<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-7" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000066">
David van der Spoel wrote:
<blockquote cite="mid:49833E7F.9090600@xray.bmc.uu.se" type="cite">Dimitris
Dellis wrote:
<br>
<blockquote type="cite"><br>
Justin A. Lemkul wrote:
<br>
<blockquote type="cite"><br>
<br>
Dimitris Dellis wrote:
<br>
<blockquote type="cite">Hi.
<br>
I run the same (exactly) simulations with v3.3.3 and v4.0.3, on the
same 64bit Q6600/DDR2-1066 machine, gcc-4.3.2 ,fftw-3.2.
<br>
I found that the performance of 4.0.3 is roughly 30% lower than 3.3.3
(30% higher hours/ns), for few systems (512 molecules of 5-15 sites,
nstlist=10) I tried.
<br>
This happens with single precision serial and parallel, np=2,4
(openmpi 1.3) versions and only when electrostatics (PME) are present.
<br>
With Simple LJ potentials the performance is exactly the same.
<br>
Is there any speed comparison 3.3.3 vs 4.0.3 available ?
<br>
D.D.
<br>
<br>
</blockquote>
<br>
Can you show us your .mdp file? What did grompp report about the
relative PME load? These topics have been discussed a few times;
you'll find lots of pointers on optimizing performance in the list
archive.
<br>
</blockquote>
</blockquote>
<br>
try turning off optimize_fft
<br>
<br>
</blockquote>
Though better with optimize_fft = no,<br>
performce of 3.3.3 is still better.<br>
v4.0.3 1.389 hr/ns (from 1.665)<br>
<br>
<br>
<br>
<blockquote cite="mid:49833E7F.9090600@xray.bmc.uu.se" type="cite">
<blockquote type="cite">
<blockquote type="cite"><br>
-Justin
<br>
<br>
</blockquote>
Hi Justin,
<br>
These are from the small system, no I/O only 1k steps.
<br>
grompp.mdp
<br>
===========
<br>
integrator = md
<br>
dt = 0.0010
<br>
nsteps = 1000
<br>
nstxout = 0
<br>
nstvout = 0
<br>
nstlog = 1000
<br>
nstcomm = 10
<br>
nstenergy = 0
<br>
nstxtcout = 0
<br>
nstlist = 10
<br>
ns_type = grid
<br>
dispcorr = AllEnerPres
<br>
tcoupl = berendsen
<br>
tc-grps = System
<br>
ref_t = 293.15
<br>
gen_temp = 293.15
<br>
tau_t = 0.2
<br>
gen_vel = no
<br>
gen_seed = 123456
<br>
constraints = none
<br>
constraint_algorithm = shake
<br>
dispcorr = AllEnerPres
<br>
energygrps = System rlist = 1.6
<br>
vdw-type = Cut-off
<br>
rvdw = 1.6
<br>
coulombtype = PME
<br>
fourierspacing = 0.12
<br>
pme_order = 4
<br>
ewald_rtol = 1.0e-5
<br>
optimize_fft = yes
<br>
rcoulomb = 1.6
<br>
<br>
related 4.0.3 grompp output
<br>
Estimate for the relative computational load of the PME mesh part: 0.19
<br>
<br>
4.0.3 mdrun serial timings (near zero omitted)
<br>
<br>
Coul(T) + LJ 576.513824 31708.260 71.5
<br>
Outer nonbonded loop 8.489390 84.894 0.2
<br>
Calc Weights 6.006000 216.216 0.5
<br>
Spread Q Bspline 128.128000 256.256 0.6
<br>
Gather F Bspline 128.128000 1537.536 3.5
<br>
3D-FFT 1088.769682 8710.157 19.6
<br>
Solve PME 18.531513 1186.017 2.7
<br>
<br>
parallel 4.0.3 np=4
<br>
Average load imbalance: 5.2 %
<br>
Part of the total run time spent waiting due to load imbalance: 2.3 %
<br>
Performance: 96.086 7.380 14.414 1.665
<br>
<br>
3.3.3 mdrun serial timings
<br>
Coul(T) + LJ 576.529632 31709.129760 72.0
<br>
Outer nonbonded loop 8.487860 84.878600 0.2
<br>
Spread Q Bspline 128.128000 256.256000 0.6
<br>
Gather F Bspline 128.128000 1537.536000 3.5
<br>
3D-FFT 1088.769682 8710.157456 19.8
<br>
Solve PME 17.986469 1151.133984 2.6
<br>
<br>
parallel 3.3.3 np=4
<br>
Performance: 144.132 12.556 21.600 1.111
<br>
<br>
D.D.
<br>
<br>
<blockquote type="cite">
<blockquote type="cite"><br>
------------------------------------------------------------------------
<br>
<br>
_______________________________________________
<br>
gmx-users mailing list <a class="moz-txt-link-abbreviated" href="mailto:gmx-users@gromacs.org">gmx-users@gromacs.org</a>
<br>
<a class="moz-txt-link-freetext" href="http://www.gromacs.org/mailman/listinfo/gmx-users">http://www.gromacs.org/mailman/listinfo/gmx-users</a>
<br>
Please search the archive at <a class="moz-txt-link-freetext" href="http://www.gromacs.org/search">http://www.gromacs.org/search</a> before
posting!
<br>
Please don't post (un)subscribe requests to the list. Use the www
interface or send it to <a class="moz-txt-link-abbreviated" href="mailto:gmx-users-request@gromacs.org">gmx-users-request@gromacs.org</a>.
<br>
Can't post? Read <a class="moz-txt-link-freetext" href="http://www.gromacs.org/mailing_lists/users.php">http://www.gromacs.org/mailing_lists/users.php</a>
<br>
</blockquote>
<br>
</blockquote>
<br>
<br>
------------------------------------------------------------------------
<br>
<br>
_______________________________________________
<br>
gmx-users mailing list <a class="moz-txt-link-abbreviated" href="mailto:gmx-users@gromacs.org">gmx-users@gromacs.org</a>
<br>
<a class="moz-txt-link-freetext" href="http://www.gromacs.org/mailman/listinfo/gmx-users">http://www.gromacs.org/mailman/listinfo/gmx-users</a>
<br>
Please search the archive at <a class="moz-txt-link-freetext" href="http://www.gromacs.org/search">http://www.gromacs.org/search</a> before
posting!
<br>
Please don't post (un)subscribe requests to the list. Use the www
interface or send it to <a class="moz-txt-link-abbreviated" href="mailto:gmx-users-request@gromacs.org">gmx-users-request@gromacs.org</a>.
<br>
Can't post? Read <a class="moz-txt-link-freetext" href="http://www.gromacs.org/mailing_lists/users.php">http://www.gromacs.org/mailing_lists/users.php</a>
<br>
</blockquote>
<br>
<br>
</blockquote>
<br>
</body>
</html>