[QE-users] Fwd: [Pw_forum] how to collect dvscf files?

Lorenzo Paulatto paulatz at gmail.com
Wed Mar 21 17:33:32 CET 2018


On 21/03/18 16:15, Hari Paudyal wrote:
> Thank you for your response. The way you described (alat byte level) may 
> be appropriate for the experts, but for me, I am sorry, it is not 
> familiar. I tried your second suggestion instead to store the dvscf in 
> formatted file, but I meshed up. Could you please give me a direction 
> which file I need to change and how, to make dvscf formatted? It would 
> be great for the learner like me.

Hello,
I hardly have the time to do this. But to be honest doing it directly on 
the files is much easier than it look, except for the bad spelling in my 
message (I blame autocorrect).

Let's say you have 6 modes, and you run them in groups of 2.

Then in each directory you will have a dvscf file, you can check their 
sizes with "ls -l"

In directory 1:
-rw-r--r-- 1 xx yy   442368 Jan 19 14:22 silicon.dvscf1

directory 2:
-rw-r--r-- 1 xx yy   884736 Jan 19 14:22 silicon.dvscf1

directory 3:
-rw-r--r-- 1 xx yy  1327104 Jan 19 14:22 silicon.dvscf1

Which means that each mode takes 1327104/6 = 221184 bytes

The size of each mode is 1327104/6 = 221184
The first file has 2 modes
The second file has 2 modes filled with zeros, then two actual modes
The third file has 4 modes filled with zeros, then two actual modes

You can use dd to recompose the file

dd bs=221184 count=2 if=FILE1 of=TOTAL
dd bs=221184 count=2 skip=2 seek=2 if=FILE2 of=TOTAL
dd bs=221184 count=2 skip=4 seek=4 if=FILE3 of=TOTAL

Where bs is the size of the mode in bytes (the size of the block that dd 
will copy) and "count" is the number of blocks to copy.

I.e. copy 2 blocks of size 221184 from FILE1 to file TOTAL

Then copy 2 blocks of size 221184 from FILE2 to file TOTAL, bust skip 
the first 2 blocks of both FILE2 and TOTAL

...


hth


> 
> Thank you for the time and consideration.
> 
> Hari Paudyal
> SUNY Binghamton, USA
> 
> On Mon, Feb 26, 2018 at 4:37 AM, Lorenzo Paulatto <paulatz at gmail.com 
> <mailto:paulatz at gmail.com>> wrote:
> 
>     Dear Hari,
>     There is no automatic way to do it, it quite simple in practice: you
>     can do it alat byte level, or from fortran. If your final file size
>     is N bytes and you have M modes you take the first N/M bytes from
>     the node that computed the first mode, and so on. The final file
>     size is the size of the file on the node that computed the last mode.
> 
>     That said, it is a bit risky, and jt would be better to store the
>     dvscf in formatted file. However, I cannot spend my limited time on
>     this issue for the moment.
> 
>     HTH
> 
>     -- 
>     Lorenzo Paulatto
>     Written on a virtual keyboard with real fingers
> 
>     On Feb 25, 2018 23:26, "Hari Paudyal" <hpaudya1 at binghamton.edu
>     <mailto:hpaudya1 at binghamton.edu>> wrote:
> 
>         Dear experts,
> 
>         Does anybody have an idea how to collect the dvscf files after
>         splitting the phonon jobs in different machines for different
>         representations/q-points using QE? The dynmat and patterns files
>         can be collected as described in the examples, but not the dvscf
>         one.I need a single dvscf file for one q-point so that it can be
>         used further in the EPW calculations. Your suggestions in this
>         regard will be highly appreciated.
> 
>         Thank You,
>         Hari Paudyal
>         SUNY Binghamton, USA.
> 
>         _______________________________________________
>         Pw_forum mailing list
>         Pw_forum at pwscf.org <mailto:Pw_forum at pwscf.org>
>         http://pwscf.org/mailman/listinfo/pw_forum
>         <http://pwscf.org/mailman/listinfo/pw_forum>
> 
> 
> 
>     _______________________________________________
>     Pw_forum mailing list
>     Pw_forum at pwscf.org <mailto:Pw_forum at pwscf.org>
>     http://pwscf.org/mailman/listinfo/pw_forum
>     <http://pwscf.org/mailman/listinfo/pw_forum>
> 
> 
> 
> 
> 
> _______________________________________________
> users mailing list
> users at lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/users
> 

-- 
Lorenzo Paulatto - Paris


More information about the users mailing list