Wow really nice test ! Add a red inner core for blood
Like el_diablo, I’m curious to see some stats about it ( mainly about subframe because you have high velocity implied here).
I can put it on my blog here? : http://pyroevil.com/2013/05/09/post-you-videos-here/
I love Suzanne so much, and I just want to kill her in a thousand ways. :evilgrin:
Simulation time around 2h 15min on i7-3770k (3.4GHz) 100 frames
Thank you! ^^ Yeeeah, Inner blood! That’s actually my next step! XD I’ll try to do it combining SPH + molecular script, can they work together, right?
Yes it’s fast, I started testing using 80 subframes, but finally I used around 100 and 150 subframes in the different tests you see in the video.
In order to slow down the whole simulation I decreased gravity and particle speed by half, but I think that’s not the ideal way; I wanted to make a super slow motion shot, but didn’t find how to do it, however while simulating I saw the process subframe by subframe and it looks awesome! I would love if the script lets you save 1 cache frame every X subframe, instead of saving every 1 full frame, do you think that’s possible?
You have a “time remapping” in render panel where you can play with to slow it down.
I make a timescale option in the past but it’s not working well , probably I can put it back if I find a solution for it.
I am making a Jelly fish, but currently having trouble connecting the tentacles to the head, Would I be able to use this plugin to solve that issue. And if so, how would I transfer the animation of the molecules to the tentacle geometry.
Hi Pyroevil, your videos are awesome like Multiple Cloth etc.
Yes, I tried to get something like ceramics, but other setups looked like clay or rubber …
Can you post some file?
Your monkey head is close to what I’m trying to achieve.
Thank you
Just to add to what i mentioned before, i have stabilised the simulation now, no more crashes (fingers crossed) i did this by increasing the scale of the source object, and decreasing the rendered particle size to compensate, not sure why this works but it does.
Anyway onto today’s question. The machine i’m running it on, should bake quickly, but it maxes out its ram after about 200 frames or so. now that’s 32GB ram gone very quickly. i’m using about 65000 particles, made up from 3 source objects grouped together. they’re not too complex a shape either, just standard ico-spheres slightly modified. i even lowered the substep own to 1 to see if that made any difference, but it doesn’t
do you know of any memory leak problems?, or is it normal to max out that much ram so quickly? i can probably attach the file somewhere if you want to look.
I get a stable version now where I never get crashs for about two weeks.
I did some optimisation of code and algorytme ( get near 2 times faster ) and I continue to improve it.
Now I try to multithread it entirely ( before only 60-70% of the code are multhithreaded )before releasing a new version.
Maverick: I never get memory issues( I simulate manytime more than 200k particles with only 6gig of ram ). Always check the option to cache your particles on disk because you max out your memory for sure quickly. This the only case where I explode my ram.
I do have cache enabled, and have it set to 1 step, however i have the particles colliding with an object with a complicated surface (30000 surfaces) would this be the reason?, if so is it just a case of reducing the surfaces to reduce RAM usage?
Hi !
I just wrote this little post to tell the developpement on Molecular Script is not dead. I still working on it and the progress is really great. I fix all crash now ( I not crash at all after more than 6 weeks of test ) and improve a lot the speed of it. More improvement still coming but for now here is the stats of where I’m :
Numbers of Particles: 1 000 000
Numbers of links created: 15 671 796
Original code from beta release 1 may 2013: 12.98 sec per subframe
Light structure of particles for kdtree creation : 9.02 sec per subframe
Multithread particles collision and links solving: 6.55 sec per subframe
Code my own quick_sort algorythm function: 5.57 sec per subframe
For now the script is a bit more than 2 times faster. More improvement still going and I think for now I can reach 3 sec per subframes. So we talking about 4 times faster than before if this incoming improvement is a success. Sorry for the slow dev speed and stay tuned !
Hi !
For waiting until the stable/optimised release of my script , here is a 1920 x 1080 wallpaper of a test simulation. 4 millions particles , 256 millions of polygones , 30 minutes per frames to simulate and 40 minutes to render. Enjoy !
New candidate version available !
2 times faster and tested about 2months without crash.
Test it and let’s me know about issues ! ( for people don’t be able to active it , I don’t fix the problem because I’m not be able to reproduce, sorry )
I put the OSX version on hold. Until I find a volunteer to compile it.
I use Cython 0.19 and python 3.3.0 ( to match the python version in Blender ).
Sorry , I don’t have mac to maintain it myself
Here I store in Angular_Velocity of the particles the UVW coordinate. I retrive it with Particles Info nodes in Cycles and use UV to map the color_grid textures with the unwrap UVs from the emmiter. Because the W is not used is 2D textures , I store in it the distance of the particles from the surface. So I can use it two create two material: One for the outer ( color_grid textures with diffuse shader) and one for the inner ( orange glowing with emmision shader ).