New Major BerkeleyGW-4.0 Release
Mauro Del Ben
Dear All,
We are happy to announce the major 4.0 release of the BerkeleyGW software package using the GW and GW plus Bethe-Salpeter equation (GW-BSE) approaches and beyond for the ab-initio computation of quasiparticle excitations and optical responses of materials. BerkeleyGW is a versatile, massively parallel, GPU-enabled, modular software package based on quantum many-body perturbation theory. Whether you are investigating bulk crystals, clusters, or 2D/1D materials, BerkeleyGW caters to your needs, adapting to different dimensionalities and to insulating, metallic, and semi-metallic systems.
Our latest release includes new methods and algorithms to significantly speed up calculations, supports a wider range of GPUs, implements novel approaches to study new physical phenomena, and interfaces with more external tools to increase researcher usability and productivity.
The source code is available at https://berkeleygw.org/download/ and distributed under a permissive free software license.
Below are some highlights of the new features included in the 4.0 release:
- Full GPU acceleration for the entire GW and GW-BSE workflow using portable programming models, supporting NVIDIA, AMD, and Intel GPUs.
- Stochastic pseudoband methods and tools to accelerate convergence with respect to empty states. Efficient implementation in the parabands code.
- Portable GPU implementation of the epsilon full-frequency static subspace approximation, including the capability to evaluate the RPA correlation energy.
- New implementation in the epsilon code to overcome the cubic scaling memory bottleneck (NV-block algorithm).
- Implementation of the partial occupations for efficient treatment of metallic systems.
- Patched sampling method to accelerate the convergence of exciton binding energies and wavefunctions with respect to k-point sampling.
- Capability to include external screening (such as those from a substrate or liquid environment) to that of the intrinsic material electronic one calculated by BerkeleyGW.
- Interface to the PRIMME library including GPU offload of the BSE's matvec driver for efficient iterative diagonalization.
- New tools, such as interfacing to the Wannier90 code, analyzing circularly polarized optical properties, exciton-phonon coupling, and performing wavefunction self-consistent calculations.
- Improved documentation (http://manual.berkeleygw.org/4.0/) and testing for new and existing features, as well as expanded set of examples (https://github.com/BerkeleyGW/BerkeleyGW-examples).
Sincerely,
The BerkeleyGW Development Team
Amos Afugu
I have compiled its on NERCS using GPU modules but i still get an error when I am running epsilon calculations.
# arch.mk for BerkeleyGW codes
#
# suitable for Cori-GPU
#
# MDB
# 2021, NERSC
# PGI compiler
#
# Do:
# module purge ; module load cgpu ; module load nvhpc/21.3 ; module load cuda/11.2.2 ; module load openmpi/4.0.3
#
COMPFLAG = -DPGI
PARAFLAG = -DMPI -DOMP
MATHFLAG = -DUSESCALAPACK -DUNPACKED -DUSEFFTW3 -DHDF5 -DOPENACC
FCPP = /usr/bin/cpp -P -ansi -nostdinc -C -E -std=c11
F90free = ftn -Mfree -acc -mp -target-accel=nvidia80 -gpu=cc80 -Mcuda=cuda11.7 -Mcudalib=cublas,cufft -Mcuda=lineinfo -traceback -Minfo=mp
LINK = ftn -acc -mp -target-accel=nvidia80 -gpu=cc80 -Mcuda=cuda11.7 -Mcudalib=cublas,cufft -Minfo=mp
FOPTS = -fast -Mfree -Mlarge_arrays
FNOOPTS = $(FOPTS)
#MOD_OPT = -J
MOD_OPT = -module
INCFLAG = -I
C_PARAFLAG = -DPARA -DMPICH_IGNORE_CXX_SEEK
CC_COMP = CC
C_COMP = cc
C_LINK = CC
C_OPTS = -fast -mp
C_DEBUGFLAG =
REMOVE = /bin/rm -f
MKLLIB = /global/homes/a/aafugu/BGW-all-files/intel/oneapi/mkl/2021.4.0/lib/intel64
MKLINC = /global/homes/a/aafugu/BGW-all-files/intel/oneapi/mkl/2021.4.0/include
FFTWLIB = $(MKLLIB)/libmkl_scalapack_lp64.a -Wl,--start-group $(MKLLIB)/libmkl_cdft_core.a $(MKLLIB)/libmkl_intel_lp64.a $(MKLLIB)/libmkl_pgi_thread.a $(MKLLIB)/libmkl_core.a $(MKLLIB)/libmkl_blacs_intelmpi_lp64.a -Wl,--end-group -pgf90libs -mp -lpthread -lm -ldl
FFTWINCLUDE = $(MKLINC)/fftw $(MKLINC) $(MKLINC)/intel64
PERFORMANCE =
HDF5_LDIR = /global/homes/a/aafugu/BGW-all-files/hdf5-1.10.7-GPU/build/lib
HDF5INCLUDE = /global/homes/a/aafugu/BGW-all-files/hdf5-1.10.7-GPU/build/include
HDF5LIB = $(HDF5_LDIR)/libhdf5hl_fortran.a \
$(HDF5_LDIR)/libhdf5_hl.a \
$(HDF5_LDIR)/libhdf5_fortran.a \
$(HDF5_LDIR)/libhdf5.a -lz -ldl