Moleculars physics

yes , so I need volunteers to compile it because I don’t have MAC at home.
Somebody told me probably extension compile in linux can work on MAC because OSX is linux based.

depends on libraries , but generally speaking it may work :smiley:

anyone want to fellow my progress can subscribes to my youtubes channel or to this threads but now you can subscribes to the newsletter of my blog too ! Just enter your email on the sidebar.

beta version of my C version coming soon ( 30times faster , calculate 300k particles in 140sec !)

I did a stress test this weekend with 1millions of particles.
It’s took about a average of 10minutes per frame ( 2 days for a animation of 300frames ).
I thinking to got all render done this morning but it’s seem than the render never start because my video cards don’t have enough memory for render all that particles ( cuda memalloc error ) ! On the CPU only , it’s take about 6-7minutes per frames with a render samples of 96. My addons become to be faster than cycles :wink:

Here is a little test I did couples of days ago:

Now I need to work a bit on how particles react when collision happens between them.

that’s very cool - it’s great to see your progress here.

@Pyroevil
what a great new video demo! thanks for your work on this awesome tool!

impressive stuff as always!!

PS: Is there some news about a particle remesher for blender?

Top stuff. Could this work with metaballs and become - theoretically - a fluid simulation/flui solver too? Thanks sorry if it sounds basic.

Thanks for all comments ! @MohamedSakr currently work on a CUDA particles mesher. I think he working on some issues with CUDA and cards compatibility.

Soon , a beta version of my C addon is coming with a “teaser” video of it’s “features”.
The C version is 30times faster, multithread and have many features than the python version don’t have.
Now I try to compile it for all Blender version : Windows 32bit , Windows 64bit , Linux 32bit and Linux 64bit.
Yesterday I trying to install VirtualBox and Ubuntu as a virtual machine to compile on Linux. It’s work fine. I finish to wrote the docs today.

Thanks for all supporters and donators !

Good news ! Beta version is now available here:

More parameters is avaialble for the user and better physics integrated. And it’s 30times faster than the old python version ! Available for windows 32bit/64bit and Linux 32bit/64bit.

With link to documentation. Test it and let’s me know what you are thinking about it.
Soon I want to release a teaser video for BlenderNation and try to package zipped samples .blend files.

Enjoy !

sorry about anybody download the Windows 64bit version and have a runtime error when he’s start Blender.
It’s a microsoft manifest problem.
I fix it now so just go on my blog again and redownload and reinstall it.
sorry again

Installed fine - it seems to work ( crashed once when playing with it and hit ‘start sim’ twice in a row while it was already simulating.
There are so many options http://pyroevil.com/molecular-script-docs/ which is GREAT to document. But I was wondering, you have some great WIP videos , can you create a simple step by step tutorial to get some relatively neat effects?

awesome work PyroEvil :smiley:
it is crashing a little (I think it is initializing issue) when I bake and bake again after finish (so to overcome this i delete particle system and create another one)
it has now very good speed too,did you make the whole code multi Core?

about the mesher:
I’m still very busy (will take a month I think or 2 till I look at it and fix the bugs,kinda know where is the problem so it will be fixed :))

Thanks spatdsg and mohamedsakr!

I think a did a error and compile a previous code with memory problem.
Tomorrow I try to compile it with the fixed issue. You use the addons on which platform? Disabling the start button when already simulating can be a good things to integrate !

Mohamed , for now it’s not entire multithreaded for now. Some part need to space partitionning the particles for to be fully multithreaded. I will see in the future. I need to do more profiling befire to see if better part of the code can be optimised.

Okay I find time to look and my code is the uptodate one. I crash sometime too but not a lot. It’s difficult to fix this kind of bugs when it’s just crash 1 times on 20. If somebody have a simple blend file than crash a lot , just send me please to help me to debug.

About tutorials , I want to create a samples files zip package. With that I think it’s can help user to guess which parameters is used for a kind of effect. Simulation is a lot of guessing and you developp a fifth sense for that with a bit of testing.

thx pyro – i cant get it to crash reliably either. Thanks for the zip files idea - it would be appreciated.

one note though
when you finish the simulation,all data are (shaded) so I can’t try a new simulation unless I delete the particle system and create new one

but overall performance is great :smiley:
if you want any help I can make it parallel with OpenMP very easy (may add like 5 lines to your code to be parallel :D)

Mohamed , if you want to do a simulation again , you don’t have to delete your particles system, just push the “free all bake” button in the normal Blender particles parameters tab. I can repeat this button next to my “start simulation button”.

Yes I already use OpenMP in my code. it’s really easy. The problem is not here , is you have to space partionning your particles for make it “parrallisable”. 60% of my code use openmp ( kdtree creation , neighbours searching … etc). The only one missing is the collision and links calculation and for that I need to separate particles to make sure two CPU don’t do math on the same particles. I read a bit on that. It’s not too hard , I just need to find time for that. :slight_smile:
I want to use particles pointer like your script too to see if i can get speed improvement.

cool :smiley:
how do you make the collisions?iterate over the whole particles?