be possible.
Simulations like Clothray or Mechsim can be parallelized very well but you
have to do this on the level of the simulation loop. Otherwise it will be
like trying to build a house's roof before the walls. Of course parallel
calculations on this level will require a lot of communication between the
processes - the data will have to be synchronized after every simulation
step.
Christoph
--
POV-Ray tutorials, include files, Sim-POV,
HCR-Edit and more: http://www.tu-bs.de/~y0013390/
Last updated 28 Feb. 2003 _____./\/^>_*_<^\/\.______
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Hello!
I'm just searching the code and wondering, why the function
"Iterate_Cloth_Simulation" -clothray.cpp- has two parameters instead of one.
I can't find, where the second one ("vector * resul") is used.
I need this to figure out, how and where I can try to parallelize the code.
("clothray.cpp" seems easier than "mechsim.cpp" to me for this ;-) )
Thank you for any information about this!
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On Thu, 22 May 2003 18:27:43 +0200, "Daniel Genrich" <dan### [at] inventivesoftcom>
wrote:
> I'm just searching the code and wondering, why the function
> "Iterate_Cloth_Simulation" -clothray.cpp- has two parameters instead of one.
> I can't find, where the second one ("vector * resul") is used.
> I need this to figure out, how and where I can try to parallelize the code.
> ("clothray.cpp" seems easier than "mechsim.cpp" to me for this ;-) )
It is used in macro AccessResul (defined earlier) which is used in
Iterate_Cloth_Simulation. Otherwise no usage would be reported by compilers
which I have set to highiest possible warning level to make sources warning
free.
ABX
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Daniel Genrich wrote:
>
> Hello!
>
> I'm just searching the code and wondering, why the function
> "Iterate_Cloth_Simulation" -clothray.cpp- has two parameters instead of one.
> I can't find, where the second one ("vector * resul") is used.
> I need this to figure out, how and where I can try to parallelize the code.
> ("clothray.cpp" seems easier than "mechsim.cpp" to me for this ;-) )
In Clothray you can distribute the content the main double loop in
'Iterate_Cloth_Simulation()'. But after every execution of the loop the
content of the position and velocity arrays have to be synchronized
between the nodes. I doubt this would be efficient because this is a lot
of data and one simulation step does not take very long.
Mechsim can be handled similarly but i would suggest you wait for the next
version because there will be a change in the code structure (the
different integration methods will be in separate functions).
Christoph
--
POV-Ray tutorials, include files, Sim-POV,
HCR-Edit and more: http://www.tu-bs.de/~y0013390/
Last updated 28 Feb. 2003 _____./\/^>_*_<^\/\.______
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
*smile* Take a look at line 333 in the file "parse.cpp" of Megapov.
----
imax = l;
for(i = 1; i <= imax; i++)
{
----
:-)
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On Wed, 4 Jun 2003 19:39:44 +0200, "Daniel Genrich" <dan### [at] inventivesoftcom>
wrote:
> *smile* Take a look at line 333 in the file "parse.cpp" of Megapov.
> ----
>
> imax = l;
> for(i = 1; i <= imax; i++)
> {
>
> ----
Is this a joke ? I do not get it. Everything seems fine in these two lines, and
it is not MegaPOV specific, so ?
ABX
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
ABX <abx### [at] abxartpl> wrote:
> On Wed, 4 Jun 2003 19:39:44 +0200, "Daniel Genrich" <dan### [at] inventivesoftcom>
> wrote:
> > *smile* Take a look at line 333 in the file "parse.cpp" of Megapov.
> > ----
> >
> > imax = l;
> > for(i = 1; i <= imax; i++)
> > {
> >
> > ----
>
> Is this a joke ? I do not get it. Everything seems fine in these two
> lines, and it is not MegaPOV specific, so ?
There is nothing wrong here.
I guess he is reading imax = l ('el' L) as imax = 1 (one) :-)
This can happen with some fonts.
yvo
--
MacMegaPOV at:
http://users.skynet.be/smellenbergh
E-mail: yvo### [at] gmxnet
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Thank you for the hint.
I was really wondering about some lines of the code..I better change the
font :-)
cu
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
