Nicola et al.,<br><br>I am attempting to setup such a test under controlled conditions.<br><br>It would be helpful to clarify a point, because we want to avoid comparing apples with oranges.<br><br>There are two ways to solve for the KS ground state.
<br>1. Direct minimization of KS functional<br>2. Self-consistently solving the KS eqn.<br><br>The conjugate-gradient (cg) in PWSCF and CP are not doing different things same thing. In PWSCF we are doing #2, while in CP we are doing #1.
<br><br>What would best for comparison purposes then? Have each do 10 iterations with CG using random wavefunctions as a starting guess. I would think that even with those settings that PWSCF still has the advantage due to have startingpot = atomic.
<br><br>Hopefully, I am not saying jibberish. I having not thought how the PW codes work in a while so I am rusty.<br><br><br><div><span class="gmail_quote">On 2/15/07, <b class="gmail_sendername">Nicola Marzari</b> <<a href="mailto:marzari@mit.edu">
marzari@mit.edu</a>> wrote:</span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><br><br>Paolo Umari has also implemented a conjugate gradients minimization
<br>in CP.<br><br>so, for a CP calculation, in which the only thing you need is the<br>electronic ground state of a given structure, this would be the best<br>strategy - our few tests were that it was 3-4 times more effective than
<br>damped dynamics. If you want to relax a structure, you should do, after<br>this first electronic calculation, a combined damped dynamics on<br>electrons and ions.<br><br>One thing we haven't tested is the relative performance of PWSCF and
<br>CP - my guess is that the algorithms of PWSCF are much faster, although<br>the balance might change as the system becomes larger and larger.<br>In principle, the CP algorithm is variational, so should always<br>converge, but again this is rarely a problem.
<br><br>Bottom line - advantages for PWSCF is that it should be faster for most<br>systems of interest. CP uses much less memory, and should be very<br>efficient in calculating H|psi>, so the balance when going towards very
<br>large scale or very parallel calculations would be extremely interesting<br>to explore.<br><br>Any comparative test, under controlled conditions, would be very useful.<br><br> nicola<br>
<br><br>Paolo Giannozzi wrote:<br>><br>> On Feb 14, 2007, at 18:48 , Nichols A. Romero wrote:<br>><br>>> If one were to use CP across such a large number of processors,<br>>> what algorithm is your best bet for minimizing your electronic
d.o.f.,<br>>> sd or cg<br>><br>> steepest descent is very inefficient and should be used only<br>> for the initial steps of a simulation or in desperate cases. The<br>> "standard" way of minimizing electronic degrees of freedom
<br>> in CP is damped dynamics<br>><br>> Paolo<br>> ---<br>> Paolo Giannozzi, Democritos and University of Udine, Italy<br>><br>><br>> _______________________________________________<br>> Pw_forum mailing list
<br>> <a href="mailto:Pw_forum@pwscf.org">Pw_forum@pwscf.org</a><br>> <a href="http://www.democritos.it/mailman/listinfo/pw_forum">http://www.democritos.it/mailman/listinfo/pw_forum</a><br><br><br>--<br>---------------------------------------------------------------------
<br>Prof Nicola Marzari Department of Materials Science and Engineering<br>13-5066 MIT 77 Massachusetts Avenue Cambridge MA 02139-4307 USA<br>tel 617.4522758 fax 2586534 <a href="mailto:marzari@mit.edu">marzari@mit.edu
</a> <a href="http://quasiamore.mit.edu">http://quasiamore.mit.edu</a><br>_______________________________________________<br>Pw_forum mailing list<br><a href="mailto:Pw_forum@pwscf.org">Pw_forum@pwscf.org</a><br><a href="http://www.democritos.it/mailman/listinfo/pw_forum">
http://www.democritos.it/mailman/listinfo/pw_forum</a><br></blockquote></div><br><br clear="all"><br>-- <br>Nichols A. Romero, Ph.D.<br>1613 Denise Dr. Apt. D<br>Forest Hill, MD 21050<br>443-567-8328 (C)<br>410-306-0709 (O)
<br>