[AMBER] usage of nmr (between COMs) restraints with pmem.cuda

From: Dmitry Karlov <dkar89.gmail.com>
Date: Fri, 25 Sep 2020 00:50:48 +0300

Dear Amber community,
I am trying to perform SILCS simulation (https://www.ncbi.nlm.nih.gov/
pmc/articles/PMC4685950/) using Amber18 (all patches have been applied).
Currently, I prepared a water box containing benzene, propane, acetate.
methylammonium, formamide and methanol molecules distributed across the box
(400 copies for each item). I tried to use nmr restraints to prevent
aggregation between benzene/benzene, benzene/propane and propane/propane
molecules. The examples of restraint syntax used you can see below:
geometric center of carbon atoms of benzene 1 / geometric center of carbon
atoms of benzene 2
&rst iat=-1,-1, r1=6.0, r2=7.5, r3=10000., r4=20000., rk2=15., rk3=50.,
igr1 = 1,2,3,4,5,6, igr2 = 13,14,15,16,17,18, &end
geometric center of carbon atoms of benzene 1 / central carbon atom of
propane 2
&rst iat=-1,-1, r1=6.0, r2=7.5, r3=10000., r4=20000., rk2=15., rk3=50.,
igr1 = 3553,3554,3555,3556,3557,3558, igr2 = 7111,0,0,0,0,0, &end
central carbon atom of propane 1 / central carbon atom of propane 2
&rst iat=6462,7584, r1=6.0, r2=7.5, r3=10000., r4=20000., rk2=15., rk3=50.,
&end

The meaningful is only lower bound while the upper bound I set was big
enough compared to box size

On this way I get about 500000 individual restraints and it works fine
serial version of amber. But at the same time the gpu version of pmemd do
not work with more than 30 individual restraints yielding an error:
cudaMemcpy GpuBuffer::Download failed an illegal memory access was
encountered.
Do you have any idea about how to deal with it?





-- 
DK
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Sep 24 2020 - 15:00:03 PDT
Custom Search