<div dir="ltr">Dear Prof. Hutter,<div><br></div><div>Thank you very much for your reply and instruction.</div><div><br></div><div>In the second step, do you mean increasing the grids might spend more time on the FFT calculation?</div><div><br></div><div>Regards,</div><div><br></div><div>Huan</div><div>.</div></div><div class="gmail_extra"><br><div class="gmail_quote">On Tue, Mar 29, 2016 at 11:33 AM, <span dir="ltr"><<a href="mailto:hut...@chem.uzh.ch" target="_blank">hut...@chem.uzh.ch</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Hi<br>
<br>
the three inputs "CUTOFF", "NGRIDS", and "REL_CUTOFF" are important<br>
for accuracy and efficiency. The problem is that both, "accuracy and<br>
efficiency" depend on the combination of chosen values.<br>
<br>
In an optimal setting you can proceed the following way:<br>
<br>
1) choose a high enough value for REL_CUTOFF<br>
this is system and basis set independent. It determines the<br>
minimal cutoff used to bring a Gaussian function on the real space grid.<br>
High values mean high accuracy, low values faster execution.<br>
<br>
2) choose your cutoff and number of grids. This will determine the real space grids<br>
for your calulcation, e.g. cutoff=400 ngrids=4<br>
grid1 = 400, grid2 = 200, grid3 =100, grid4 = 50 (not actual numbers)<br>
The cutoff should be high enough to take on the Gaussian with the largest exponent,<br>
i.e. exponent*REL_CUTOFF <= cutoff (where exponent is for a product Gaussian)<br>
If you choose more grids the mapping (Gaussian->Grid) is better but you have more<br>
overhead from more grids (mostly FFTs).<br>
<br>
regards<br>
<br>
Juerg Hutter<br>
<br>
--------------------------------------------------------------<br>
Juerg Hutter Phone : ++41 44 635 4491<br>
Institut für Chemie C FAX : ++41 44 635 6838<br>
Universität Zürich E-mail: <a href="mailto:hut...@chem.uzh.ch">hut...@chem.uzh.ch</a><br>
Winterthurerstrasse 190<br>
CH-8057 Zürich, Switzerland<br>
---------------------------------------------------------------<br>
<br>
-----<a href="mailto:cp...@googlegroups.com">cp...@googlegroups.com</a> wrote: -----To: cp2k <<a href="mailto:cp...@googlegroups.com">cp...@googlegroups.com</a>><br>
From: <a href="mailto:huan...@mail.huji.ac.il">huan...@mail.huji.ac.il</a><br>
Sent by: <a href="mailto:cp...@googlegroups.com">cp...@googlegroups.com</a><br>
Date: 03/27/2016 10:16PM<br>
Subject: [CP2K:7621] How to choose the NGRIDS in &MGRID<br>
<span class=""><br>
Dear CP2K developers and users, <br>
</span>I found a usage hint in the BASIS_MOLOPT file."NGRIDS 5" (section &MGRID) will deal more efficiently (2x speedup) with the diffuse nature of the basis.Whereas in the CP2K manual, the default value of the NGRIDS is 4.<br>
My question is that is there any rule of thumb for choosing the NGRIDS?For example, lager basis set needs lager value of NGRIDS? Or how many Gaussian functions mapped on each grid would be efficient and accurate for the simulations?<br>
I noticed that the CP2K website gave an example on setting the CUTOFF and REL_CUTOFF, <a href="https://www.cp2k.org/howto:converging_cutoffwhich" rel="noreferrer" target="_blank">https://www.cp2k.org/howto:converging_cutoffwhich</a> might related to the NGRIDS.<br>
<span class="">However, the descriptions made me confused. It might lead to another question.<br>
In the introduction, it said:If CUTOFF is too low, then all grids will be coarse and the calculation may become inaccurate.<br>
</span>While at the end of "Converging CUTOFF" section, it goes,The reader may also notice that as CUTOFF increases, the number of Gaussians being assigned to the finest grids decreases. <br>
<br>
It looks like the "NG on grid 1" is the finest grid, since as the description of the above sentence, the number of Gaussians decreased significantly on grid 1.However, this would be contrary to the introduction. Because the introduction indicates that the lower CUTOFF, the coarser the grid. As a result, the higher CUTOFF should lead to finer grid. But the example showed that increasing the CUTOFF pushed the Gaussian to the coarse grid.<br>
<div class="HOEnZb"><div class="h5">I appreciate it if anyone could correct me if I am wrong.<br>
Regards,Huan.<br>
<br>
<br>
<br>
<br>
</div></div><span class="HOEnZb"><font color="#888888">--<br>
<br>
You received this message because you are subscribed to the Google Groups "cp2k" group.<br>
<br>
To unsubscribe from this group and stop receiving emails from it, send an email to <a href="mailto:cp2k%2Bun...@googlegroups.com">cp2k+uns...@googlegroups.com</a>.<br>
<br>
To post to this group, send email to <a href="mailto:cp...@googlegroups.com">cp...@googlegroups.com</a>.<br>
<br>
Visit this group at <a href="https://groups.google.com/group/cp2k" rel="noreferrer" target="_blank">https://groups.google.com/group/cp2k</a>.<br>
<br>
For more options, visit <a href="https://groups.google.com/d/optout" rel="noreferrer" target="_blank">https://groups.google.com/d/optout</a>.<br>
<br>
<br>
--<br>
You received this message because you are subscribed to a topic in the Google Groups "cp2k" group.<br>
To unsubscribe from this topic, visit <a href="https://groups.google.com/d/topic/cp2k/UvhmfO_LZxY/unsubscribe" rel="noreferrer" target="_blank">https://groups.google.com/d/topic/cp2k/UvhmfO_LZxY/unsubscribe</a>.<br>
To unsubscribe from this group and all its topics, send an email to <a href="mailto:cp2k%2Bun...@googlegroups.com">cp2k+uns...@googlegroups.com</a>.<br>
To post to this group, send email to <a href="mailto:cp...@googlegroups.com">cp...@googlegroups.com</a>.<br>
Visit this group at <a href="https://groups.google.com/group/cp2k" rel="noreferrer" target="_blank">https://groups.google.com/group/cp2k</a>.<br>
For more options, visit <a href="https://groups.google.com/d/optout" rel="noreferrer" target="_blank">https://groups.google.com/d/optout</a>.<br>
</font></span></blockquote></div><br></div>