VASP Forum

I recently reinstalled my Linux workstation to a more recent Ubuntu version as the previous one was reaching end of support and had many bugs making some functionality unusable. I had to install the most recent (2025) version of intel oneapi base toolkit and hpc toolkit, as older ones are no longer available. I tried recompiling VASP 5.4.4.pl2, which is still the version I work wirh, checked that the makefile.include points to correct folders in the new toolkit and got an error that "ifort" is an unknown command. Turns out, intel has stopped including ifort and only includes ifx, which offloads some work to the GPU. I have not seen anything about this error in the wiki. Are there any workarounds for recompiling VASP with the most recent intel toolkits?

Hello!

Unfortunately, we do not have any installation instructions for this old VASP version available, I would highly recommend to upgrade. However, if you have a recipe for older Intel compiler versions you may still try to modify them for the latest oneAPI. Although Intel switched to their new compiler ifx instead of the now removed "classic" ifort this does not mean that the new compiler only works with offloading. By default, ifx still targets only CPUs and only if you add special compiler flags a GPU offloading is performed. Please be warned that there may be more obstacles if you proceed with this attempt. For example, there will be compiler flags in your makefile.include which were renamed or do not even have an equivalent flag in the current oneAPI compilers.

All the best,
Andreas Singraber

This worked for me:
# Precompiler options
CPP_OPTIONS= -DHOST=\"LinuxIFX\" \
-DMPI -DMPI_BLOCK=64000 \
-Duse_collective \
-DscaLAPACK \
-DCACHE_SIZE=32000 \
-Davoidalloc \
-Duse_bse_te \
-Dtbdyn \
-Duse_shmem

CPP = fpp -f_com=no -free -w0 $*$(FUFFIX) $*$(SUFFIX) $(CPP_OPTIONS)

# Fortran Compiler
FC = mpiifx
FCL = mpiifx -qmkl=sequential -lstdc++

FREE = -free -names lowercase
FFLAGS = -assume byterecl -w -xHOST
OFLAG = -O2
OFLAG_IN = $(OFLAG)
DEBUG = -O0

# MKL Math Libraries (SCALAPACK + BLACS)
MKL_PATH = $(MKLROOT)/lib
SCALAPACK = $(MKL_PATH)/libmkl_scalapack_lp64.a -lmkl_blacs_intelmpi_lp64

INCS = -I$(MKLROOT)/include/fftw -I/usr/include
LLIBS = $(SCALAPACK)

# Objects
OBJECTS = fftmpiw.o fftmpi_map.o fft3dlib.o fftw3d.o

# vasp.5.lib
CPP_LIB = $(CPP)
FC_LIB = $(FC)
CC_LIB = icx
CFLAGS_LIB = -O
FFLAGS_LIB = -O1
FREE_LIB = $(FREE)
OBJECTS_LIB = linpack_double.o getshmem.o

# Parser library
CXX_PARS = icpx
CC_PARSER = icx
CFLAGS_PARSER = -x c -O
CXXFLAGS_PARSER = -std=c++17 -I/usr/include/c++/13 -I/usr/include/x86_64-linux-gnu/c++/13 -O -Wall -DYY_parse_DEBUG=1
LIBS += parser
LLIBS += -Lparser -lparser -lstdc++

# Source and Binary Paths
SRCDIR = ../../src
BINDIR = ../../bin

# GPU Flags (disabled)
CPP_GPU = -DCUDA_GPU -DRPROMU_CPROJ_OVERLAP -DUSE_PINNED_MEMORY -DCUFFT_MIN=28 -UscaLAPACK
OBJECTS_GPU = fftmpiw.o fftmpi_map.o fft3dlib.o fftw3d_gpu.o fftmpiw_gpu.o

CC = icx
CXX = icpx
CFLAGS = -fPIC -DADD_ -Wall -qopenmp -DMAGMA_WITH_MKL -DMAGMA_SETAFFINITY -DGPUSHMEM=300 -DHAVE_CUBLAS -std=c++17

CUDA_ROOT ?= /usr/local/cuda/
NVCC := $(CUDA_ROOT)/bin/nvcc -ccbin=icx
CUDA_LIB := -L$(CUDA_ROOT)/lib64 -lnvToolsExt -lcudart -lcuda -lcufft -lcublas

GENCODE_ARCH := -gencode=arch=compute_30,code=\"sm_30,compute_30\" \
-gencode=arch=compute_35,code=\"sm_35,compute_35\" \
-gencode=arch=compute_60,code=\"sm_60,compute_60\"

MPI_INC = $(I_MPI_ROOT)/include64

CPPFLAGS = -I/usr/include/c++/13 -I/usr/include/x86_64-linux-gnu/c++/13

Hello!

This issue has also been discussed in another topic, please see here:

https://vasp.at/forum/viewtopic.php?t=20273

All the best,
Andreas Singraber