OK - this time I really really think I've got it fixed.
https://github.com/Amber-MD/cpptraj/pull/985
Version is 6.14.0 - please beat it up as much as you can. I've tested
with my own files and the files you sent me and everything looks like
it makes sense now, but I've thought that before...
As a bonus, I've now openmp-parallelized the diffusion and unwrap
calcs (cpptraj.OMP), and sped up the unwrap calc by just over 2x by
only calculating the unwrap vector when the entity in question has
traveled more than half the box length in any dimension. The unwrap
results are consistent with previous cpptraj results as far as I can
tell, but again tell me if you see anything weird.
Thanks again for reporting this!
-Dan
On Tue, Aug 9, 2022 at 7:08 PM Cianna Calia <ccalia.ucsd.edu> wrote:
>
> Ah, yeah I just thought of that - The traj I sent you has frames spaced far apart just because I was trying to make the file size smaller for easy sending and I forgot that this might create a different issue, oops. The original 100 ns traj I was testing this on had frames saved every 2 ps. In the meantime I'll see if I can find a segment of my traj that shows the problem but is short enough that I can send it with the full density of frames without the file being extremely big. Sorry about that.
>
> Thanks,
> Cianna
>
> On Tue, Aug 9, 2022 at 3:42 PM Daniel Roe <daniel.r.roe.gmail.com> wrote:
>>
>> Hi,
>>
>> Thanks for the files!
>>
>> So one difference I see right off the bat is that in your test
>> trajectory you're saving frames every 400 ps while in my test
>> trajectories I was saving frames every 0.5-10 ps at most. I will look
>> in more detail tomorrow but it's possible the imaging code in the
>> "diffusion" action (which is a bit simpler than the code in the
>> "unwrap" action) is getting confused by the much larger distance jumps
>> in-between frames (due to the lower frequency at which the trajectory
>> is being written). I'll dig into this more tomorrow.
>>
>> -Dan
>>
>> On Tue, Aug 9, 2022 at 3:58 PM Cianna Calia <ccalia.ucsd.edu> wrote:
>> >
>> > I tried V6.13.0 by doing the following:
>> > git clone https://urldefense.com/v3/__https://github.com/Amber-MD/cpptraj.git__;!!Mih3wA!EGutLqXWJ4wRIIkndG8fvlbQIUef5NXczKaxgob--QNevHQ_mc81T-7rtD3pIXtkRUgL3BAZUhQHf3_Fdy5M$
>> > Inside the cpptraj folder:
>> > ./configure gnu
>> > make install
>> >
>> > On Tue, Aug 9, 2022 at 12:44 PM Daniel Roe <daniel.r.roe.gmail.com> wrote:
>> >>
>> >> On Tue, Aug 9, 2022 at 3:15 PM Cianna Calia <ccalia.ucsd.edu> wrote:
>> >> >
>> >> > Do diff.1.dat and diff.2.dat come out the same for you?
>> >>
>> >> Yes, they're identical except for the header:
>> >>
>> >> $ diff diff.1.dat diff.2.dat
>> >> 1c1
>> >> < #Time D1[R] D1[X] D1[Y] D1[Z] D1[A]
>> >> ---
>> >> > #Time D2[R] D2[X] D2[Y] D2[Z] D2[A]
>> >>
>> >> > My current test system has 9620 waters; the 100 ns trajectory is > 23 GB which makes attaching it a little hard. Let me see if I can reproduce the issue using a shorter, sparser version of my traj and if I can then I will share that.
>> >>
>> >> Understood, thanks! I can also look into trying to make my traj
>> >> available to you to try. How was your cpptraj compiled?
>> >>
>> >> -Dan
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Aug 10 2022 - 13:30:03 PDT
