From owner-chemistry@ccl.net Tue Sep 21 08:46:01 2021 From: "Anson Thomas a_thomas(_)cy.iitr.ac.in" To: CCL Subject: CCL: Error installing QE-GPU Message-Id: <-54458-210921032533-22357-CdsJv/G+uc/5ozJkC0n6IQ+/-server.ccl.net> X-Original-From: Anson Thomas Content-Type: multipart/alternative; boundary="000000000000347e2c05cc7c4c56" Date: Tue, 21 Sep 2021 12:54:33 +0530 MIME-Version: 1.0 Sent to CCL by: Anson Thomas [a_thomas^cy.iitr.ac.in] --000000000000347e2c05cc7c4c56 Content-Type: text/plain; charset="UTF-8" Thank you Dr. Phil. I am using an nvfortran compiler downloaded from the Nvidia site ( nvidia_hpc_sdk ). As you suggested in point 1, I tried the modified the syntax it has configured successfully. However, on doing make all, or make pw, I now encounter the following: make[1]: Entering directory '/home/anson/qe/qe-6.8/install' *** Unable to download EIGENSOLVER_GPU. Test whether curl or wget is installed and working, *** if you have direct access to internet. If not, copy into archive/ the file *** located here https://github.com/NVIDIA/Eigensolver_gpu/archive/v0.3.1.tar.gz extlibs_makefile:81: recipe for target 'libcuda_eigensolver' failed make[1]: *** [libcuda_eigensolver] Error 1 make[1]: Leaving directory '/home/anson/qe/qe-6.8/install' Makefile:230: recipe for target 'libcuda' failed make: *** [libcuda] Error 2 wget and curl are not able to download from this github site, although it is working fine for others. I manually downloaded this gz file and copied it, but still I get the same error. Any suggestions? *Anson Thomas* M.Sc. Chemistry, IIT Roorkee Mob: *9910294743* a_thomas],[cy.iitr.ac.in https://www.linkedin.com/in/ansonthms/ On Tue, Sep 21, 2021 at 5:40 AM Phil Hasnip phil.hasnip-,-york.ac.uk < owner-chemistry],[ccl.net> wrote: > Dear Anson, > > There are several things you could check: > > 1. Are you using pgf90 or nvfortran, and which versions? The command-line > arguments have changed for more recent nvfortran compilers. Instead of > using > > -Mcuda= etc. > > the syntax is > > -cuda -gpu=,cuda > > 2. Are you compiling with MPI? Is the mpif90 (or mpifort) using the > correct Fortran compiler? You can use "mpif90 --show" to see which compiler > it's using. It should be pgf90 or nvfortran. > > 3. Have you tried to compile a simple test program using CUDA Fortran? > > Hope that helps, > > Phil Hasnip > > > ------------------------------------------------------------------------------------ > Dr Phil Hasnip Email: phil.hasnip~~york.ac.uk > EPSRC RSE Fellow Web: www-users.york.ac.uk/~pjh503 > Dept of Physics > University of York Tel: +44 (0)1904 322225 > York YO10 5DD > > > On Mon, 20 Sept 2021 at 21:56, Anson Thomas a_thomas:_:cy.iitr.ac.in < > owner-chemistry~~ccl.net> wrote: > >> >> Sent to CCL by: "Anson Thomas" [a_thomas%%cy.iitr.ac.in] >> I am trying to install the latest version of quantum espresso (6.8) with >> GPU- >> support on an Ubuntu 18.04.5 LTS (GNU/Linux 4.15.0-135-generic x86_64) >> >> System Configuration: >> Processor: Intel Xeon Gold 5120 CPU ],[ 2.20 GHz (2 Processor) >> RAM: 96 GB >> HDD: 6 TB >> Graphics Card: NVIDIA Quadro P5000 (16 GB) >> >> Following the steps given at https://gitlab.com/QEF/q-e-gpu/-/wikis/home, >> I >> installed all 3 required packages (CUDA Toolkit v8+, PGI Compilers >> v17.10+, >> OpenMP package v3+), and tried configuring using >> >> ./configure >> --with-cuda=""/opt/nvidia/hpc_sdk/Linux_x86_64/21.7/cuda/11.4/"" >> --with-cuda-runtime=11.4 --with-cuda-cc=6.1 --enable-openmp --with- >> scalapack=no >> >> However, every time I get an error saying that >> >> checking whether Fortran compiler accepts -Mcuda=cuda11.4... no >> configure: error: You do not have the cudafor module. Are you using a PGI >> compiler? >> >> What can be done, any suggestions?>> E-mail to subscribers: CHEMISTRY~~ccl.net or use:>> >> E-mail to administrators: CHEMISTRY-REQUEST~~ccl.net or use>> >> >> --000000000000347e2c05cc7c4c56 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable
make[1]: *** [libcuda_eigensolver] Error 1
make[1]: L= eaving directory '/home/anson/qe/qe-6.8/install'
Makefile= :230: recipe for target 'libcuda' failed
make: *** [libcu= da] Error 2


wget and curl are= not able to download from this github site, although it is working fine fo= r others.=C2=A0
I manually downloaded this gz file and copied it,= but still I get the same error.=C2=A0

Any suggest= ions?


Anson Thomas
M.Sc. Chemi= stry, IIT Roorkee
Mob:=C2=A09910294743

On Tue, Sep 21, 2021 at 5:40 AM Phil Hasnip phil.hasnip-,-york.ac.uk <owner-chemistry],[ccl.net> wrote:
Dear Anson,

There are several things you could check:

1. Are you using pgf= 90 or nvfortran, and which versions? The command-line arguments have change= d for more recent nvfortran compilers. Instead of using

-Mcuda=3D&l= t;cuda_version> etc.=C2=A0

the syntax=C2=A0is <= br>
-cuda=C2=A0-gpu=3D<compute_capability>,cuda<cuda_version>= ;

2. Are you compiling with MPI? Is the mpif90 (or mpifort) using th= e correct Fortran compiler? You can use "mpif90 --show" to see wh= ich compiler it's using. It should be pgf90 or nvfortran.
3. Have you tried to compile a simple test program using CUDA Fortran?
=
Hope that helps,

Phil Hasnip

---------------------------= ---------------------------------------------------------
Dr Phil Hasnip= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 Em= ail: phil.hasn= ip~~york.ac.uk
EPSRC RSE Fellow=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0= =C2=A0=C2=A0 Web:=C2=A0 www-users.york.ac.uk/~pjh503
Dept of Physics
Univ= ersity of York =C2=A0 =C2=A0 =C2=A0 =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 Tel: =C2= =A0=C2=A0=C2=A0 +44 (0)1904 322225
York YO10 5DD


On Mon, 20 Sept 2021 at 21:56, Anson Thomas a_thomas:_:cy.iitr.ac.in <owner-chemistry~~cc= l.net> wrote:

Sent to CCL by: "Anson=C2=A0 Thomas" [a_thomas%%cy.iitr.ac.in]
I am trying to install the latest version of quantum espresso (6.8) with GP= U-
support on an Ubuntu 18.04.5 LTS (GNU/Linux 4.15.0-135-generic x86_64)

System Configuration:
Processor: Intel Xeon Gold 5120 CPU ],[ 2.20 GHz (2 Processor)
RAM: 96 GB
HDD: 6 TB
Graphics Card: NVIDIA Quadro P5000 (16 GB)

Following the steps given at https://gitlab.com/QEF/q-e-g= pu/-/wikis/home, I
installed all 3 required packages (CUDA Toolkit v8+, PGI Compilers v17.10+,=
OpenMP package v3+), and tried configuring using

./configure --with-cuda=3D""/opt/nvidia/hpc_sdk/Linux_x86_64/21.7= /cuda/11.4/""
--with-cuda-runtime=3D11.4 --with-cuda-cc=3D6.1 --enable-openmp --with-
scalapack=3Dno

However, every time I get an error saying that

checking whether Fortran compiler accepts -Mcuda=3Dcuda11.4... no
configure: error: You do not have the cudafor module. Are you using a PGI <= br> compiler?

What can be done, any suggestions?



-=3D This is automatically added to each message by the mailing script =3D-=
E-mail to subscribers: CHEMISTRY~~ccl.net or use:
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message

E-mail to administrators: CHEMISTRY-REQUEST~~ccl.net or use
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/chemistry/sub_un= sub.shtml

Before posting, check wait time at: http://www.ccl.net

Job: http://www.ccl.net/jobs
Conferences: http://server.ccl.net/chemist= ry/announcements/conferences/

Search Messages: http://www.ccl.net/chemistry/sear= chccl/index.shtml
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/spammers.txt

RTFI: http://www.ccl.net/chemistry/aboutccl/ins= tructions/


--000000000000347e2c05cc7c4c56--