Re: [SIESTA-L] Parallel siesta crashing during SCF

Hi, Without more information we can't help, 1) compiler and library versions 2) your arch.make 3) your input files Den tir. 17. nov. 2020 kl. 22.01 skrev D. Bennett : > Dear siesta users, > > I'm running some calculations using a parallel version of siesta-psml, > and I'm finding that some of

[SIESTA-L] Parallel siesta crashing during SCF

Dear siesta users, I'm running some calculations using a parallel version of siesta-psml, and I'm finding that some of them are crashing during the SCF loop. Things seem to go fine, and then I get a message like --- Primary job terminated n

Re: [SIESTA-L] parallel version issue of 4-1.-b3, m_orderbands.o: undefined reference to `indxg2p_'

-- Forwarded message - From: Alberto Garcia Date: Sun, Sep 23, 2018 at 4:03 PM Subject: Re: [SIESTA-L] parallel version issue of 4-1.-b3, m_orderbands.o: undefined reference to `indxg2p_' To: siesta-l not useful, I am looking for clear answer Hi, You seem to be missin

Re: [SIESTA-L] parallel version issue of 4-1.-b3, m_orderbands.o: undefined reference to `indxg2p_'

Hi, You seem to be missing the blacs and pblas libraries. These are sometimes included in scalapack, but not on all systems, apparently. Since you seem to be working on an IBM machine, you might get some inspiration from this snippet of an arch.make file that I used in the past. Details might b

[SIESTA-L] parallel version issue of 4-1.-b3, m_orderbands.o: undefined reference to `indxg2p_'

Hi Issue is : siesta 4.1-b3 MPI" Installation. I found the following issue and quite unsuccessful . I have following module's installed in our HPC. ps-01:siesta-4.1-b3] module load lapack/3.4.2 ps-01:siesta-4.1-b3] module load openmpi/1.8.8 ps-01:siesta-4.1-b3] module load numa/2.0.11 ps-01:sie

Re: [SIESTA-L] parallel version

Hai, Are you trying to install in Ubuntu? I was also struggling with compilation of parallel version, If you want, I could share the arch file which worked for me. Thanks Virus-free

[SIESTA-L] parallel version

Dear all users: I just start to use SIESTA. But I found that I installed a serial version. I want to compile a parallel version. But I can't get success only with: MPI_INTERFACE=libmpi_f90.a MPI_INCLUDE=/usr/local/include DEFS_MPI=-DMPI (see the manual) Please tell me how to compile it. My i

Re: [SIESTA-L] Parallel SIESTA: segmentation error depending on the number of atoms and cores.

Dear Peter, Thank you VERY-VERY MUCH!!! This is indeed some kind of magic command! Unlimited stack (ulimit -s unlimited) allowed me to run the siesta smoothly on 8 cores and with any number of atoms in the system as it should be. There is no seg-faults anymore up to siesta “end of run”. Fantast

Re: [SIESTA-L] Parallel SIESTA: segmentation error depending on the number of atoms and cores.

Dear Dima, let me comment on this familiar sucking behaviour. Ifort is fast but it is known for crashing with larger sizes at some innocent places such as intrinsic reshape, transpose, or even copy commands... Using unlimited stack helps in some cases (ulimit -s unlimited in bash) You could try

Re: [SIESTA-L] Parallel SIESTA: segmentation error depending on the number of atoms and cores.

Sadly we often see these kind of problems with Intel MKL and Intel MPI, and sometimes also with MKL and OpenMPI. We have had numerous tries on trying to narrow down these seg-faults, but since they are often non-reproducible it is very hard (if somebody wishes to help us, please let us know!). I w

[SIESTA-L] Parallel SIESTA: segmentation error depending on the number of atoms and cores.

Dear Siesta users and Developers For a long time I try to overcome the following problem: Increasing number of atoms in the system leads to the need to increase the number of cores used for the calculation. In general, this sounds logical, but there is one flaw. I have 8 cores and 32GB of memory

Re: [SIESTA-L] parallel installation of netcdf4

Dear siesta users, Does anybody know how to solve this error when compiling Denchar? "make: *** No rule to make target `Denchar'. Stop." Best.

Re: [SIESTA-L] parallel installation of netcdf4

TBtrans does not require a parallel netcdf version either. And if you use a parallel netcdf, you should expect little to no difference, so don't bother. 2017-09-28 6:22 GMT+02:00 Ankita Joshi : > Dear Nick, > > Thanks for your suggestion and I will do the same. As far as parallel > version of n

Re: [SIESTA-L] parallel installation of netcdf4

Dear Nick,Thanks for your suggestion and I will do the same. As far as parallel version of netcdf4 is concerned, I needed this for running tbtrans in parallel. Could you please that if there be huge difference (in terms of computational time) in running calculations using serial or parallel vers

Re: [SIESTA-L] parallel installation of netcdf4

Please note that there is no need to install the parallel version of netcdf4. It isn't required, in fact it will probably not make a huge difference. Secondly, you should use the script shipped with siesta-4.1-b3, not the one you linked. It contains the latest updates. So my suggestion, stick wit

Re: [SIESTA-L] parallel installation of netcdf4

I am sure you know that the *install_netcdf4.bash *scripts required zlib-1.2.11.tar.gz, hdf5-1.8.18.tar.bz2, netcdf-c-4.4.1.1.tar.gz and netcdf-fortran-4.4.4.tar.gz files to be downloaded in a folder. Copy *install_netcdf4.bash *to this folder. Run "*bash* *install_netcdf4.bash*" and add all the pa

[SIESTA-L] parallel installation of netcdf4

Dear Nick,I am facing difficulty in installing the parallel version of netcdf4. From the link below, I have installed the serial version of netcdf4.https://www.mail-archive.com/siesta-l@uam.es/msg09275.htmlCould you please tell, how to install the parallel version by modifying the script given

Re: [SIESTA-L] Parallel version of TBTrans with Netcdf4 support

Please update to 4.1-b3. Once you have downloaded the siesta sources (tbtrans are included with siesta) you may find a file called install_netcdf4.bash in the Docs directory. This file will enable you to install all required libraries to enable netcdf4 support for siesta AND tbtrans. Please read

[SIESTA-L] Parallel version of TBTrans with Netcdf4 support

Dear Siesta users,I have serial version of TBtrans 4.1 beta 2 with NetCDF4 support installed in my system. But I am unable to install its parallel version. Could anyone explain to me in detail about parallel installation of TBtrans with NetCDF4 support? Thanks RegardsAnkita> > > > > >

Re: [SIESTA-L] Parallel method by siesta

Hi Yangchuan, 2017-03-03 0:25 GMT+01:00 Yangchuan Li : > Hi Nick, > > Thank you so much for your help. > Following by meshgrid InitMesh: MESH = 216 x 216 x80 = 3732480 > If the ProcessorY is 8 and I have 64 cores. As I understand from the > manual, the mesh will divide into YZ direct

Re: [SIESTA-L] Parallel method by siesta

Hi Nick, Thank you so much for your help. Following by meshgrid InitMesh: MESH = 216 x 216 x80 = 3732480 If the ProcessorY is 8 and I have 64 cores. As I understand from the manual, the mesh will divide into YZ direction. means the total workload is divide into section of 216x1x1 with

Re: [SIESTA-L] Parallel method by siesta

Those output numbers are not intended for the users. They are mainly used by developers iff bugs occur. So an end user can easily neglect the output you refer to. However, the line: InitMesh: MESH = 216 x 216 x80 = 3732480 tells you the real space grid used internally (1st, 2nd, 3rd la

[SIESTA-L] Parallel method by siesta

Dear Siesta users, I'm trying to understanding how siesta running in parallel. I don't understand the meaning of the grid distribution part of the output file. For example like the following New grid distribution: 1 1 1: 1081: 271:8 2 1: 108

[SIESTA-L] parallel siesta 4.1 b1 with intel mkl libraries

Dear Siesta users I tried to compile Siesta-4.1 b-1 with intel mkl libraries but I am getting following errors: m_zminim.o: In function `__m_zminim_MOD_calc_densmat': /home/rajan.singh/siesta-4.1-b1/Src/m_zminim.F90:2566: undefined reference to `pzgemm_' /home/rajan.singh/siesta-4.1-b1/Src/m_zmin

[SIESTA-L] parallel Tbtrans stop

Dear all I would really appreciate if someone kindly answer to my question.my calculations with transiesta was successfully but tbtrans warned me : Number of Nodes greater than number of k-points!!Some nodes will be inactive  in k-point parallelization!! I increased k-points in x,y and z dire

Re: [SIESTA-L] Parallel compile and run

Good news! I finally built the parallel version on Rocks and I can see multiple processes on the hosts. I will prepare a document containing the documents about that later and will post it in a new post since this topic is very lengthy... Thanks to all who helped me. Regards, Mahmood

Re: [SIESTA-L] Parallel compile and run

Hi Mahmood, I know next to nothing about Rocks, so I'll leave that to more knowledgeable people, On 07/08/2016 12:44 PM, Mahmood Naderan wrote: ... ​As I borrowed From Robert (RCP). I have to change those line, but I am not sure what to do. Here is my suggestion LAPACK_LIBS=liblapack.a

Re: [SIESTA-L] Parallel compile and run

m_shabani1...@yahoo.com On Jul 8, 2016 11:09 PM, "Arthur France-Lanord" wrote: > Hi Mahmood, > > Another possibility, if you are managing that rocks cluster, would be to > propagate packages installed on the head node to the slave nodes. Basically > you’d have to edit a new xml config file, and b

Re: [SIESTA-L] Parallel compile and run

Hi Mahmood, Another possibility, if you are managing that rocks cluster, would be to propagate packages installed on the head node to the slave nodes. Basically you’d have to edit a new xml config file, and build a new rocks distro. See eg. http://www.rocksclusters.org/roll-documentation/bas

Re: [SIESTA-L] Parallel compile and run

Hi again, There are some problems with the linking the libraries for parallel execution. The system is based on ROCKS and if an MPI jobs is going to be run on the compute nodes, all libraries should be available on the nodes. Anyway... As another try, I am rebuilding siesta parallel by downloading

Re: [SIESTA-L] Parallel siesta using openmpi 1.10.3

Hi Nick, Thanks for the feedback. Responses below: > On 06 Jul 2016, at 15:30, Nick Papior wrote: > > Hi Arthur, > > 2016-07-06 15:13 GMT+02:00 Arthur France-Lanord >: > Dear all, > > I’ve recently compiled a parallel version of siesta 4.0b-485 using openmp

Re: [SIESTA-L] Parallel siesta using openmpi 1.10.3

Hi Arthur, 2016-07-06 15:13 GMT+02:00 Arthur France-Lanord : > Dear all, > > I’ve recently compiled a parallel version of siesta 4.0b-485 using openmpi > 1.10.3, and I’m seeing horrible parallel performances. I was previously > building siesta with openmpi 1.6.x (on another box though) and clear

[SIESTA-L] Parallel siesta using openmpi 1.10.3

Dear all, I’ve recently compiled a parallel version of siesta 4.0b-485 using openmpi 1.10.3, and I’m seeing horrible parallel performances. I was previously building siesta with openmpi 1.6.x (on another box though) and clearly didn’t get the same behaviour. For instance, when running Tests/s

Re: [SIESTA-L] Parallel compile and run

hi, just a simple thought... do you have the same name for siesta serial & parallel excutables ? try to name them siesta-s & siesta-p, so for mpirun use siesta-p s.auluck > Good news! > Finally I built parallel siesta > > root@cluster:ObjPar# ./siesta > ./siesta: error while loading shared

Re: [SIESTA-L] Parallel compile and run

As you can see it is saying PARALLEL version, but I see "Running in serial mode with MPI". What does that mean? Ok, no problem. It means you just typed "siesta", so openMPI handles it as a serial code. More info: root@cluster:ObjPar# ldd siesta        linux-vdso.so.1 => (0x

Re: [SIESTA-L] Parallel compile and run

Good news! Finally I built parallel siesta root@cluster:ObjPar# ./siesta ./siesta: error while loading shared libraries: libscalapack.so.2: cannot open shared object file: No such file or directory root@cluster:ObjPar# export LD_LIBRARY_PATH=/usr/lib64/openmpi/lib root@cluster:ObjPar# ./siesta

Re: [SIESTA-L] Parallel compile and run

Of course I have those libraries because I did compile them by myself (Slackware Linux distro). Probably you don't need them because the package in your distro is organized differently. Namely, -lmpiblacs alone might already work. Anyway, check the contents of blacs-openmpi-devel.x86

Re: [SIESTA-L] Parallel compile and run

We are progressing step by step! The next error is about -lblacsCinit library which is mentioned in BLACS_LIBS. I have installed blacs related packages as below root@cluster:ObjPar# find /usr/ -name *blacs* /usr/lib64/openmpi/lib/libmpiblacs.so /usr/lib64/openmpi/lib/libmpiblacs.so.2 /usr/lib64/op

Re: [SIESTA-L] Parallel compile and run

Seems your linker does not search /usr/lib64/openmpi/lib for libraries. So, yes, try -L/usr/lib64/openmpi/lib in the LDFLAGS R. On 07/05/2016 12:31 PM, Mahmood Naderan wrote: I searched a lot about the parallel compilation of siesta and saw some question on the list. However, they don't r

Re: [SIESTA-L] Parallel compile and run

I searched a lot about the parallel compilation of siesta and saw some question on the list. However, they don't report any success! Regarding the sckalapack, I may get related errors later! but let me explain step by step. I installed netcdf-devel.x86_64 (4.1.2) from the yum repository. Files act

Re: [SIESTA-L] Parallel compile and run

You must have a version of the netcdf library newer than mine. Those do not include the fortran interface anymore. Now, the latter is a separate package, but can't recall its name. It should be more or less obvious, search the web. Also, netcdf is optional for Siesta. You might comment it altoget

Re: [SIESTA-L] Parallel compile and run

What I did in the previous posts was 1) Running ../Src/obj_setup.sh 2) Copy arch.make from serial siesta and modified FC MPI_INTERFACE=libmpi_f90.a MPI_INCLUDE=. The reason I didn't use your arch.make was that I saw some differences between yours and my serial siesta. It seems that some

Re: [SIESTA-L] Parallel compile and run

Ok, weird. If I do ldd on mine it also shows these two components ... libopen-rte.so.0 => /usr/lib64/libopen-rte.so.0 (0x2b247d9cb000) libopen-pal.so.0 => /usr/lib64/libopen-pal.so.0 (0x2b247dc18000) ... that are basic MPI stuff and do not appear in yours.

Re: [SIESTA-L] Parallel compile and run

OK. I understand that. But still I have problem. Please see below. 1) mpif90 exists as below # which mpif90 /opt/openmpi/bin/mpif90 2) I modifed FC variable in arch.make from f95 to mpif90 and built siesta. Fortunately, ldd command shows the binary depends on some MPI libraries root@cluster:

Re: [SIESTA-L] Parallel compile and run

Hi, On 07/03/2016 03:40 PM, Mahmood Naderan wrote: ... mahmood@cluster:~$ mpirun -np 2 /export/apps/chemistry/siesta-3.2/ObjPar/siesta < file.fdf Siesta Version:                                       siesta-3.2 Architecture : x86_64-unknown-linux-gnu--un

Re: [SIESTA-L] Parallel compile and run

Hello again, I tried to build the parallel version. Though I had some progress, but still have problem. I created another Obj folder called ObjPar/ modified arch.make to include the MPI libraries. MPI_INTERFACE=libmpi_f90.a MPI_INCLUDE=. MPI_INCLUDE2=/usr/include/openmpi-x86_64 ​Then I ran make a

Re: [SIESTA-L] Parallel compile and run

Mahmood, There is only a ONE right manual: the corresponding one to the version you are trying to compile/use. Take a look in the DOCS folder inside the SIESTA distro: there is a SIESTA.TEX file. Compiling in using LATEX you will get THE RIGHT ONE. Also, in the SIESTA distro, there is a folder S

Re: [SIESTA-L] Parallel compile and run

>There’s an appendix (16) on parallel siesta in the documentation (page 131 in doc v.492), which explains >many things including compilation, and additional libraries required. The users gave me siesta-3.2-manual which is dated as March-24-2013 and has 124 pages! Please let me know which manual i

Re: [SIESTA-L] Parallel compile and run

There’s an appendix (16) on parallel siesta in the documentation (page 131 in doc v.492), which explains many things including compilation, and additional libraries required. Arthur > On 24 Jun 2016, at 15:04, Mahmood Naderan wrote: > > OK. It seems that it uses the serial version only.. >

Re: [SIESTA-L] Parallel compile and run

OK. It seems that it uses the serial version only.. mahmood@cluster:siesta-test$ mpirun -np 4 siesta < file.fdf Siesta Version: SIESTA_VERSION Architecture : SIESTA_ARCH Compiler flags: FFLAGS SERIAL version * Running in serial mode >> Start of run: 24-JUN-2016 17:31:30 Siesta Version: SIEST

Re: [SIESTA-L] Parallel compile and run

If your code was compiled for parallel execution, it should clearly say so in the head of the outpur file, just before the welcome banner. Regards, Roberto On 06/23/2016 08:04 PM, Mahmood Naderan wrote: The serial mode works for that fdf. Regards, Mahmood

Re: [SIESTA-L] Parallel compile and run

The serial mode works for that fdf. Regards, Mahmood

Re: [SIESTA-L] Parallel compile and run

Hi, The problem is not MPI but your data file. So, try a data file you know it works, even if in serial mode. You could grab one from the Tests folder, or, better still, run the the tests, it is a good policy. Cheers, Roberto On 06/23/2016 05:46 PM, Mahmood Naderan wrote: Hi, As a cluster ad

[SIESTA-L] Parallel compile and run

Hi, As a cluster administrator, users say that siesta is capable to run in parallel via MPI. I am not an expert with Siesta, so I saw the command as mpirun -np 8 siesta < file.fdf However, I get reinit: --- reinit: System Nam

Re: [SIESTA-L] Parallel SIESTA

Hi Nick ! Thanks. I can use both the serial and parallel version in the super computer. After this post I have figured out that using 4 nodes and ParallelOverK makes the process faster than the serial version. Using more than 4 nodes again makes it slower. I am still confused on how to optimize

Re: [SIESTA-L] Parallel SIESTA

We need much more information. For all we know, you could be using the serial version. Please elaborate on all you have done. -- Kind regards Nick Papior On 20 Oct 2015 23:48, "Anup Pandey" wrote: > Hello everyone, > > I have a simple and tricky question about running the SIESTA in parallel. >

[SIESTA-L] Parallel SIESTA

Hello everyone, I have a simple and tricky question about running the SIESTA in parallel. I am running my 512-atom system using PARALLEL siesta in 24 nodes in a supercomputer. What surprises me is that the program is running well but seems to be running slower than the SERIAL version. I have alre

Re: [SIESTA-L] Parallel Siesta

Building, Room 327 Atlanta, GA 30332-0700 Office: (404) 385-0045 From: siesta-l-requ...@uam.es on behalf of Nick Papior Sent: Friday, August 21, 2015 3:36 PM To: siesta-l@uam.es Subject: Re: [SIESTA-L] Parallel Siesta Show all the steps you perform. And please

Re: [SIESTA-L] Parallel Siesta

m:* siesta-l-requ...@uam.es on behalf of > Nick Papior > *Sent:* Friday, August 21, 2015 12:33 PM > *To:* siesta-l@uam.es > *Subject:* Re: [SIESTA-L] Parallel Siesta > > Yes, every available version from the website supports parallel > compilation. > > To successfully compil

Re: [SIESTA-L] Parallel Siesta

(404) 385-0045 From: siesta-l-requ...@uam.es on behalf of Nick Papior Sent: Friday, August 21, 2015 12:33 PM To: siesta-l@uam.es Subject: Re: [SIESTA-L] Parallel Siesta Yes, every available version from the website supports parallel compilation. To successfully compile with MPI you nee

Re: [SIESTA-L] Parallel Siesta

Yes, every available version from the website supports parallel compilation. To successfully compile with MPI you need to do this to your arch.make file: FPPFLAGS += -DMPI MPI_INTERFACE=libmpi_f90.a As always, remember to start all over to ensure everything is from scratch. 2015-08-21 16:08 GMT+

[SIESTA-L] Parallel Siesta

Dear Sir, I am trying to install siesta-3.2 on our local cluster, and when I put -DMPI in compilation, I am getting error /usr/local/build/Cherry/siesta-3.2-pl-5_orig_MoreDecimals/Src/sys.F(81): error #7002: Error in opening the compiled module file. Check INCLUDE paths. [MPI_SIESTA]

[SIESTA-L] parallel siesta as an spawned process

Dear Siesta Developers, I am developing this mpi genetic algorithm app to search for the optimal configuration of clusters of atoms. So far, serial siesta (evoked by "call system") has been used quite successfully as an external app to evaluate the cluster energy. I am now testing parallel siesta

Re: [SIESTA-L] parallel scaling

Look at this link (there is also a reference to a paper): http://departments.icmab.es/leem/siesta/siestimator/siestimator.php?Na=1&Nopa=4&NMD=1000&NSCFpMD=10&supercomputer=Hermit 2014-12-15 13:29 GMT+01:00 Herbert Fruchtl : > > Folks, > > Has anybody done a citable study on parallel scaling o

[SIESTA-L] parallel scaling

Folks, Has anybody done a citable study on parallel scaling of SIESTA on a large number of processors? Thanks in advance, Herbert -- Herbert Fruchtl Senior Scientific Computing Officer School of Chemistry, School of Mathematics and Statistics University of St Andrews -- The University of St

[SIESTA-L] parallel spin calculation

Dear siesta users, I'm trying to do the spin calculations for the geometry relaxation of my system within an efficient way by Siesta code. The number of orbitals are very much more than the k grid in my system. Which option Should I use for parallel spin calculation? I have used " SpinPolarized

Re : [SIESTA-L] parallel installation of siesta-ldau-308

Dear siesta users , I installed siesta-ldau-308 in serial version successfully after the editing in f2kcli.F90 file .I change #ifndef GFORTRAN to #ifndef __GFORTRAN__ to remove the denchar error while installation. Its installed successfully. But when I try to install in parrallel version to rem

[SIESTA-L] ??????[SIESTA-L] parallel version of siesta-3.2.

With these information, we can see nothing and I think nobody can answer your question. You should post the complete output file. -- -- ??: "somayeh fotoohi";; : 2014??5??10??(??) 8:18 ??: "siesta-l"; :

[SIESTA-L] parallel version of siesta-3.2.

Dear siesta users, I have installed parallel version of siesta-3.2. I run the program in parallel by submitting my job using .pbs script file, The calculation started normally, but it stopped suddengly as follows: Siesta Version:siesta-3.2 Architecture

[SIESTA-L] parallel version of siesta-3.2

Dear siesta users, I have installed parallel version of siesta-3.2.  I run the program in parallel by submitting my job using .pbs script file, The calculation started normally, but it stopped suddengly as follows:   Siesta Version:    siesta-3.2 Architecture 

[SIESTA-L] parallel and anti parallel magnetism orientation

hi dear users I want calculate TMR because it should be defined parallel and anti parallel configuration.every one can help me how dose it possible define these configurations? thanks alot.

Re : [SIESTA-L] parallel version

Dear siesta user I Install siesta 3.1 in parallel without any error . when I run the programme with one node it run successfully but when I increase the number of nodes the programmes terminated in the middle as shown below .I attached my arch .make file and h20.sh.e35125 file which showing the er

[SIESTA-L] parallel siesta Install and help

Hi all, I noticed the following thing: When I compile the siesta the FFLAGS setting like this: FFLAGS= -O2 -I$(MKLROOT)/include/intel64/lp64/ -I$(MKLROOT)/include Then, when I run the siesta after the compiling process finished, I see the following information on the stdout: ---

Re: [SIESTA-L] parallel compilation of siesta-3.

Dear siesta users I try to install siesta in parallel but I am getting the error which I can't able to analyse I edit the arch.make file gave path to libraries but when I try to make it I get the following error plz help me out Compilation architecture to be used: x86_64-unknown-linux-gnu-- unkn

Re: [SIESTA-L] Parallel compilation error

I'll try to compile siesta-3.1. But notthing new. The same error. ifort -c -g -O2 -I../objs/finclude -DFC_HAVE_FLUSH -DFC_HAVE_ABORT -DFC_ABORT_ARG -DFC_EOR_LF -DRESTRICTED_ASSOCIATED_BUG /home/users/cvashnini/src/siesta-3.1/Src/FoX/utils/fox_m_utils_uri.F90 /home/users/cvashnini/src/siesta-3.1/Sr

Re: [SIESTA-L] Parallel compilation error

I need to do transiesta calculations. But now I find that in siesta-3.1 also included transiesta calculations. Sorry, I have not followed by the updates,I used the version which had been downloaded a lot of time ago. Thanks, I'll try to compile siesta-3.1. On 4 October 2011 16:11, Yun-Peng Wang

Re: [SIESTA-L] Parallel compilation error

it is clear from the errors, .F90 file containing fox_m_fsys_string module should be compiled before fox_m_utils_uri.F90. That is why I say a bad makefile. On 10/04/2011 08:12 PM, Nick Papior Andersen wrote: Sorry not libraries, object files, it is clearly complaining about objects which com

Re: [SIESTA-L] Parallel compilation error

Sorry not libraries, object files, it is clearly complaining about objects which comes from another compiler, or in this case module files >This module file was not generated by any release of this compiler. Kind regards Nick 2011/10/4 Yun-Peng Wang > I think the errors happened before the lin

Re: [SIESTA-L] Parallel compilation error

I think the errors happened before the link stage, so I do not think older libraries is critical here. Yun-Peng On 10/04/2011 08:07 PM, Nick Papior Andersen wrote: It seems as though you are compiling a with a gcc and trying to compile together with older compiled libraries from intel. Try mak

Re: [SIESTA-L] Parallel compilation error

It seems as though you are compiling a with a gcc and trying to compile together with older compiled libraries from intel. Try make clean then make... 2011/10/4 Dmitry G. Kvashnin > Dear siesta users! > > > I'll try to compile siesta-trunk-364 on Moscow state university cluster: > MPI: openmpi-1

Re: [SIESTA-L] Parallel compilation error

it seems there is some error in makefile. Why not use formal siesta releases, such as siesta-3.1? best, Yun-Peng On 10/04/2011 08:02 PM, Dmitry G. Kvashnin wrote: Dear siesta users! I'll try to compile siesta-trunk-364 on Moscow state university cluster: MPI: openmpi-1.5 COMPILER: intel-12.0

[SIESTA-L] Parallel compilation error

Dear siesta users! I'll try to compile siesta-trunk-364 on Moscow state university cluster: MPI: openmpi-1.5 COMPILER: intel-12.0 Configure process was well. But when I want to make, I got the following error: Compilation architecture to be used: x86_64-unknown-linux-gnu--Gfortran If this is no

[SIESTA-L] Parallel compilation error

Dear siesta users. I'll try to compile siesta-trunk-364 on Moscow state university cluster: MPI: openmpi-1.5 COMPILER: intel-12.0 Configure process was well. But when I want to make, I got the following error: Compilation architecture to be used: x86_64-unknown-linux-gnu--Gfortran If this is no

[SIESTA-L] Parallel Installation

Dear Siesta Users, As a new user of this code, I need help installing it in parallel. I have Mpich2, blacs and scalapack installed but I am getting this error when i type make. Please help me diagnose and solve the problem /home/cecil/applications/siesta-3.0-rc2/Src/cdiag.F:438: undefined referen

Re: [SIESTA-L] parallel compilation

On Fri, 2011-01-14 at 01:45 +0100, enrique martinez wrote: > Hi there, > > When compiling in parallel with the option --enable-mpi in the > configure, is it mandatory to add the blacs and scalapack libraries??? Yes, it is. > The problem is that I cannot find the libraries in the cluster I am > u

Re: [SIESTA-L] parallel compilation

Hi there, When compiling in parallel with the option --enable-mpi in the configure, is it mandatory to add the blacs and scalapack libraries??? I am getting the following messages when doing the configure: -- configure: Don't know how to compile with BLACS - edit arch.make by hand checking if

Re: [SIESTA-L] parallel compilation

Thanks a lot Jan, That removed the problem with mpi but now I am getting the following errors: cdiag.o: In function `cdiag_': cdiag.F:(.text+0x392): undefined reference to `descinit_' cdiag.F:(.text+0x956): undefined reference to `blacs_gridinfo_' cdiag.F:(.text+0x973): undefined reference to `in

Re: [SIESTA-L] parallel compilation

Hi, I'm quite new to siesta too, but when I compiled it in parallel version, I used the --enable-mpi option with configure. It than ran some tests, if all libraries and compilers were working and stated errors or things I had to add manually in the arch.make-file. Dont know, if this helps.

[SIESTA-L] parallel compilation

Hi there, I am trying to compile version siesta-3.0-rc2 in parallel unsuccessfully. In the Obj/ folder I am doing ../Src/obj_setup.sh and then I am running ../Src/configure FC=mpif90 FCFLAGS=-O2 FPPFLAGS=-DMPI. In the arch.make I get the MPI_INTERFACE is blank. If I do make as it is I get the foll

Re: [SIESTA-L] Parallel Version of Siesta is only running in serial mode

I forgot to say this errors only occurs if i run this job on more than 1 processor. > Finally I get managed to compile the sourcecode but if i try to run a > test, for instance 32_h20 (in the Test folder) with mpiexec or mpirun i > get after a while following more or less cryptic error message and

Re: [SIESTA-L] Parallel Version of Siesta is only running in serial mode

Finally I get managed to compile the sourcecode but if i try to run a test, for instance 32_h20 (in the Test folder) with mpiexec or mpirun i get after a while following more or less cryptic error message and a subsequent killing of the siesta job: siesta: malloc.c:3096: sYSMALLOc: Assertion `(old

Re: [SIESTA-L] Parallel Version of Siesta is only running in serial mode

I tried this but I get with the default values follwoing errormessage gfortran: libmpi_f90.a: No such file or directory BUt this is interesting because other reasons , I this case i do not use gfortran as FC! I use mpif90! I found a libmpi_f90.a in /Obj/MPI and /Src which one i should use? Since

Re: [SIESTA-L] Parallel Version of Siesta is only running in serial mode

Dear Michael, >From my experience one shouldn't change the standard MPI definitions at all meaning it should be MPI_INTERFACE=libmpi_f90.a MPI_INCLUDE=./ I think there are some special things for siesta in them. I checked your arch.make and you had always MPI_INCLUDE defined to something, I have

Re: [SIESTA-L] Parallel Version of Siesta is only running in serial mode

Hello Everyone, I tried your suggestion ( i use instead of your suggestion this line "MPI_LIBS= -L/usr/local/openmpi -lmpi -lmpi_f90" and i add MPI_LIBS to the "LIBS" in addition i tried to link all libraries in usr/local/openmpi to this variable) Now I get a big bundle of new error messages durin

Re: [SIESTA-L] Parallel Version of Siesta is only running in serial mode

Michael, Try specifying the mpi libraries explicitly by adding the following line MPI_LIBS= -L/usr/local/lib -lfmpich -lmpich -lmpif90 changing the /usr/local/lib for whatever is appropriate, of course. Marcos On Tue, Dec 7, 2010 at 10:27 PM, wrote: > Dear Siesta-Users > Now I found out what

Re: [SIESTA-L] Parallel Version of Siesta is only running in serial mode

Dear Siesta-Users Now I found out whats going wrong. I have simply forgot to clean the directory, i.e. using serial libs leads to serial behavior. But now a new problem appears. If I try to make siesta using mpif90 or gfortran ( i tried both) and using scalapack, blacs, blas, lapack which was insta

Re: [SIESTA-L] Parallel Version of Siesta is only running in serial mode

Michael, Was everything compiled from scratch? OpenMPI, lapack, scalapack, blas, blacs? I would, as Tran suggested, use mpif90 instead of gfortran as FC, and also change the lines with libraries: SCALAPACK_LIBS=/usr/lib/libscalapack-openmpi.a to SCALAPACK_LIBS=-L/usr/lib -lscalapack-openmpi a

Re: [SIESTA-L] Parallel Version of Siesta is only running in serial mode

Hi Michel. If u are using the hlrn Cluster in Hannover-Berlin, i have a make file which Compiler the paralel Version. Cheers On Monday, December 6, 2010, wrote: > Thank I immediately tried you suggestion but this seems not to be the > reason for my problem > > Sincerelly > Michael > > >> I t

Re: [SIESTA-L] Parallel Version of Siesta is only running in serial mode

Thank I immediately tried you suggestion but this seems not to be the reason for my problem Sincerelly Michael > I think that in your arch.make the line "FC=gfortran" should be replaced > by > "FC=mpif90". > > On Mon, Dec 6, 2010 at 7:45 PM, wrote: > >> Ok sorry I forgot to say that i use gfort

Re: [SIESTA-L] Parallel Version of Siesta is only running in serial mode

I think that in your arch.make the line "FC=gfortran" should be replaced by "FC=mpif90". On Mon, Dec 6, 2010 at 7:45 PM, wrote: > Ok sorry I forgot to say that i use gfortran, openmpi and the following > arch.make file. > > Sincerelly > Michael Schuch > > > > This is definitely not due to pseudo

  1   2   >