Chris Neale wrote:
contrary to my previous stament, I am no longer convinced that it is
simply an initial constraints issue.
There are two things to do:
1. run gmx3.3 and gmx4.0 with the 3.3 tpr file. This should give the
same result in principle.
2. run gmx3.3 and gmx4.0 with their own tpr files. Here you can use
gmxcheck -s1 -s2 to compare the files.
The differences due to initial constraints are small usually, but they
seem comparable in your case. Defining flexible should remove all
difference due to constraints (unless constraints are turned on in the
mdp file which selects LINCS over the default SETTLE).
Anyway it is important to distinguish the apples from the oranges, hence
point 1 is the best to test it. Obviously you are aware that the MW
velocity does not in anyway influence the energy or the dynamics, since
it merely shows the displacement of the particle divided by the time
step. Therefore it would be good to compare energies as well from a 3.3
run and a 4.0 run.
In my last post, I stated: "when I define -DFLEXIBLE and make the tpr
using gromacs 3 and then run using gromacs 4, I do get gromacs 4 -
identical velocities.". However, thinking more on this has me believing
that this only indicates that the difference occurs at the mdrun stage.
I did a few more tests to see what the difference was when turning on or
off the unconstrained start option
in either gromacs 3 or gromacs 4. I see a very small diffence in
velocities compared to the MW difference between
gromacs 3 or gromacs 4 mdruns.
Below showing gromacs 4, using unconstrained_start=no vs.
unconstrained_start=yes has differences, but very minor.
diff gmx4.0.3_noUCS/feoff.gro gmx4.0.3/feoff.gro ...
< 51SOL OW 1151 6.816 4.071 1.558 0.0549 0.4777 -0.4315
< 51SOL HW1 1152 6.851 4.082 1.470 -0.1304 0.6992 -0.4795
< 51SOL HW2 1153 6.727 4.106 1.553 -0.2840 -0.3333 -0.3852
< 52SOL OW 1155 3.564 2.550 1.079 0.0951 0.0648 0.1748
< 52SOL HW1 1156 3.521 2.598 1.009 1.5023 0.4847 -0.4757
< 52SOL HW2 1157 3.551 2.604 1.157 0.0156 0.5140 -0.1424
---
51SOL OW 1151 6.816 4.071 1.558 0.0554 0.4775 -0.4315
51SOL HW1 1152 6.851 4.082 1.470 -0.1306 0.6992 -0.4793
51SOL HW2 1153 6.727 4.106 1.553 -0.2903 -0.3308 -0.3857
52SOL OW 1155 3.564 2.550 1.079 0.0951 0.0648 0.1747
52SOL HW1 1156 3.521 2.598 1.009 1.5034 0.4843 -0.4719
52SOL HW2 1157 3.551 2.604 1.157 0.0147 0.5151 -0.1438
...
with some, but not all MW showing small differences:
< 54SOL OW 1163 5.835 7.123 3.589 -0.0214 -0.1603 0.5173
< 54SOL HW1 1164 5.788 7.068 3.651 0.4698 -0.5196 0.5833
< 54SOL HW2 1165 5.841 7.208 3.632 -0.8849 -0.0052 0.3674
< 54SOL MW 1166 5.830 7.127 3.602 -0.0578 -0.2071 0.4856
---
54SOL OW 1163 5.835 7.123 3.589 -0.0214 -0.1604 0.5172
54SOL HW1 1164 5.788 7.068 3.651 0.4694 -0.5205 0.5835
54SOL HW2 1165 5.841 7.208 3.632 -0.8844 -0.0018 0.3686
54SOL MW 1166 5.829 7.127 3.602 -0.0578 -0.2071 0.4856
Below showing gromacs 3, using unconstrained_start=no vs.
unconstrained_start=yes has differences, but very minor.
diff gmx3.3.1_noUCS/feoff.gro gmx3.3.1/feoff.gro ...
< 51SOL OW 1151 6.816 4.071 1.558 0.0550 0.4778 -0.4315
< 51SOL HW1 1152 6.851 4.082 1.470 -0.1304 0.6992 -0.4795
< 51SOL HW2 1153 6.727 4.106 1.553 -0.2841 -0.3333 -0.3852
< 51SOL MW 1154 6.809 4.077 1.546 0.0254 -0.0273 0.0172
< 52SOL OW 1155 3.564 2.550 1.079 0.0951 0.0648 0.1749
< 52SOL HW1 1156 3.521 2.598 1.009 1.5023 0.4846 -0.4757
< 52SOL HW2 1157 3.551 2.604 1.157 0.0157 0.5140 -0.1425
< 52SOL MW 1158 3.557 2.563 1.080 -0.0486 0.0225 -0.0022
---
51SOL OW 1151 6.816 4.071 1.558 0.0554 0.4775 -0.4315
51SOL HW1 1152 6.851 4.082 1.470 -0.1306 0.6992 -0.4793
51SOL HW2 1153 6.727 4.106 1.553 -0.2903 -0.3308 -0.3857
51SOL MW 1154 6.809 4.077 1.546 0.0219 -0.0279 0.0237
52SOL OW 1155 3.564 2.550 1.079 0.0951 0.0648 0.1747
52SOL HW1 1156 3.521 2.598 1.009 1.5034 0.4843 -0.4719
52SOL HW2 1157 3.551 2.604 1.157 0.0147 0.5151 -0.1438
52SOL MW 1158 3.557 2.563 1.080 -0.0421 0.0144 0.0060
...
< 54SOL OW 1163 5.835 7.123 3.589 -0.0215 -0.1603 0.5173
< 54SOL HW1 1164 5.788 7.068 3.651 0.4698 -0.5196 0.5834
< 54SOL HW2 1165 5.841 7.208 3.632 -0.8849 -0.0051 0.3674
< 54SOL MW 1166 5.830 7.127 3.602 0.1837 -0.0398 0.1193
---
54SOL OW 1163 5.835 7.123 3.589 -0.0214 -0.1604 0.5172
54SOL HW1 1164 5.788 7.068 3.651 0.4694 -0.5205 0.5835
54SOL HW2 1165 5.841 7.208 3.632 -0.8844 -0.0018 0.3686
54SOL MW 1166 5.830 7.127 3.602 0.1879 -0.0399 0.1103
...
In the above cases, other specified .mdp options were:
integrator = md
comm_mode = linear
nstcomm = 1
comm_grps = System
nstlist = 5
ns_type = grid
pbc = xyz
coulombtype = PME
rcoulomb = 0.9
fourierspacing = 0.12
pme_order = 4
vdwtype = cut-off
rvdw_switch = 0
rvdw = 1.4
rlist = 0.9
constraints = none
energygrps = System
nsteps = 0
dt = 0.004
gen_vel = no
unconstrained-start = <yes or no depending on the test>
######
And note that if I set
constraints = all-bonds
constraint_algorithm= lincs
lincs-iter = 1
lincs-order = 6
then the difference between unconstrained-start=yes or =no was similarly
small.
Therefore it seems to me that something other than initial constraints
is leading to this difference.
Chris.
-- original message --
Thanks David, you are correct about initial step constraints. More
information inline below, but my question has been resolved.
>Chris Neale wrote:
>> Hello,
>>
>> Does anybody know if there is a reason why the .gro output velocities
>> would be different for tip4p MW in a zero-step mdrun between gromacs 3
>> and gromacs 4 (3.3.1 and 3.3.3 are the same, and are different from
>> 4.0.2 and 4.0.3, which are themselves the same).
>Is this with the same tpr?
This was utilizing a tpr generated by the same version of gromacs as was
used from the mdrun (3.3.1 grompp for 3.3.1 mdrun, etc.) although the
input was identical (with the exception of any undefined .mdp options
whose default values changes between v3 and v4, although I can't think
of one that affects only MW).
Based on your question, I did redo the 3.3.1 grompp to generate a single
.tpr that was used for both 3.3.1 mdrun and 4.0.3 mdrun. The results are
the same:
$ diff 4.gro 3.gro | head
1156c1156
< 51SOL MW 1154 6.809 4.077 1.546 0.0120 0.3740 -0.4572
---
> 51SOL MW 1154 6.809 4.077 1.546 0.0219 -0.0279 0.0237
1160c1160
< 52SOL MW 1158 3.557 2.563 1.080 0.2917 0.2007 0.0277
---
> 52SOL MW 1158 3.557 2.563 1.080 -0.0421 0.0144 0.0060
>
>It could have to do with initial step constraints.
I had considered that, but my .mdp uses unconstrained-start = yes.
Based on your suggestion I have attempted setting:
1. constraints = none (get the same difference)
2. also define = -DFLEXIBLE (get the same difference)
However, when I define -DFLEXIBLE and make the tpr using gromacs 3 and
then run using gromacs 4
I do get gromacs 4 - identical velocities.
$ diff 4from3.gro 4.gro | head
$ diff 4from3.gro 3.gro | head
1156c1156
< 51SOL MW 1154 6.809 4.077 1.546 0.0120 0.3740 -0.4572
---
> 51SOL MW 1154 6.809 4.077 1.546 0.0219 -0.0279 0.0237
1160c1160
< 52SOL MW 1158 3.557 2.563 1.080 0.2917 0.2007 0.0277
---
> 52SOL MW 1158 3.557 2.563 1.080 -0.0421 0.0144 0.0060
1164c1164
< 53SOL MW 1162 2.458 4.404 3.060 -0.0763 0.0165 -0.5712
So thank you, this explains it.
Chris.
>>
>> diff gmx4.0.3/feoff.gro gmx3.3.1/feoff.gro | head
>> 1156c1156
>> < 51SOL MW 1154 6.809 4.077 1.546 0.0120 0.3740 -0.4572
>> ---
>> > 51SOL MW 1154 6.809 4.077 1.546 0.0219 -0.0279 0.0237
>> 1160c1160
>> < 52SOL MW 1158 3.557 2.563 1.080 0.2917 0.2007 0.0277
>> ---
>> > 52SOL MW 1158 3.557 2.563 1.080 -0.0421 0.0144 0.0060
>> ...
>> (continues for all MW)
>>
>> This difference does not appear to be related to solvent optimization:
>>
>> $ diff gmx4.0.3/feoff.gro gmx4.0.3_GMX_NO_SOLV_OPT/feoff.gro
>>
>> I might just chalk this up to different mdp option defaults, but it
>> seems strange that there is a difference for all MW, but not for any
>> other atoms in the system.
>>
>> $ cat dpc50_md11.mdp
>> integrator = md
>> comm_mode = linear
>> nstcomm = 1
>> comm_grps = System
>> nstlist = 5
>> ns_type = grid
>> pbc = xyz
>> coulombtype = PME
>> rcoulomb = 0.9
>> fourierspacing = 0.12
>> pme_order = 4
>> vdwtype = cut-off
>> rvdw_switch = 0
>> rvdw = 1.4
>> rlist = 0.9
>> constraints = all-bonds
>> constraint_algorithm= lincs
>> lincs-iter = 1
>> lincs-order = 6
>> energygrps = System
>> nsteps = 0
>> dt = 0.004
>> gen_vel = no
>> unconstrained-start = yes
>>
>> Note that I explicitly included the tip4p.itp from gromacs 3.3.1 in
both
>> runs.
>>
>> Perhaps this is related to the gromacs 3 vsite problem?
>> http://www.gromacs.org/pipermail/gmx-users/2008-November/037659.html
>>
>> Thanks,
>> Chris.
_______________________________________________
gmx-users mailing list gmx-users@gromacs.org
http://www.gromacs.org/mailman/listinfo/gmx-users
Please search the archive at http://www.gromacs.org/search before posting!
Please don't post (un)subscribe requests to the list. Use the www
interface or send it to gmx-users-requ...@gromacs.org.
Can't post? Read http://www.gromacs.org/mailing_lists/users.php
--
David van der Spoel, Ph.D., Professor of Biology
Molec. Biophys. group, Dept. of Cell & Molec. Biol., Uppsala University.
Box 596, 75124 Uppsala, Sweden. Phone: +46184714205. Fax: +4618511755.
sp...@xray.bmc.uu.se sp...@gromacs.org http://folding.bmc.uu.se
_______________________________________________
gmx-users mailing list gmx-users@gromacs.org
http://www.gromacs.org/mailman/listinfo/gmx-users
Please search the archive at http://www.gromacs.org/search before posting!
Please don't post (un)subscribe requests to the list. Use the
www interface or send it to gmx-users-requ...@gromacs.org.
Can't post? Read http://www.gromacs.org/mailing_lists/users.php
