Respected Amber Community,
While I could not complete the dialanine peptide tutorial because of the reasons stated above, The pmemd.cuda.MPI command gave the following error :
```
gpu_neighbor_list_setup :: Small box detected, with <= 2 cells in one or more
dimensions. The current GPU code has been deemed
unsafe for these situations. Please alter the
cutoff to increase the number of hash cells, make
use of the CPU code, or (if absolutely necessary)
````
However, I was able to complete the proline pentapeptide abmd tutorial after using 40 cores on 8 walkers, using mpirun and pmemd.MPI. The run took about 2.5 hours. I ran the nfe-umbrella-slice to generate the FE.dat file, and then used the plot.sh​ file to generate the plot, however, that gave the image :
https://imgur.com/MfnFsG
[
https://i.imgur.com/MfnFsGJ.png?fb]<
https://imgur.com/MfnFsGJ>
imgur.com<
https://imgur.com/MfnFsGJ>
Imgur: The magic of the Internet
imgur.com
Which did not match with the one in the tutorial. However, when I plot the data in FE.dat using matplotlib, it gave me the following image :
https://imgur.com/rhSHe6A
[
https://i.imgur.com/rhSHe6A.png?fb]<
https://imgur.com/rhSHe6A>
imgur.com<
https://imgur.com/rhSHe6A>
Imgur: The magic of the Internet
imgur.com
Which matches very closely with the tutorial. The questions that I'd like to ask, are as follows :
1. Has anyone encountered the vlimit error before and managed to fix it?
2. Why doesn't pmemd.cuda.MPI work, or kindly explain how to fix the small box error, or any workarounds?
3. Is 2.5 hours for a run on 40 cores for 8 walkers too long for a small system like proline pentapeptide?
4. Is there any issue with how I ran the gnuplot command ? Here is my FE.dat file output :
https://pastebin.pl/view/108d6d2a
Any help would be appreciated.
Regards,
Shaunak Badani,
4th Year Undergraduate at IIIT Hyderabad
________________________________
From: SHAUNAK BADANI
Sent: Tuesday, September 29, 2020 2:45 PM
To: amber.ambermd.org <amber.ambermd.org>
Subject: Query regarding Dialanine peptide 2d ABMD tutorial in the NFE category
Respected AMBER community,
I was following the ABMD dialanine peptide tutorial for generating the 2D Free energy profile using multiple walker method. I was able to successfully complete the first step, however, in the second step of the tutorial where 4 walkers are used, I encountered the following error :
```
Running multipmemd version of pmemd Amber18
Total processors = 40
Number of groups = 4
vlimit exceeded for step ******; vmax = 26.2469
vlimit exceeded for step ******; vmax = 26.0264
Program received signal SIGSEGV: Segmentation fault - invalid memory reference.
```
I looked in the amber docs if there was some way to set vmax, but I did not find anything.
I have tried this simulation twice, and it has given me the same error at both times.
All I have changed from the tutorial is the number of processors I have used in MPI, and the normal version of pmemd.MPI instead of pmemd.cuda.MPI :
mpirun -np 40 pmemd.MPI -O -ng 4 -groupfile groups
Instead of
mpirun -np 4 pmemd.cuda.MPI -O -ng 4 -groupfile groups
I am currently running Amber18, instead of Amber16 software used in the tutorial.
Could anyone guide me on how to fix this error, or what could be the potential source of it? Any help would be highly appreciated.
Regards,
Shaunak Badani,
4th Year Undergraduate at IIIT Hyderabad
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Oct 01 2020 - 13:30:03 PDT
