The statistics are as attached in the screenshot: I tested IVDW 10, 11, 12, 20, 21, and 202, with LSORBIT = .TRUE. (spin-orbit coupling).
For DFT-Dx corrections, the GPU and CPU runs are giving the same answers. However, for Tkatchenko-Scheffler and many-body corrections, the GPU results deviate from the CPU ones, or even crashed. My colleague tested version 6.4.0 on another GPU machine, and the number agreed with CPU (IVDW = 20). To the best of my observation, this issue does not appear if LSORBIT = .FALSE.
The four input files for my test system are attached here, as well as the makefile.include file that I used to compile the code on GPU. Could the VASP team, or anyone else who is interested, try running them and see if the same problem occurs? Thanks.
Code: Select all
# Default precompiler options
CPP_OPTIONS = -DHOST=\"LinuxNV\" \
-DMPI -DMPI_INPLACE -DMPI_BLOCK=8000 -Duse_collective \
-DscaLAPACK \
-DCACHE_SIZE=4000 \
-Davoidalloc \
-Dvasp6 \
-Duse_bse_te \
-Dtbdyn \
-Dqd_emulate \
-Dfock_dblbuf \
-D_OPENMP \
-D_OPENACC \
-DUSENCCL -DUSENCCLP2P
CPP = nvfortran -Mpreprocess -Mfree -Mextend -E $(CPP_OPTIONS) $*$(FUFFIX) > $*$(SUFFIX)
# N.B.: you might need to change the cuda-version here
# to one that comes with your NVIDIA-HPC SDK
FC = mpif90 -acc -gpu=cc60,cc70,cc80,cuda12.2 -mp
FCL = mpif90 -acc -gpu=cc60,cc70,cc80,cuda12.2 -mp -c++libs
FREE = -Mfree
FFLAGS = -Mbackslash -Mlarge_arrays
OFLAG = -fast
DEBUG = -Mfree -O0 -traceback
OBJECTS = fftmpiw.o fftmpi_map.o fftw3d.o fft3dlib.o
LLIBS = -cudalib=cublas,cusolver,cufft,nccl -cuda
# Redefine the standard list of O1 and O2 objects
SOURCE_O1 := pade_fit.o minimax_dependence.o
SOURCE_O2 := pead.o
# For what used to be vasp.5.lib
CPP_LIB = $(CPP)
FC_LIB = nvfortran
CC_LIB = nvc -w
CFLAGS_LIB = -O
FFLAGS_LIB = -O1 -Mfixed
FREE_LIB = $(FREE)
OBJECTS_LIB = linpack_double.o
# For the parser library
CXX_PARS = nvc++ --no_warnings
##
## Customize as of this point! Of course you may change the preceding
## part of this file as well if you like, but it should rarely be
## necessary ...
##
# When compiling on the target machine itself , change this to the
# relevant target when cross-compiling for another architecture
VASP_TARGET_CPU ?= -tp host
FFLAGS += $(VASP_TARGET_CPU)
# Specify your NV HPC-SDK installation (mandatory)
#... first try to set it automatically
NVROOT = /home/libraries/nvhpc_2023_237_Linux_x86_64_cuda_12.2/hpc_sdk/Linux_x86_64/23.7
# If the above fails, then NVROOT needs to be set manually
#NVHPC ?= /opt/nvidia/hpc_sdk
#NVVERSION = 21.11
#NVROOT = $(NVHPC)/Linux_x86_64/$(NVVERSION)
## Improves performance when using NV HPC-SDK >=21.11 and CUDA >11.2
OFLAG_IN = -fast -Mwarperf
SOURCE_IN := nonlr.o
# Software emulation of quadruple precsion (mandatory)
QD ?= $(NVROOT)/compilers/extras/qd
LLIBS += -L$(QD)/lib -lqdmod -lqd
INCS += -I$(QD)/include/qd
# BLAS (mandatory)
BLAS = -lblas
# LAPACK (mandatory)
LAPACK = -llapack
# scaLAPACK (mandatory)
SCALAPACK = -Mscalapack
LLIBS += $(SCALAPACK) $(LAPACK) $(BLAS)
# FFTW (mandatory)
FFTW_ROOT ?= /home/libraries/fftw/gpu/fftw_install
LLIBS += -L$(FFTW_ROOT)/lib -lfftw3 -lfftw3_omp
INCS += -I$(FFTW_ROOT)/include
# HDF5-support (optional but strongly recommended)
CPP_OPTIONS+= -DVASP_HDF5
HDF5_ROOT ?= /home/libraries/hdf5/gpu/hdf5_install
LLIBS += -L$(HDF5_ROOT)/lib -lhdf5_fortran
INCS += -I$(HDF5_ROOT)/include
# For the VASP-2-Wannier90 interface (optional)
CPP_OPTIONS += -DVASP2WANNIER90
WANNIER90_ROOT ?= /home/ibraries/wannier90-3.1.0/gpu
LLIBS += -L$(WANNIER90_ROOT)/lib -lwannier