Differences

This shows you the differences between two versions of the page.

--- quantum_espresso [2020/11/20 22:01] – root
+++ quantum_espresso [2025/10/15 19:51] (current) – external edit 127.0.0.1
@@ Line 1: / Line 1: @@
 ===== Quantum Espresso =====
-Version 5.1
-** Compilation **
-With Intel compiler and either OpenMPI or MVAPICH2:
+Versions 6.8/7.1
-<code>
-OpenMPI:
-DFLAGS         = -D__INTEL -D__FFTW3 -D__MPI -D__PARA -D__SCALAPACK $(MANUAL_DFLAGS)
-IFLAGS         = -I../include
-MPIF90         = mpif90
-CFLAGS         = -O3 -xSSE2 -axavx $(DFLAGS) $(IFLAGS)
-F90FLAGS       = $(FFLAGS) -nomodule -fpp $(FDFLAGS) $(IFLAGS) $(MODFLAGS)
-FFLAGS         = -O2 -xSSE2 -axavx -assume byterecl -g -traceback -par-report0 -vec-report0
-FFLAGS_NOOPT   = -O0 -assume byterecl -g -traceback
-FFLAGS_NOMAIN  = -nofor_main
-LD             = mpif90
-LDFLAGS        = -static-intel
-SCALAPACK_LIBS = -lmkl_scalapack_lp64 -lmkl_blacs_openmpi_lp64
-FFT_LIBS       = -L ${MKL_ROOT}/interfaces/fftw3xf -lfftw3xf_intel
-MVAPICH2: same except
+** Compilation **
-SCALAPACK_LIBS = -lmkl_scalapack_lp64 -lmkl_blacs_intelmpi_lp64
+With Intel compiler, Intel MPI, and MKL
-trestles: same except
-no -axavx (though an "optional" code path, it makes the program fail on AMD)
-</code>
-** Benchmarks **
-We run AUSURF112 from [[http://qe-forge.org/gf/project/q-e/frs/?action=FrsReleaseBrowse&frs_package_id=36|Espresso Benchmarks]] and compare with [[http://glennklockwood.blogspot.com/2014/02/quantum-espresso-performance-benefits.html|Glenn Lockwood]] who ran the AUSURF112 benchmark on SDSC Comet and on the Trestles system when it was at SDSC.  Unfortunately, the AUSURF112 benchmark generally ends the simulation with
-''convergence NOT achieved after   2 iterations: stopping'',
-but it does so fairly repeatably so may be timed.
 <code>
-OpenMPI:
+#COMPUTER=skylake
-module load intel/14.0.3 mkl/14.0.3 openmpi/1.8.8
+#OPT="-xHOST"
-mpirun -np 64  -machinefile $PBS_NODEFILE -x LD_LIBRARY_PATH \
+COMPUTER=bulldozer
-/share/apps/espresso/espresso-5.1-intel-openmpi/bin/pw.x -npools 1 <ausurf.in
+OPT="-msse3 -axsse3,sse4.2,AVX,core-AVX2,CORE-AVX512"
-MVAPICH2:
+VERSION=7.1
-module load intel/14.0.3 mkl/14.0.3 mvapich2/2.1
+HDF5=1.12.0
-mpirun -np 64  -machinefile $PBS_NODEFILE \
+module purge
-/share/apps/espresso/espresso-5.1-intel-mvapich2/bin/pw.x -npools 1 <ausurf.in
+module load intel/19.0.5 mkl/20.0.4 impi/17.0.4
+OMP="--enable-openmp"
+make clean
+./install/configure MPIF90=mpiifort F90=ifort F77=ifort FC=ifort CC=icc \
+SCALAPACK_LIBS="-L$MKLROOT/lib/intel64 -lmkl_scalapack_lp64 -lmkl_blacs_intelmpi_lp64" \
+LAPACK_LIBS="-L$MKLROOT/lib/intel64 -lmkl_lapack95_lp64 -lmkl_blas95_lp64" \
+BLAS_LIBS="-lmkl_intel_lp64  -lmkl_intel_thread -lmkl_core -liomp5 -thread" \
+FFT_LIBS="-L$MKLROOT/interfaces/fftw3xf -lfftw3xf_intel" \
+FFLAGS="-O3 $OPT -D__INTEL -D__GNUC__ -D__FFTW3 -D__MPI -D__PARA -D__SCALAPACK -assume byterecl \-I$MKLROOT/include/fftw" \
+CFLAGS="-O3 $OPT -D__INTEL -D__GNUC__ -D__FFTW3 -D__MPI -D__PARA -D__SCALAPACK" \
+--with-hdf5=/share/apps/hdf5/$HDF5/intel/impi -with-scalapack=intel --enable-parallel \
+$OMP --prefix=/share/apps/espresso/espresso-$VERSION-intel-impi-mkl-$COMPUTER
+make depends
+make all
+make install
 </code>
-The tables shows Lockwood's and our times.  We add 32x4 core runs for Trestles as we think node-to-node is the more representative comparison.  Our newer versions of OpenMPI show better results on the almost-identical hardware than Lockwood.
-<csv>
-Walltime,CoresxNodes,Intel/Mvapich, Intel/OpenMPI
-Lockwood Gordon E5-2670,16x4,470,580
-Lockwood Trestles AMD6136,32x2,1060,1440
-Our E5-2650V2,16x4,na,475
-Our E5-2670,16x4,456,488
-Our Trestles AMD6136,32x2,(1),1007
-Our Trestles AMD6136,32x4,642,762
-</csv>
-(1) Fails with error [[http://www.quantum-espresso.org/faq/frequent-errors-during-execution/#5.6|charge is wrong]].
-** Notes **
-Each run fails with error messages (depending on MPI type) and RC 1 after terminating normally according to the log. This appears harmless:
+Runtime:
 <code>
-   This run was terminated on:  13: 2:44  11Nov2015
+module load intel/18.0.2 impi/17.0.4 mkl/20.0.4 {qe/7.1 or qe/6.8}
-=------------------------------------------------------------------------------=
+trestles:module load intel/18.0.2 impi/17.0.4 mkl/20.0.1 {qe/7.1 or qe/6.8}
-   JOB DONE.
-=------------------------------------------------------------------------------=
--------------------------------------------------------
-Primary job  terminated normally, but 1 process returned
-a non-zero exit code.. Per user-direction, the job has been aborted.
--------------------------------------------------------
-------------------------------------------------------------
-A process or daemon was unable to complete a TCP connection
-to another process:
-etc.
 </code>
-** Continuing Work **
+The performance is not sensitive to qe version between 6.8 and 7.1, but is quite sensitive to MKL version.  Newest MKL (20.0.4) is best on all platforms except on trestles (20.0.1) is best.  There are two executable sets selected by the module at runtime ("skylake" for Pinnacle-I and "bulldozer" for all other platforms).  Performance with OpenMP is slightly slower.
-ELPA in newer versions of Espresso is reportedly faster than Scalapack.
-OpenMPI threading.
-MKL threading.
-FFTW fft vs. Intel fft on AMD.
-=== 2020 Update q-e 6.6===
-On Trestles with Intel tools.  It's difficult to find a combination of versions that is new enough to compile qe-6.6 (compiler > 18) yet will still produce a binary that runs on AMD Bulldozer (mkl <20) while avoiding most mkl bugs (mkl > 18).
+The AUSURF112 benchmark is used for comparison with "-nk 2" and both CPUs on one node
 <code>
-module load intel/18.0.2 mkl/19.0.5 impi/17.0.4
+System     QE version cores OMP  time
-MKL_NUM_THREADS=# OMP_NUM_THREADS=# mpirun -np ## -machinefile machinefile /share/apps/espresso/q-e-qe-6.6/bin/pw.x <ausurf.in
+Pinnacle II-AMD7543  7.1 64   1    86
+Pinnacle II-AMD7543  7.1 32   2    89
+Pinnacle I-Intel6130 7.1 32   1   133
+Pinnacle I-Intel6130 7.1 16   2   137
+Trestles-AMD6136     7.1 32   1   718
+Trestles-AMD6136     7.1 16   2   858
 </code>
-q-e appears to be a code that does not like mpi threads x OMP threads > physical cores.
-Performance on two trestles nodes is a lot better than the previous 5.1 benchmarks, but it doesn't scale past two nodes on this small problem.
-<code>
-cores Node type    #mpi   #nodes  #OMP #MKL   WALL
-  Trestles AMD    32       1      1    1  12m11s
-  Trestles AMD    32       1      2    1  50m41s
-  Trestles AMD    32       1      1    2  >14m
-  Trestles AMD    16       1      2    1  >14m
-  Trestles AMD    64       2      1    1   8m33s
-  Trestles AMD   128       4      1    1   8m42s
-</code>
-Install script
-<code>
-OMP="--enable-openmp"
-./install/configure MPIF90=mpiifort F90=ifort F77=ifort FC=ifort CC=icc SCALAPACK_LIBS="-L$MKLROOT/lib/intel64 -lmkl_scalapack_lp64 -lmkl_blacs_intelmpi_lp64" LAPACK_LIBS="-L$MKLROOT/lib/intel64 -lmkl_lapack95_lp64 -lmkl_blas95_lp64" BLAS_LIBS="-lmkl_intel_lp64  -lmkl_intel_thread -lmkl_core -liomp5 -thread" FFT_LIBS="-L$MKLROOT/interfaces/fftw3xf -lfftw3xf_intel" FFLAGS="-O3 -xHOST -D__INTEL -D__GNUC__ -D__FFTW3 -D__MPI -D__PARA -D__SCALAPACK -assume byterecl -I$MKLROOT/include/fftw" CFLAGS="-O3 -xHOST -D__INTEL -D__GNUC__ -D__FFTW3 -D__MPI -D__PARA -D__SCALAPACK" --with-hdf5=/share/apps/hdf5/1.10.5/intel/impi -with-scalapack=intel --enable-parallel $OMP --prefix=/share/apps/espresso/espresso-6.6-intel-impi-mkl-trestles
-</code>