Hi people,
I've pushed another patch today to improve the scalability of particle insertion in LIGGGHTS. You can find a description here:
https://github.com/ParticulateFlow/LIGGGHTS-PUBLIC/pull/1
It's yet another one of those features I had to hold back for months. Now I'm free to publicly release it. Better late than never I guess...
Cheers,
Richard
| Attachment | Size |
|---|---|
 Scalability of insert/pack | 69.37 KB |
richti83 | Thu, 07/23/2015 - 11:01
Thank You !
Hi Richard,
thank you very much for releasing this improvement !
If anyone is interested, I've made a patch for current LIGGGHTS 3.2.1.
rberger | Thu, 07/23/2015 - 15:54
Great!
Hey Christian,
You're welcome. Thanks for backporting! :-)
Cheers,
Richard
shawnwuch | Tue, 07/28/2015 - 23:15
Thank you very much!
Thank you very much Richard and Christian,
I truly think this is a great improvement and is certainly very useful as I am trying to achieve ~ 0.5-1 million particles insertion in my CFDEM simulation. In the past, I often encounter errors during LIGGGHTS and OpenFOAM communication when the number of particle in insertion gets large. Hope this modification can help with that issue.
Christian, If possible, could you share the patch with me and instruct me on how I can modify my LIGGGHTS? Sorry, I am not really familiar with git features and it is my first attempt beyond simply clone and pull .... my email is shawnwu@utexas.edu should you need it.
Best regards,
Shawn
richti83 | Wed, 07/29/2015 - 10:08
files
Hi all, as I can not fork the public repo bc. I allready have one here are the files:
https://drive.google.com/file/d/0B16_X8nrVquiZ20zOVFLM1p5VEE/view?usp=sh... (patched source, 3.5MB)
https://drive.google.com/file/d/0B16_X8nrVquiMWdXMkNzd190d1E/view?usp=sh... (patch, 77kb)
https://drive.google.com/file/d/0B16_X8nrVquiYnlEbzJLTkFkMmM/view?usp=sh... (benchmark, 2kb)
The easiest way is to download the full patched source and replace in LIGGGHTS folder the original src folder, than run cfdemCompLIG and cfdemCompCFDEM again.
shawnwuch | Wed, 07/29/2015 - 18:49
Thank you
Thank you for sharing this!
ckloss | Mon, 07/27/2015 - 12:45
Hi Richard,
Hi Richard,
thanks - this is great!
Best wishes
Christoph
NTT1508 | Fri, 07/31/2015 - 01:23
For 3.2.1 Only ?
Many thanks Berger and Christian,
Are your uploaded files just for LIGGGHTS Version 3.2.1? I could not see bond_gran files there. Can I copy and replace these files to my current LIGGGHTS version 3.0.7.
Regards,
richti83 | Fri, 07/31/2015 - 10:04
No
The files I uploaded are only usable for liggghts 3.2.1. But Richards github fork bases on 3.0.7, so you can use these files directly.
https://github.com/ParticulateFlow/LIGGGHTS-PUBLIC/commit/1c86e65638008f...
richti83 | Fri, 07/31/2015 - 10:08
On the other hand
.. I think it will not break bond-feature when you copy the full src folder over yours, except christoph removed something necessary for bond gran when he cleaned up the src in 3.1.0 tag.
I'll update my bond gran fork at github after my summer holiday and add insert- and (if possible omp) patches.
Make a backup of your src folder and try :-)
NTT1508 | Mon, 08/03/2015 - 06:56
Bond_gran
Thank Christian,
I have tested bond_gran for versions 3.2.0 and 3.0.7. The results are not different, including bond_break.
Regards,
NTT1508 | Fri, 04/29/2016 - 03:04
Available in the recent versions ?
Hi Chiristian,
Can I know whether this improvement has been incorporated into the most recent version, i.e., LIGGGHTS 3.3 ?
Thank you,
Regards,
richti83 | Fri, 04/29/2016 - 07:33
Yes, see release notes of 3.3
Yes, see release notes of 3.3.0
Added an optimization/improvement to the performance of particle insertion. This breaks down the overlap check upon insertion from O(n^2) to O(n), which means considerable improvement for insertion of large # of particles, as described here: https://github.com/ParticulateFlow/LIGGGHTS-PUBLIC/pull/1 Thanks to Richard Berger (JKU) for the contribution of this very nice feature!
http://www.cfdem.com/liggghtsr-version-history