Re: [QE-users] Finite electric field polarisation

[Terry Frankcombe]

Hi Mostaffa

To close the loop here, I appreciate your response, though I am familiar with 
the Modern Theory and agree that the dipole from any particular cell is not 
useful in isolation.  However, the calculated dipole numbers should not jump 
from e.g. -10 to +5 in the last iteration as the SCF converges!

Nonetheless, it turns out the problem is something wrong with the 
centrally-provided build on the machine I'm using.  When I compile pw.x myself 
(ver. 6.7, but meh) on the same or another machine, the dipoles converge as I 
expect them to.

Thanks for your help.

Ciao
Terry



From: users  on behalf of Mostafa 
Youssef via users 
Sent: Tuesday, 19 January 2021 7:44 PM
To: users@lists.quantum-espresso.org 
Subject: Re: [QE-users] Finite electric field polarisation

Dear Terry,

The dipole results need post-processing. One needs to subtract n x polarization 
quantum to obtain a meaningful polarization.  n is an integer.

Consider a simple dielectric like MgO at zero field. pw.x may give you 
different dipoles in all directions, although physically in all directions the 
net dipoles are zero. If you post-process as I indicated above you get zero 
dipole as it should.

A good starter on this topic is this paper:
https://www.sciencedirect.com/science/article/abs/pii/S0022459612003234
A free arXiv version of it is also available.



Mostafa
AUC

___
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list users@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users

Re: [QE-users] percision of eigenvalues

[Stefano Baroni]

My question is: how far can you trust the overall accuracy of a numerical dft 
calculation to that extreme limit? SB

___
Stefano Baroni, Trieste -- http://stefano.baroni.me

> On 26 Jan 2021, at 18:24, Zahra Khatibi  wrote:
> 
> 
> Dear all, 
> 
> 
> I am currency using ncsf calculation to produce band dispersion. I need to 
> generate high precision eigenvalues, e.g. as small as 0.01 meV. I managed to 
> increase the number of decimal digits shown after the decimal point in the 
> standard output by modifying subroutine print_ks_energies.f90 in QE version 
> 6.5 (format 9030 from 8F9.4 to 8F9.5). 
> My question is, how far can I go for increasing the number of decimal digits 
> shown in the output, and what parameter dictates the precision of eigenvalues?
> 
> 
> Kind regards,
> --
> Z. Khatibi
> Postdoctoral fellow
> School of Physics
> Trinity College Dublin
> ___
> Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
> users mailing list users@lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/users
___
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list users@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users

Re: [QE-users] Problem when compiling qe-6.7

[Fabrizio Ferrari]

You can definitely use libxc, but with version 5.0.0 you need to pay
attention to the f03 interfaces, which have been removed and then restored
after v5.0.0. In the user guide in doc folder, in the section 'libxc
library' you will find everything you need to know, especially in the 'note
for libxc 5.0.0'.
Cheers,
Fabrizio

Il mar 26 gen 2021, 20:57 José Carlos Conesa Cegarra 
ha scritto:

> Dear Fabrizio,
>
> I am compiling with libxc 5.0.0, which I installed previously. But because
> the first attempt was not successful, I used the followig recipe, that I
> found sometime ago:
>
> - run configure without any libxc flag
> - go to the make.inc file and, by following the comments inside, add the
> needed flags to DFLAGS, IFLAGS and LD_LIBS.
>
> Do you think that it is better not to use libxc at all?
>
> José C. Conesa
> El 26/01/2021 a las 19:10, Fabrizio Ferrari escribió:
>
> Hello,
> are you compiling with libxc? Which version?
>
> Fabrizio
>
> On Tue, Jan 26, 2021 at 6:39 PM José Carlos Conesa Cegarra <
> jccon...@icp.csic.es> wrote:
>
>> Dear all,
>>
>> I tried compiling qe-6.7. The compilation fails with a number of
>> messages like these:
>>
>> funct.f90(40): error #7002: Error in opening the compiled module file.
>> Check INCLUDE paths.   [XC_F03_LIB_M]
>>USE xc_f03_lib_m
>> --^
>> funct.f90(424): error #6457: This derived type name has not been
>> declared.   [XC_F03_FUNC_T]
>>  TYPE(xc_f03_func_t) :: xc_func03
>> -^
>> funct.f90(776): error #6404: This name does not have a type, and must
>> have an explicit type.   [XC_FUNC03]
>>  CALL xc_f03_func_init( xc_func03, id_vec(ii), 1 )
>> ---^
>> funct.f90(926): error #6054: A CHARACTER data type is required in this
>> context.   [XC_F03_FUNCTIONAL_GET_NAME]
>> name = xc_f03_functional_get_name( i )
>> --^
>> funct.f90(962): error #6601: In a CASE statement, the case-value must be
>> a constant expression.   [XC_FAMILY_GGA]
>>CASE( XC_FAMILY_GGA, XC_FAMILY_HYB_GGA )
>> ^
>> funct.f90(962): error #6612: In a CASE statement, the case-value must be
>> of type INTEGER, CHARACTER, or LOGICAL.   [XC_FAMILY_GGA]
>>CASE( XC_FAMILY_GGA, XC_FAMILY_HYB_GGA )
>> ^
>> funct.f90(962): error #6611: The case-value must be of the same type as
>> the case-expr.   [XC_FAMILY_GGA]
>>CASE( XC_FAMILY_GGA, XC_FAMILY_HYB_GGA )
>> ^
>>
>> All of them relate to funct.f90. Maybe the key issue relates to error
>> #7002; but I could not find any module named xc_f03_lib_m.mod. Which can
>> be the reason?
>>
>> Regards,
>>
>> José C. Conesa
>>
>> ___
>> Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
>> users mailing list users@lists.quantum-espresso.org
>> https://lists.quantum-espresso.org/mailman/listinfo/users
>
>
> ___
> Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
> users mailing list 
> users@lists.quantum-espresso.orghttps://lists.quantum-espresso.org/mailman/listinfo/users
>
> --
> José C. Conesa
> Research Professor
> Instituto de Catálisis y Petroleoquímica, CSIC
> Marie Curie, 2; Campus de Cantoblanco
> 28028 Madrid (Spain)
> Phone +34 915854766
>
>
>
> 
>  Libre
> de virus. www.avast.com
> 
> <#m_-7070990978830589904_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
> ___
> Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
> users mailing list users@lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/users
___
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list users@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users

[QE-users] QE + Wannier90 + problem with spn file

[Arena Konta]

Dear QE Users

I have a problem with spn files. I prepared calculations with various version 
of QE (6.6, 6.5, 6.2.1) with p2wannier90 and wannier90 (version 3.1). 
Everything seems to be ok, expect, when I run postwannier90 app I got error:

* Energy bands in eV, coloured by SHC

   * Berry curvature-like term for spin Hall conductivity in Ang^2

 Reading overlaps from Wannier.mmn in get_AA_R   :  Created on 26Jan2021 at 17: 
3:27

 Reading spin matrices from Wannier.spn in get_SS_R : reated on 26Jan2021 at 
17: 31: 3
      34     1000
  0.8621686762
 Exiting...
 Error: Problem reading input file Wannier.spn


What can be wrong with this file? I did diffrent calculations using ELK and the 
same wannier90 library, and everything seems to be also OK.


-- 
with regards
 
Arena Konta
The Institute of Thermophysics in Novosibirsk Scientific Center

___
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list users@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users

Re: [QE-users] Problem when compiling qe-6.7

[José Carlos Conesa Cegarra]

Dear Fabrizio,

I am compiling with libxc 5.0.0, which I installed previously. But 
because the first attempt was not successful, I used the followig 
recipe, that I found sometime ago:


- run configure without any libxc flag
- go to the make.inc file and, by following the comments inside, add the 
needed flags to DFLAGS, IFLAGS and LD_LIBS.


Do you think that it is better not to use libxc at all?

José C. Conesa

El 26/01/2021 a las 19:10, Fabrizio Ferrari escribió:

Hello,
are you compiling with libxc? Which version?

Fabrizio

On Tue, Jan 26, 2021 at 6:39 PM José Carlos Conesa Cegarra 
mailto:jccon...@icp.csic.es>> wrote:


Dear all,

I tried compiling qe-6.7. The compilation fails with a number of
messages like these:

funct.f90(40): error #7002: Error in opening the compiled module
file.
Check INCLUDE paths.   [XC_F03_LIB_M]
   USE xc_f03_lib_m
--^
funct.f90(424): error #6457: This derived type name has not been
declared.   [XC_F03_FUNC_T]
 TYPE(xc_f03_func_t) :: xc_func03
-^
funct.f90(776): error #6404: This name does not have a type, and must
have an explicit type.   [XC_FUNC03]
 CALL xc_f03_func_init( xc_func03, id_vec(ii), 1 )
---^
funct.f90(926): error #6054: A CHARACTER data type is required in
this
context.   [XC_F03_FUNCTIONAL_GET_NAME]
    name = xc_f03_functional_get_name( i )
--^
funct.f90(962): error #6601: In a CASE statement, the case-value
must be
a constant expression.   [XC_FAMILY_GGA]
   CASE( XC_FAMILY_GGA, XC_FAMILY_HYB_GGA )
^
funct.f90(962): error #6612: In a CASE statement, the case-value
must be
of type INTEGER, CHARACTER, or LOGICAL.   [XC_FAMILY_GGA]
   CASE( XC_FAMILY_GGA, XC_FAMILY_HYB_GGA )
^
funct.f90(962): error #6611: The case-value must be of the same
type as
the case-expr.   [XC_FAMILY_GGA]
   CASE( XC_FAMILY_GGA, XC_FAMILY_HYB_GGA )
^

All of them relate to funct.f90. Maybe the key issue relates to error
#7002; but I could not find any module named xc_f03_lib_m.mod.
Which can
be the reason?

Regards,

José C. Conesa

___
Quantum ESPRESSO is supported by MaX (www.max-centre.eu
)
users mailing list users@lists.quantum-espresso.org

https://lists.quantum-espresso.org/mailman/listinfo/users



___
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list users@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users


--
José C. Conesa
Research Professor
Instituto de Catálisis y Petroleoquímica, CSIC
Marie Curie, 2; Campus de Cantoblanco
28028 Madrid (Spain)
Phone +34 915854766



--
El software de antivirus Avast ha analizado este correo electrónico en busca de 
virus.
https://www.avast.com/antivirus
___
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list users@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users

Re: [QE-users] Problem when compiling qe-6.7

[Fabrizio Ferrari]

Hello,
are you compiling with libxc? Which version?

Fabrizio

On Tue, Jan 26, 2021 at 6:39 PM José Carlos Conesa Cegarra <
jccon...@icp.csic.es> wrote:

> Dear all,
>
> I tried compiling qe-6.7. The compilation fails with a number of
> messages like these:
>
> funct.f90(40): error #7002: Error in opening the compiled module file.
> Check INCLUDE paths.   [XC_F03_LIB_M]
>USE xc_f03_lib_m
> --^
> funct.f90(424): error #6457: This derived type name has not been
> declared.   [XC_F03_FUNC_T]
>  TYPE(xc_f03_func_t) :: xc_func03
> -^
> funct.f90(776): error #6404: This name does not have a type, and must
> have an explicit type.   [XC_FUNC03]
>  CALL xc_f03_func_init( xc_func03, id_vec(ii), 1 )
> ---^
> funct.f90(926): error #6054: A CHARACTER data type is required in this
> context.   [XC_F03_FUNCTIONAL_GET_NAME]
> name = xc_f03_functional_get_name( i )
> --^
> funct.f90(962): error #6601: In a CASE statement, the case-value must be
> a constant expression.   [XC_FAMILY_GGA]
>CASE( XC_FAMILY_GGA, XC_FAMILY_HYB_GGA )
> ^
> funct.f90(962): error #6612: In a CASE statement, the case-value must be
> of type INTEGER, CHARACTER, or LOGICAL.   [XC_FAMILY_GGA]
>CASE( XC_FAMILY_GGA, XC_FAMILY_HYB_GGA )
> ^
> funct.f90(962): error #6611: The case-value must be of the same type as
> the case-expr.   [XC_FAMILY_GGA]
>CASE( XC_FAMILY_GGA, XC_FAMILY_HYB_GGA )
> ^
>
> All of them relate to funct.f90. Maybe the key issue relates to error
> #7002; but I could not find any module named xc_f03_lib_m.mod. Which can
> be the reason?
>
> Regards,
>
> José C. Conesa
>
> ___
> Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
> users mailing list users@lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/users
___
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list users@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users

[QE-users] Problem when compiling qe-6.7

[José Carlos Conesa Cegarra]

Dear all,

I tried compiling qe-6.7. The compilation fails with a number of 
messages like these:


funct.f90(40): error #7002: Error in opening the compiled module file.  
Check INCLUDE paths.   [XC_F03_LIB_M]

  USE xc_f03_lib_m
--^
funct.f90(424): error #6457: This derived type name has not been 
declared.   [XC_F03_FUNC_T]

    TYPE(xc_f03_func_t) :: xc_func03
-^
funct.f90(776): error #6404: This name does not have a type, and must 
have an explicit type.   [XC_FUNC03]

    CALL xc_f03_func_init( xc_func03, id_vec(ii), 1 )
---^
funct.f90(926): error #6054: A CHARACTER data type is required in this 
context.   [XC_F03_FUNCTIONAL_GET_NAME]

   name = xc_f03_functional_get_name( i )
--^
funct.f90(962): error #6601: In a CASE statement, the case-value must be 
a constant expression.   [XC_FAMILY_GGA]

  CASE( XC_FAMILY_GGA, XC_FAMILY_HYB_GGA )
^
funct.f90(962): error #6612: In a CASE statement, the case-value must be 
of type INTEGER, CHARACTER, or LOGICAL.   [XC_FAMILY_GGA]

  CASE( XC_FAMILY_GGA, XC_FAMILY_HYB_GGA )
^
funct.f90(962): error #6611: The case-value must be of the same type as 
the case-expr.   [XC_FAMILY_GGA]

  CASE( XC_FAMILY_GGA, XC_FAMILY_HYB_GGA )
^

All of them relate to funct.f90. Maybe the key issue relates to error 
#7002; but I could not find any module named xc_f03_lib_m.mod. Which can 
be the reason?


Regards,

José C. Conesa

___
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list users@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users

[QE-users] percision of eigenvalues

[Zahra Khatibi]

Dear all,


I am currency using ncsf calculation to produce band dispersion. I need to
generate high precision eigenvalues, e.g. as small as 0.01 meV. I managed
to increase the number of decimal digits shown after the decimal point in
the standard output by modifying subroutine *print_ks_energies.f90* in QE
version 6.5 (format 9030 from 8F9.4 to 8F9.5).
My question is, how far can I go for increasing the number of decimal
digits shown in the output, and what parameter dictates the precision of
eigenvalues?


Kind regards,
--
Z. Khatibi
Postdoctoral fellow
School of Physics
Trinity College Dublin
___
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list users@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users

Re: [QE-users] vc-relax with PBE0 not converging

[Bidault, Xavier]

Hello,

My system is an organic molecular crystal (atoms C, H, N, O). The cutoff is 
chosen so there is 2 meV/atom error for SCF energy. The k-point grid 3x2x3 
(off-centered) yields perfectly converged SCF energy. But to get faster with 
the vc-relax, I lowered it to 2x1x2 (off-centered), which introduces an error 
of only -0.1 meV/atom. I don't know if this is the reason why vc-relax density 
and pressure oscillate without converging.

I'm currently running v-relax with 3x2x3 grid... waiting for the result (10x 
slower)...

Thank you,
Xavier

From: users  on behalf of Paolo 
Giannozzi 
Sent: Monday, January 25, 2021 9:38 AM
To: Quantum ESPRESSO users Forum 
Subject: Re: [QE-users] vc-relax with PBE0 not converging

Maybe the same problem reported here?
   
https://gitlab.com/QEF/q-e/-/issues/204
Paolo

On Mon, Jan 25, 2021 at 4:19 PM Bidault, Xavier 
mailto:xavb...@uic.edu>> wrote:
Hello,

I have a molecular crystal (1 unit cell = 56 atoms). I use SG15 ONCV 1.2 
pseudopotential and Grimme D2 correction.
Using vc-relax with PBE, the structure converges, no problem.
But with PBE0, density and pressure oscillate without dampening (see figure 
below - Pressure remains negative, and not close to zero). Is there any 
particular strategy with PBE0?

  *   vc-relax with PBE and then starting from the result and use PBE0?
  *   relax with PBE0 and then vc-relax?
  *   any simulation parameter to change (ion dynamics and/or cell dynamics - 
currently BFGS)?
  *   any strategy with PBE0?

[cid:1773a31d564cb971f161]
Thank you,
Xavier

___
Quantum ESPRESSO is supported by MaX 
(www.max-centre.eu)
users mailing list 
users@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users


--
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 206, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222

___
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list users@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users

Re: [QE-users] Out of Memory on large available RAM for QE-GPU version

[Romero Molina, Sandra]

Dear Pietro,

Thank you very much for the fast answer, it was very useful.

Best regards,
Sandra

-Original Message-
From: users [mailto:users-boun...@lists.quantum-espresso.org] On Behalf Of 
Pietro Bonfa'
Sent: Tuesday, January 26, 2021 2:14 PM
To: users@lists.quantum-espresso.org
Subject: Re: [QE-users] Out of Memory on large available RAM for QE-GPU version

Dear Sandra,

unfortunately PW still doesn't predict the amount of GPU memory to be used 
during the simulation, but the estimate for RAM is also a good guess for the 
GPU memory.

The error message that you see is actually a failed allocation on the GPU side, 
not on the RAM.

Even if you had nvidia cards with 16 GB of memory, the prediction, that in your 
case reads

Estimated max dynamical RAM per process >  13.87 GB

is, as you can see, generally underestimated. I wouldn't be surprised of a 
15-20% inaccuracy.

Hope this helps,
best,
Pietro

--
Pietro Bonfà
Department of Mathematical, Physical and Computer Sciences University of Parma


On 1/26/21 12:58 PM, Romero Molina, Sandra wrote:
> Dear Community,
> 
> I have compiled Quantum ESPRESSO (Program PWSCF v.6.7MaX) for GPU 
> acceleration (hybrid MPI/OpenMP) with the next options:
> 
>    module load compiler/intel/2020.1
> 
>    module load hpc_sdk/20.9
> 
>    ./configure F90=pgf90 CC=pgcc MPIF90=mpif90 
> --with-cuda=yes --enable-cuda-env-check=no --with-cuda-runtime=11.0 
> --with-cuda-cc=70 --enable-openmp BLAS_LIBS='-lmkl_intel_lp64 
> -lmkl_intel_thread -lmkl_core'
> 
>    make -j8 pw
> 
> Apparently the compilation ends successfully and then, I execute the 
> program:
> 
>    module load compiler/intel/2020.1
> 
>    module load hpc_sdk/20.9
> 
>    export OMP_NUM_THREADS=1
> 
>    mpirun -n 2 /home/my_user/q-e-gpu-qe-gpu-6.7/bin/pw.x < 
> silverslab32.in > silver4.out
> 
> Then, the program starts and output:
> 
>      Parallel version (MPI & OpenMP), running on   8 processor cores
> 
>      Number of MPI processes: 2
> 
>      Threads/MPI process: 4
> 
>    ...
> 
>    GPU acceleration is ACTIVE
> 
>    ...
> 
>      Estimated max dynamical RAM per process >  13.87 GB
> 
>      Estimated total dynamical RAM >  27.75 GB
> 
> But after 2 minutes of execution the job ends with error:
> 
> 0: ALLOCATE: 4345479360 bytes requested; status = 2(out of memory)
> 
> 0: ALLOCATE: 4345482096 bytes requested; status = 2(out of memory)
> 
> --
> 
> Primary job  terminated normally, but 1 process returned
> 
> a non-zero exit code. Per user-direction, the job has been aborted.
> 
> --
> 
> --
> 
> mpirun detected that one or more processes exited with non-zero status, 
> thus causing
> 
> the job to be terminated. The first process to do so was:
> 
>    Process name: [[47946,1],1]
> 
>    Exit code:    127
> 
> --
> 
> This node has > 180GB of available RAM. With the top commands this is 
> the memory consume:
> 
>     PID USER  PR  NI    VIRT    RES    SHR S  %CPU %MEM TIME+ 
> COMMAND
> 
> 89681 my_user   20   0   30.1g   3.6g   2.1g R 100.0  1.9   1:39.45 pw.x
> 
> 89682 my_user   20   0   29.8g   3.2g   2.0g R 100.0  1.7   1:39.30 pw.x
> 
> When the RES memory arise the 4GB the processes stop and the error is 
> displayed
> 
> This are the characteristics of the node:
> 
> (base) [my_user@gpu001]$ numactl --hardware
> 
> available: 2 nodes (0-1)
> 
> node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 28 29 30 31 32 33 34 35 36 
> 37 38 39 40 41
> 
> node 0 size: 95313 MB
> 
> node 0 free: 41972 MB
> 
> node 1 cpus: 14 15 16 17 18 19 20 21 22 23 24 25 26 27 42 43 44 45 46 47 
> 48 49 50 51 52 53 54 55
> 
> node 1 size: 96746 MB
> 
> node 1 free: 70751 MB
> 
> node distances:
> 
> node   0   1
> 
>    0:  10  21
> 
>    1:  21  10
> 
> (base) [my_user@gpu001]$ free -lm
> 
>    total    used    free  shared  buff/cache   
> available
> 
> Mem: 192059    2561  112716 260   76781  
> 188505
> 
> Low: 192059   79342  112716
> 
> High: 0   0   0
> 
> Swap:      8191   0    8191
> 
> (base) [my_user@gpu001]$ ulimit -a
> 
> core file size     (blocks, -c) 0
> 
> data seg size      (kbytes, -d) unlimited
> 
> scheduling priority (-e) 0
> 
> file size   (blocks, -f) unlimited
> 
> pending signals (-i) 768049
> 
> max locked memory   (kbytes, -l) unlimited
> 
> max memory size (kbytes, -m) unlimited
> 
> open files  (-n) 10
> 
> 

Re: [QE-users] Out of Memory on large available RAM for QE-GPU version

[Pietro Bonfa']

Dear Sandra,

unfortunately PW still doesn't predict the amount of GPU memory to be 
used during the simulation, but the estimate for RAM is also a good 
guess for the GPU memory.


The error message that you see is actually a failed allocation on the 
GPU side, not on the RAM.


Even if you had nvidia cards with 16 GB of memory, the prediction, that 
in your case reads


Estimated max dynamical RAM per process >  13.87 GB

is, as you can see, generally underestimated. I wouldn't be surprised of 
a 15-20% inaccuracy.


Hope this helps,
best,
Pietro

--
Pietro Bonfà
Department of Mathematical, Physical and Computer Sciences
University of Parma


On 1/26/21 12:58 PM, Romero Molina, Sandra wrote:

Dear Community,

I have compiled Quantum ESPRESSO (Program PWSCF v.6.7MaX) for GPU 
acceleration (hybrid MPI/OpenMP) with the next options:


   module load compiler/intel/2020.1

   module load hpc_sdk/20.9

   ./configure F90=pgf90 CC=pgcc MPIF90=mpif90 
--with-cuda=yes --enable-cuda-env-check=no --with-cuda-runtime=11.0 
--with-cuda-cc=70 --enable-openmp BLAS_LIBS='-lmkl_intel_lp64 
-lmkl_intel_thread -lmkl_core'


   make -j8 pw

Apparently the compilation ends successfully and then, I execute the 
program:


   module load compiler/intel/2020.1

   module load hpc_sdk/20.9

   export OMP_NUM_THREADS=1

   mpirun -n 2 /home/my_user/q-e-gpu-qe-gpu-6.7/bin/pw.x < 
silverslab32.in > silver4.out


Then, the program starts and output:

     Parallel version (MPI & OpenMP), running on   8 processor cores

     Number of MPI processes: 2

     Threads/MPI process: 4

   ...

   GPU acceleration is ACTIVE

   ...

     Estimated max dynamical RAM per process >  13.87 GB

     Estimated total dynamical RAM >  27.75 GB

But after 2 minutes of execution the job ends with error:

0: ALLOCATE: 4345479360 bytes requested; status = 2(out of memory)

0: ALLOCATE: 4345482096 bytes requested; status = 2(out of memory)

--

Primary job  terminated normally, but 1 process returned

a non-zero exit code. Per user-direction, the job has been aborted.

--

--

mpirun detected that one or more processes exited with non-zero status, 
thus causing


the job to be terminated. The first process to do so was:

   Process name: [[47946,1],1]

   Exit code:    127

--

This node has > 180GB of available RAM. With the top commands this is 
the memory consume:


    PID USER  PR  NI    VIRT    RES    SHR S  %CPU %MEM TIME+ 
COMMAND


89681 my_user   20   0   30.1g   3.6g   2.1g R 100.0  1.9   1:39.45 pw.x

89682 my_user   20   0   29.8g   3.2g   2.0g R 100.0  1.7   1:39.30 pw.x

When the RES memory arise the 4GB the processes stop and the error is 
displayed


This are the characteristics of the node:

(base) [my_user@gpu001]$ numactl --hardware

available: 2 nodes (0-1)

node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 28 29 30 31 32 33 34 35 36 
37 38 39 40 41


node 0 size: 95313 MB

node 0 free: 41972 MB

node 1 cpus: 14 15 16 17 18 19 20 21 22 23 24 25 26 27 42 43 44 45 46 47 
48 49 50 51 52 53 54 55


node 1 size: 96746 MB

node 1 free: 70751 MB

node distances:

node   0   1

   0:  10  21

   1:  21  10

(base) [my_user@gpu001]$ free -lm

   total    used    free  shared  buff/cache   
available


Mem: 192059    2561  112716 260   76781  
188505


Low: 192059   79342  112716

High: 0   0   0

Swap:      8191   0    8191

(base) [my_user@gpu001]$ ulimit -a

core file size     (blocks, -c) 0

data seg size      (kbytes, -d) unlimited

scheduling priority (-e) 0

file size   (blocks, -f) unlimited

pending signals (-i) 768049

max locked memory   (kbytes, -l) unlimited

max memory size (kbytes, -m) unlimited

open files  (-n) 10

pipe size    (512 bytes, -p) 8

POSIX message queues (bytes, -q) 819200

real-time priority  (-r) 0

stack size  (kbytes, -s) unlimited

cpu time   (seconds, -t) unlimited

max user processes  (-u) 4096

virtual memory  (kbytes, -v) unlimited

file locks  (-x) unlimited

The version of MPI is: (Open MPI) 3.1.5

This node is a compute node in a cluster, but no matter if I submit the 
job with SLURM or run it directly on the node, the error is the same.


Note that I compile it on the login node and run it on this GPU node, 
the difference is that on the 

[QE-users] Out of Memory on large available RAM for QE-GPU version

[Romero Molina, Sandra]

Dear Community,

I have compiled Quantum ESPRESSO (Program PWSCF v.6.7MaX) for GPU acceleration 
(hybrid MPI/OpenMP) with the next options:

  module load compiler/intel/2020.1
  module load hpc_sdk/20.9

  ./configure F90=pgf90 CC=pgcc MPIF90=mpif90 --with-cuda=yes 
--enable-cuda-env-check=no --with-cuda-runtime=11.0 --with-cuda-cc=70 
--enable-openmp BLAS_LIBS='-lmkl_intel_lp64 -lmkl_intel_thread -lmkl_core'

  make -j8 pw

Apparently the compilation ends successfully and then, I execute the program:
  module load compiler/intel/2020.1
  module load hpc_sdk/20.9

  export OMP_NUM_THREADS=1
  mpirun -n 2 /home/my_user/q-e-gpu-qe-gpu-6.7/bin/pw.x < 
silverslab32.in > silver4.out


Then, the program starts and output:
Parallel version (MPI & OpenMP), running on   8 processor cores
Number of MPI processes: 2
Threads/MPI process: 4
  ...
  GPU acceleration is ACTIVE
  ...
Estimated max dynamical RAM per process >  13.87 GB
Estimated total dynamical RAM >  27.75 GB

But after 2 minutes of execution the job ends with error:

0: ALLOCATE: 4345479360 bytes requested; status = 2(out of memory)
0: ALLOCATE: 4345482096 bytes requested; status = 2(out of memory)
--
Primary job  terminated normally, but 1 process returned
a non-zero exit code. Per user-direction, the job has been aborted.
--
--
mpirun detected that one or more processes exited with non-zero status, thus 
causing
the job to be terminated. The first process to do so was:

  Process name: [[47946,1],1]
  Exit code:127
--

This node has > 180GB of available RAM. With the top commands this is the 
memory consume:

   PID USER  PR  NIVIRTRESSHR S  %CPU %MEM TIME+ COMMAND
89681 my_user   20   0   30.1g   3.6g   2.1g R 100.0  1.9   1:39.45 pw.x
89682 my_user   20   0   29.8g   3.2g   2.0g R 100.0  1.7   1:39.30 pw.x

When the RES memory arise the 4GB the processes stop and the error is displayed

This are the characteristics of the node:
(base) [my_user@gpu001]$ numactl --hardware
available: 2 nodes (0-1)
node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 28 29 30 31 32 33 34 35 36 37 38 
39 40 41
node 0 size: 95313 MB
node 0 free: 41972 MB
node 1 cpus: 14 15 16 17 18 19 20 21 22 23 24 25 26 27 42 43 44 45 46 47 48 49 
50 51 52 53 54 55
node 1 size: 96746 MB
node 1 free: 70751 MB
node distances:
node   0   1
  0:  10  21
  1:  21  10

(base) [my_user@gpu001]$ free -lm
  totalusedfree  shared  buff/cache   available
Mem: 1920592561  112716 260   76781  188505
Low: 192059   79342  112716
High: 0   0   0
Swap:  8191   08191

(base) [my_user@gpu001]$ ulimit -a
core file size (blocks, -c) 0
data seg size  (kbytes, -d) unlimited
scheduling priority (-e) 0
file size   (blocks, -f) unlimited
pending signals (-i) 768049
max locked memory   (kbytes, -l) unlimited
max memory size (kbytes, -m) unlimited
open files  (-n) 10
pipe size(512 bytes, -p) 8
POSIX message queues (bytes, -q) 819200
real-time priority  (-r) 0
stack size  (kbytes, -s) unlimited
cpu time   (seconds, -t) unlimited
max user processes  (-u) 4096
virtual memory  (kbytes, -v) unlimited
file locks  (-x) unlimited

The version of MPI is: (Open MPI) 3.1.5

This node is a compute node in a cluster, but no matter if I submit the job 
with SLURM or run it directly on the node, the error is the same.
Note that I compile it on the login node and run it on this GPU node, the 
difference is that on the login node it has no GPU connected.
I would really appreciate it if you could help me figure out what could be 
going on.
Thank you.


Ms.C. Sandra Romero Molina
Ph.D. student
Computational Biochemistry
T03 R01 D48
Faculty of Biology
University of Duisburg-Essen
Universitätsstr. 2, 45117 Essen
emails: sandra.romero-mol...@uni-due.de
Phone: +49 176 2341 8772
ORCID: https://orcid.org/-0002-4990-1649

___
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list users@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users

Re: [QE-users] dynamical matrix after restart

[Lorenzo Paulatto]

Please see
PHonon/examples/GRID_example
and
PHonon/examples/GRID_recover_example
regards
LP

--
Lorenzo Paulatto - Paris
On Jan 26 2021, at 7:48 am, Бунтов Евгений Александрович  
wrote:
>
> Dear all,
>
>
> I try to calculate Raman frequencies for a 64-atom slab with ph.x in QE 
> v.6.4.1.
> I had to interrupt the calculation at certain representation, then I 
> calculated the rest using start_irr/last_irr.
> At the end the fildyn file is empty, so I can’t use dynmat.x.
> Can I create a dynamical matrix from the existing xml files in phsave 
> directory without performing all the calculations once again?
>
> Regards,
> Evgeny
> ___
> Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
> users mailing list users@lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/users

___
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list users@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users