new file: f/FFTW/FFTW-3.3.8-NVHPC-21.9.eb

new file:   o/OpenMPI/OpenMPI-4.0.7-NVHPC-21.9-CUDA-11.4.1.eb
	new file:   q/QuantumESPRESSO/QuantumESPRESSO-7.1-NVHPC-21.9.eb

f/FFTW/FFTW-3.3.8-NVHPC-21.9.eb

@@ -0,0 +1,28 @@
+# IT4Innovations 2022
+# JK
+
+name = 'FFTW'
+version = '3.3.8'
+
+homepage = 'http://www.fftw.org'
+description = """FFTW is a C subroutine library for local_computing the discrete Fourier transform (DFT)
+ in one or more dimensions, of arbitrary input size, and of both real and local_complex data."""
+
+toolchain = {'name': 'NVHPC', 'version': '21.9'}
+toolchainopts = {'pic': True}
+
+source_urls = [homepage]
+sources = [SOURCELOWER_TAR_GZ]
+checksums = ['6113262f6e92c5bd474f2875fa1b01054c4ad5040f6b0da7c03c98821d9ae303']
+
+# no quad precision, requires GCC v4.6 or higher
+# see also
+# http://www.fftw.org/doc/Extended-and-quadruple-precision-in-Fortran.html
+with_quad_prec = False
+
+# local_compilation fails when configuring with --enable-avx-128-fma, Intel local_compilers do not support FMA4 instructions
+use_fma4 = False
+
+runtest = 'check'
+
+moduleclass = 'numlib'

o/OpenMPI/OpenMPI-4.0.7-NVHPC-21.9-CUDA-11.4.1.eb

@@ -0,0 +1,43 @@
+# IT4Innovations 2021
+# LK
+
+name = 'OpenMPI'
+version = '4.0.7'
+versionsuffix = '-CUDA-11.4.1'
+
+homepage = 'https://www.open-mpi.org/'
+description = """The Open MPI Project is an open source MPI-3 implementation."""
+
+toolchain = {'name': 'NVHPC', 'version': '21.9'}
+
+source_urls = ['https://www.open-mpi.org/software/ompi/v%(version_major_minor)s/downloads']
+sources = [SOURCELOWER_TAR_GZ]
+
+dependencies = [
+    ('UCX', '1.11.2', versionsuffix),
+    ('CUDAcore', '11.4.1', '', True),
+]
+
+preconfigopts = 'export CC="pgcc -noswitcherror" && export CXX="pgc++ -noswitcherror" && export FC="pgfortran -noswitcherror" && '
+
+configopts = '--enable-shared --enable-mpi-thread-multiple --with-verbs '
+configopts += '--with-hwloc=$EBROOTHWLOC '  # hwloc support
+configopts += '--enable-mpirun-prefix-by-default '
+configopts += '--with-tm=/opt/pbs '  # Enable PBS
+configopts += '--enable-mpi-cxx '  # Enable building the C++ MPI bindings
+configopts += '--with-ucx=$EBROOTUCX '
+configopts += '--with-cuda=$EBROOTCUDACORE '
+
+moduleclass = 'mpi'
+
+#setenv("OMPI_MCA_btl_openib_if_include", "mlx5_0")
+#setenv("export OMPI_MCA_btl_openib_allow_ib", "1")
+#setenv("OMPI_MCA_pml", "^ucx")
+#setenv("OMPI_MCA_orte_base_help_aggregate", "0")
+import os
+if os.environ.get("CLUSTERNAME") in ["BARBORA"]:
+   modextravars = {'OMPI_MCA_btl_openib_if_include': 'mlx5_0',
+                'OMPI_MCA_btl_openib_allow_ib': '1',
+                'OMPI_MCA_orte_base_help_aggregate': '0',
+                'OMPI_MCA_pml': '^ucx',
+   }

q/QuantumESPRESSO/QuantumESPRESSO-7.1-NVHPC-21.9.eb

@@ -0,0 +1,70 @@
+# IT4Innovations 2022
+# JK
+
+easyblock = 'CMakeMake'
+name = 'QuantumESPRESSO'
+version = '7.1'
+
+homepage = 'https://www.quantum-espresso.org'
+description = """Quantum ESPRESSO  is an integrated suite of computer codes
+for electronic-structure calculations and materials modeling at the nanoscale.
+It is based on density-functional theory, plane waves, and pseudopotentials
+(both norm-conserving and ultrasoft).
+"""
+
+# NVHPC 21.11-22.2 have a severe bug causing hanging in runs when QE is
+# compiled with both OpenMP and OpenACC.  Use a different compiler release or
+# dislable OpenMP with potential performance loss.
+toolchain = {'name': 'NVHPC', 'version': '21.9'}
+
+import os
+if os.environ.get("CLUSTERNAME") in ["KAROLINA"]:
+   toolchainopts = {'pic': True,  'optarch': 'march=core-avx2'}
+else:
+   toolchainopts = {'pic': True}
+
+if os.environ.get("CLUSTERNAME") in ["KAROLINA"]:
+   prebuildopts = "echo %(builddir)s && while read i; do echo $i; sed 's|-xHost|-march=core-avx2|g' -i $i; done < <(grep xHost %(builddir)s -R | cut -d ':' -f 1 | sort -u) && "
+
+source_urls = ['https://www.quantum-espresso.org/rdm-download/488/v7-1/3adf3b73b84938d19f4b7ac856795acb/']
+sources = ['qe-%(version)s-ReleasePack.tar.gz']
+checksums = ['feacdbc67d084d55df464f989a916f20dfe11a50ccfda782573cdeed4fab3d3a']
+
+builddependencies = [
+    ('CMake', '3.20.1', '', ('GCCcore', '10.2.0')),
+    ('git', '2.28.0', '-nodocs', ('GCCcore', '10.2.0')),
+]
+
+dependencies = [
+#    ('ELPA', '2021.05.001', '', ('intel', '2021a')),
+#    ('libxc', '5.1.3', '', ('GCC', '10.2.0')),
+#    ('HDF5', '1.10.7', '', ('iimpi', '2021a')),
+    ('OpenMPI', '4.0.7', '-CUDA-11.4.1'),
+]
+
+preconfigopts = " export MPIF90=mpif90 && "
+preconfigopts += " export MPIFC=mpif90 && "
+preconfigopts += " export MPIF77=mpif90 && "
+preconfigopts += " export MPICC=mpicc && "
+preconfigopts += " export MPICXX=mpicxx && "
+
+configopts = 'FC=pgfortran F77=pgfortran F90=pgfortran CC=pgcc CXX=pgc++ '
+configopts += '-DQE_ENABLE_CUDA=ON -DQE_ENABLE_MPI_GPU_AWARE=ON '
+configopts += '-DNVFORTRAN_CUDA_VERSION=11.4 '
+configopts += '-DNVFORTRAN_CUDA_CC=80 '
+
+
+
+# --with-cuda=$CUDA_HOME --with-cuda-cc=80 --with-cuda-runtime=11.4 --enable-openmp'
+
+prebuildopts = "sed -i 's/-D__MPI\\b/& -D__GPU_MPI /' %(builddir)s/qe-%(version)s/make.inc && "
+
+# only pw is available for GPU
+buildopts = 'pw'
+
+# parallel build tends to fail
+parallel = 1
+
+#skipsteps = ['sanitycheck']
+
+moduleclass = 'chem'