Install request: Quantum Espresso 6.1

[heatherkellyucl]

IN:02403508

We have 5.2 and the newest is 6.1.

It has been updated so make all creates all the optional external packages as well.

http://www.quantum-espresso.org/wp-content/uploads/Doc/user_guide/node10.html

[heatherkellyucl]

Started checking how much of our current buildscript needs changing.

[heatherkellyucl]

Made a preliminary build script, currently trying build on Legion to see how many workarounds are still necessary.

[heatherkellyucl]

mpif90 -static-intel  -o pw.x \
   pwscf.o  libpw.a ../../Modules/libqemod.a ../../FFTXlib/libqefft.a ../../LAXlib/libqela.a /dev/shm/qe-6.1//clib/clib.a /dev/shm/qe-6.1//iotk/src/libiotk.a   -lfftw3xf_intel -lmkl_intel_lp64  -lmkl_sequential -lmkl_core
ld: cannot find -lfftw3xf_intel

Our install of the Intel 2017 compiler doesn't appear to have any libfftw in /shared/ucl/apps/intel/2017.Update1/mkl/lib/intel64 (Intel 2015 does).

[heatherkellyucl]

All the interfaces need building:

cd /shared/ucl/apps/intel/2017.Update1/mkl/interfaces/fftw3xf
make libintel64 compiler=gnu INSTALL_DIR=/shared/ucl/apps/intel/2017.Update1/mkl/lib/intel64

blas95 and lapack95 need ifort (so Intel compiler module loaded). fftw2x_cdft, fftw3x_cdft, mklmpi need Intel MPI loaded.

[heatherkellyucl]

Put message in Intel compiler buildscript to remind the next person to run the interface buildscript after installing.

[heatherkellyucl]

This is what needs doing, with the relevant Intel compiler module and Intel MPI loaded:

cd $MKLROOT/interfaces
for dir in */; do
    cd $dir
    make libintel64 compiler=intel INSTALL_DIR=$MKLROOT/lib/intel64
    cd ..
done

Made intel-compilers-interfaces_install and added reminder to intel-compilers-2017-update1_install.

[heatherkellyucl]

Built interfaces on:

• [x] Legion
• [x] Grace
• [x] Thomas

[heatherkellyucl]

Test build of QE finished on Legion. It doesn't have quite the same executables for this version as 5.2.0 has - I don't know if that is expected or some parts didn't build automatically.

5.2.0 version:

average.x      ev.x                           lambda.x              projwfc.x          turbo_lanczos.x
band_plot.x    fd_ef.x                        ld1.x                 pw2bgw.x           turbo_spectrum.x
bands_FS.x     fd_ifc.x                       manycp.x              pw2gw.x            unfold.x
bands.x        fd.x                           manypw.x              pw2wannier90.x     upf2casino.x
bgw2pw.x       fhi2upf.x                      matdyn.x              pw4gww.x           uspp2upf.x
blc2wan.x      fpmd2upf.x                     midpoint.x            pwcond.x           vdb2upf.x
casino2upf.x   fqha.x                         molecularpdos.x       pw_export.x        virtual.x
cmplx_bands.x  gcube2plt.x                    ncpp2upf.x            pwgui              wannier90.x
conductor.x    generate_rVV10_kernel_table.x  neb.x                 pwi2xsf.x          wannier_ham.x
cpmd2upf.x     generate_vdW_kernel_table.x    oldcp2upf.x           pw.x               wannier_plot.x
cppp.x         gww_fit.x                      path_interpolation.x  q2qstar.x          wannier.x
cp.x           gww.x                          pawplot.x             q2r.x              wfck2r.x
current.x      head.x                         phcg.x                q2trans_fd.x       wfdd.x
d3.x           importexport_binary.x          ph.x                  q2trans.x          wfk2etsf.x
decay.x        initial_state.x                plan_avg.x            read_upf_tofile.x  wfreq.x
disentangle.x  interpolate.x                  plotband.x            rrkj2upf.x         write_ham.x
dist.x         iotk_print_kinds.x             plotproj.x            sax2qexml.x        wstat.x
dos.x          iotk.x                         plotrho.x             sumpdos.x          xspectra.x
dynmat.x       kgrid.x                        plot.x                sum_sgm.x
embed.x        kpoints.x                      pmw.x                 turbo_davidson.x
epsilon.x      kvecs_FS.x                     pp.x                  turbo_eels.x

6.1

average.x     fpmd2upf.x                     manypw.x              projwfc.x             turbo_davidson.x
bands.x       fqha.x                         matdyn.x              pw2bgw.x              turbo_eels.x
bgw2pw.x      fs.x                           molecularnexafs.x     pw2gw.x               turbo_lanczos.x
bse_main.x    generate_rVV10_kernel_table.x  molecularpdos.x       pw2wannier90.x        turbo_spectrum.x
casino2upf.x  generate_vdW_kernel_table.x    ncpp2upf.x            pw4gww.x              upf2casino.x
cpmd2upf.x    gww_fit.x                      neb.x                 pwcond.x              uspp2upf.x
cppp.x        gww.x                          oldcp2upf.x           pw_export.x           vdb2upf.x
cp.x          head.x                         path_interpolation.x  pwi2xsf.x             virtual.x
dist.x        importexport_binary.x          pawplot.x             pw.x                  wannier_ham.x
dos.x         initial_state.x                phcg.x                q2qstar.x             wannier_plot.x
dynmat.x      interpolate.x                  ph.x                  q2r.x                 wfck2r.x
epsilon.x     iotk_print_kinds.x             plan_avg.x            q2trans_fd.x          wfdd.x
ev.x          iotk.x                         plotband.x            q2trans.x             xspectra.x
fd_ef.x       kpoints.x                      plotproj.x            read_upf_tofile.x
fd_ifc.x      lambda.x                       plotrho.x             rrkj2upf.x
fd.x          ld1.x                          pmw.x                 spectra_correction.x
fhi2upf.x     manycp.x                       pp.x                  sumpdos.x

[heatherkellyucl]

May just be able to copy over the pwgui binary as that comes precompiled.

[heatherkellyucl]

pwgui is a launcher script, and it needs to be able to access its other files. I think you need to copy its whole directory over. Either copy the launcher in to the main bin directory and export $PWGUI so it can find its root directory or add it to the PATH.

[heatherkellyucl]

Actually, make all still doesn't make all, so I've added in the ones we had before plus gui to see if that does everything.

[heatherkellyucl]

Some tests failed:

All done. ERROR: only 177 out of 181 tests passed (2 unknown).
Failed tests in:
        /dev/shm/quantum-espresso-build.h5og1z0a/qe-6.1/test-suite/pw_pawatom/
        /dev/shm/quantum-espresso-build.h5og1z0a/qe-6.1/test-suite/pw_vdw/

Ones I could still see were:

pw_scf - scf-1.in: Unknown.

_vdw - vdw-ts.in: **FAILED**.
p1
    ERROR: absolute error 1.50e-01 greater than 1.00e-01. (Test: 188.62.  Benchmark: 188.77.)
e1
    ERROR: absolute error 7.20e-05 greater than 1.00e-06. (Test: -44.618061.  Benchmark: -44.618133.)

[heatherkellyucl]

Tests re-running with output logged so the useful parts are visible... There is quite a lot of output so just looking in the two directories above wasn't that helpful without knowing what I am looking for.

[heatherkellyucl]

pw_pawatom - paw-vcbfgs.in: **FAILED**.
Different sets of data extracted from benchmark and test.
    More data in benchmark than in test: ef1.
p1
    ERROR: absolute error 2.20e-01 greater than 1.00e-01. (Test: -0.52.  Benchmark: -0.3.)
e1
    ERROR: absolute error 4.00e-05 greater than 1.00e-06. (Test: -328.23191.  Benchmark: -328.23187.)
n1
    ERROR: absolute error 2.00e+00 greater than 1.50e+00. (Test: 4.0.  Benchmark: 6.0.)

pw_vdw - vdw-ts.in: **FAILED**.
p1
    ERROR: absolute error 1.50e-01 greater than 1.00e-01. (Test: 188.62.  Benchmark: 188.77.)
e1
    ERROR: absolute error 7.20e-05 greater than 1.00e-06. (Test: -44.618061.  Benchmark: -44.618133.)

https://www.nsc.liu.se/systems/triolith/software/triolith-software-apps-espresso-6.1-build01.html says for the vdw-ts.in test to pass they had to compile Modules/tsvdw.f90 with -O0 instead of -O2. (They were also using Intel 2017, MKL and Intel MPI).

They also got the "More data in benchmark than in test: ef1." for paw-vcbfgs.in but no numerical errors for that one.

[heatherkellyucl]

I'm going to see whether a build with FFLAGS=-fp-model strict does any better before trying to build tsvdw alone with -O0.

[heatherkellyucl]

That fixed the vdw-ts.in test! paw-vcbfgs.in is exactly the same.

pw_pawatom - paw-vcbfgs.in: **FAILED**.
Different sets of data extracted from benchmark and test.
    More data in benchmark than in test: ef1.
p1
    ERROR: absolute error 2.20e-01 greater than 1.00e-01. (Test: -0.52.  Benchmark: -0.3.)
e1
    ERROR: absolute error 4.00e-05 greater than 1.00e-06. (Test: -328.23191.  Benchmark: -328.23187.)
n1
    ERROR: absolute error 2.00e+00 greater than 1.50e+00. (Test: 4.0.  Benchmark: 6.0.)

All done. ERROR: only 178 out of 181 tests passed (2 unknown).
Failed test in:
        /dev/shm/quantum-espresso-build.h5og1z0a/qe-6.1/test-suite/pw_pawatom/

[heatherkellyucl]

This would be the problem in the pawatom test...

<      Program PWSCF v.6.1 (svn rev. 13369) starts on  2Mar2017 at 23:52:57 
---
>      Program PWSCF v.6.1 (svn rev. 13369) starts on  5Jul2017 at 16:23:41 
11c11
<      Serial multi-threaded version, running on    4 processor cores
---
>      Parallel version (MPI), running on     1 processors

There are some +/- 0 differences at the start, one set of intermediate results a different way round, then a bunch of differences:

262c267
<   -0.000000000   2.892956709   2.892956709
---
>    0.000000000   2.892956709   2.892956709
264c269
<    2.892956709   2.892956709  -0.000000000
---
>    2.892956709   2.892956709   0.000000000

...
296c301
<      ethr =  1.21E-10,  avg # of iterations =  2.0
---
>      ethr =  6.07E-11,  avg # of iterations =  2.0
298c303
<      negative rho (up, down):  4.525E-02 0.000E+00
---
>      negative rho (up, down):  4.520E-02 0.000E+00
...
344c349
<      the Fermi energy is     5.0417 ev
---
>      the Fermi energy is     5.0414 ev
...

346,348c351,353
< !    total energy              =    -328.23191048 Ry
<      Harris-Foulkes estimate   =    -328.23000653 Ry
<      estimated scf accuracy    <          9.8E-09 Ry
---
> !    total energy              =    -328.23191046 Ry
>      Harris-Foulkes estimate   =    -328.23000637 Ry
>      estimated scf accuracy    <          4.9E-09 Ry
354,356c359,361
<      one-electron contribution =       4.98859345 Ry
<      hartree contribution      =       1.21001549 Ry
<      xc contribution           =     -32.27025092 Ry
---
>      one-electron contribution =       4.98876435 Ry
>      hartree contribution      =       1.21013745 Ry
>      xc contribution           =     -32.27050620 Ry
358,359c363,364
<      one-center paw contrib.   =    -286.39473022 Ry
<      smearing contrib. (-TS)   =       0.00004003 Ry
---
>      one-center paw contrib.   =    -286.39476787 Ry
>      smearing contrib. (-TS)   =       0.00004012 Ry
363c368
<      negative rho (up, down):  4.525E-02 0.000E+00
---
>      negative rho (up, down):  4.520E-02 0.000E+00
...

<      Computing stress (Cartesian axis) and pressure
< 
< 
<      negative rho (up, down):  4.528E-02 0.000E+00
<           total   stress  (Ry/bohr**3)                   (kbar)     P=   -0.23
<   -0.00000159  -0.00000000   0.00000000         -0.23     -0.00      0.00
<   -0.00000000  -0.00000159   0.00000000         -0.00     -0.23      0.00
<    0.00000000   0.00000000  -0.00000159          0.00      0.00     -0.23
---
>      negative rho (up, down):  4.520E-02 0.000E+00
>           total   stress  (Ry/bohr**3)                   (kbar)     P=   -0.49
>   -0.00000330   0.00000000   0.00000000         -0.49      0.00      0.00
>    0.00000000  -0.00000330  -0.00000000          0.00     -0.49     -0.00
>   -0.00000000  -0.00000000  -0.00000330         -0.00     -0.00     -0.49
521c388
<      bfgs converged in   3 scf cycles and   2 bfgs steps
---
>      bfgs converged in   2 scf cycles and   1 bfgs steps
526c393
<      Final enthalpy =    -328.2319111033 Ry
---
>      Final enthalpy =    -328.2319104638 Ry
528,529c395,396
<      new unit-cell volume =    326.58490 a.u.^3 (    48.39489 Ang^3 )
<      density =      4.98283 g/cm^3
---
>      new unit-cell volume =    326.77762 a.u.^3 (    48.42345 Ang^3 )
>      density =      4.97989 g/cm^3
532,534c399,401
<   -0.000000000   2.892387865   2.892387865
<    2.892387865  -0.000000000   2.892387865
<    2.892387865   2.892387865  -0.000000000
---
>    0.000000000   2.892956709   2.892956709
>    2.892956709  -0.000000000   2.892956709
>    2.892956709   2.892956709   0.000000000

754c598
<      the Fermi energy is     5.0472 ev
---
>      the Fermi energy is     5.0406 ev
756,758c600,602
< !    total energy              =    -328.23187026 Ry
<      Harris-Foulkes estimate   =    -328.23187026 Ry
<      estimated scf accuracy    <          4.9E-10 Ry
---
> !    total energy              =    -328.23191045 Ry
>      Harris-Foulkes estimate   =    -328.23191048 Ry
>      estimated scf accuracy    <       0.00000054 Ry
760c604
<      total all-electron energy =     -8395.996629 Ry
---
>      total all-electron energy =     -8395.996669 Ry
764,769c608,613
<      one-electron contribution =       4.99330785 Ry
<      hartree contribution      =       1.20913996 Ry
<      xc contribution           =     -32.27082136 Ry
<      ewald contribution        =     -15.76867892 Ry
<      one-center paw contrib.   =    -286.39485774 Ry
<      smearing contrib. (-TS)   =       0.00003995 Ry
---
>      one-electron contribution =       4.98883670 Ry
>      hartree contribution      =       1.20999392 Ry
>      xc contribution           =     -32.27043216 Ry
>      ewald contribution        =     -15.76557831 Ry
>      one-center paw contrib.   =    -286.39477081 Ry
>      smearing contrib. (-TS)   =       0.00004021 Ry
771c615
<      convergence has been achieved in   6 iterations
---
>      convergence has been achieved in   4 iterations
773c617
<      negative rho (up, down):  4.520E-02 0.000E+00
---
>      negative rho (up, down):  4.522E-02 0.000E+00
778c622
<      atom    2 type  1   force =     0.00000000   -0.00000000    0.00000000
---
>      atom    2 type  1   force =     0.00000000    0.00000000    0.00000000
786,790c630,634
<      negative rho (up, down):  4.520E-02 0.000E+00
<           total   stress  (Ry/bohr**3)                   (kbar)     P=   -0.30
<   -0.00000205   0.00000000   0.00000000         -0.30      0.00      0.00
<    0.00000000  -0.00000205  -0.00000000          0.00     -0.30     -0.00
<   -0.00000000  -0.00000000  -0.00000205         -0.00     -0.00     -0.30
---
>      negative rho (up, down):  4.522E-02 0.000E+00
>           total   stress  (Ry/bohr**3)                   (kbar)     P=   -0.52
>   -0.00000356   0.00000000   0.00000000         -0.52      0.00      0.00
>    0.00000000  -0.00000356   0.00000000          0.00     -0.52      0.00
>    0.00000000   0.00000000  -0.00000356          0.00      0.00     -0.52

So the 4 and 6 that were different in the test output appear to be the number of iterations each took to converge.

[heatherkellyucl]

I think we can ignore that test result as they aren't comparable simulations. There are different make options for running the tests - run-tests-serial is the default. It is possible we might want to do run-tests-pw-parallel which runs the pw tests only, in parallel. Still don't know if 4 MPI processes ought to be comparable to 4 threads for these.

[heatherkellyucl]

Nope, if you run make run-tests-pw-parallel 2>&1 | tee pw_par_test.log in the test-suite directory then more of the tests fail with larger absolute errors than expected, and pawatom is the same except the absolute error has increased from 2.20e-01 to 2.30e-01.

[heatherkellyucl]

Installed on

• [x] Thomas
• [x] Grace
• [x] Legion

[heatherkellyucl]

PWgui does still need copying in, so added that to buildscript and did the copy.

[heatherkellyucl]

pwgui requires Itcl, Itk and Iwidgets to run, though.

[cceahke@login01 ~]$ pwgui

 ==================================================
  This is PWgui version: 6.1
 --------------------------------------------------

 PWgui: using the system default "tclsh" interpreter

 PWGUI       : /shared/ucl/apps/quantum-espresso/6.1/intel-2017/PWgui-6.1
 GUIB engine : /shared/ucl/apps/quantum-espresso/6.1/intel-2017/PWgui-6.1/lib/Guib-0.6

can't find package Itk
    while executing
"package require Itk      "
    (file "/shared/ucl/apps/quantum-espresso/6.1/intel-2017/PWgui-6.1/lib/Guib-0.6/init.tcl" line 11)
    invoked from within
"source /shared/ucl/apps/quantum-espresso/6.1/intel-2017/PWgui-6.1/lib/Guib-0.6/init.tcl"
    ("package ifneeded Guib 0.6" script)
    invoked from within
"package require Guib 0.5"
    (file "/shared/ucl/apps/quantum-espresso/6.1/intel-2017/PWgui-6.1/init.tcl" line 5)
    invoked from within
"source [file join $env(PWGUI) init.tcl]"
    (file "/shared/ucl/apps/quantum-espresso/6.1/intel-2017/PWgui-6.1/pwgui.tcl" line 62)

[heatherkellyucl]

Have told the requestor that the rest of it is available, but pwgui not working yet.

Next week look at itk etc.

[Sneha112Banerjee]

In PWgui 6.1, '&bands' shows no options to choose "run calculation, run and configure calculation," etc in the Run tab. I am not sure what to do to correct it. Can anyone help me with this?

[heatherkellyucl]

@Sneha112Banerjee I think you might be asking this in the wrong place. Are you a user of University College London's clusters? We didn't install the rest of the requirements for PWgui so have no experience using it.

[Sneha112Banerjee]

I get that. Thank you for responding.

[owainkenwayucl]

Going to close for now as QE is installed. If someone needs PWGUI we'll open a new issue.