Block Installation

Source code and binary

Compile

Building Block code requires C++11 compiler, BLAS, LAPACK and Boost libraries. For distributed-memory parallel compilation, MPI and multi-threading Boost-MPI libraries are needed. When you compile Boost-MPI, you can use the following flags for b2 the Boost built-in compilation tool to generate the multi-threading -mt libraries.:

./b2 --layout=tagged link=static,shared threading=multi install

Note Boost and MPI libraries must be compiled using the same compiler as for compiling Block. See also boost documents for details of the installation of Boost with the MPI components.

To compile Block code, the following customizations need to be made to the makefile placed in the main directory.:

CXX = g++
BOOSTINCLUDE = /lib64/boost_1_55_0/include/
BOOSTLIB = /lib64/boost_1_55_0/lib/
OPENMP = yes

Please note that when choosing your compiler, either GNU or Intel, C++0x/C++11 standards must be appropriately supported, as Block requires new features for some of the modules (eg, npdm, nevpt2, etc). Here are our suggested minimum GNU/Intel compiler versions in order for the compiling process to be successful:

  • GNU g++: 4.8 or newer,

  • or Intel icpc: at least 14.0.1 (2013 SP1 Update 1) or newer.

Turn on the MPI support for Block code:

USE_MPI = yes
MPICXX = mpicxx

And supply MKL libraries:

USE_MKL = yes
MKLLIB = /opt/intel/composer_xe_2013_sp1.0.080/mkl/lib/intel64/``
MKLFLAGS = /opt/intel/composer_xe_2013_sp1.0.080/mkl/include``

For certain compilers, you may have error message:

error: ‘auto_ptr’ is deprecated (declared at /usr/include/c++/4.8.2/backward/auto_ptr.h:87) [-Werror=deprecated-declarations]

It is caused by the flag -Werror. It is safe to remove this flag from OPT variable. Some compiler/linker may issue errors if OPENMP = yes was specified in Makefile:

/usr/bin/ld: dmrg.o: undefined reference to symbol 'shm_openn@@GLIBC_2.2.5'

Appending -lpthread -lrt at the end of LIBS can solve this problem.

When the makefile is configured, run in the directory ./Block:

$ make

The successful compilation generates the executable block.spin_adapted, static and shared DMRG libraries libqcdmrg.a and libqcdmrg.so.

Interface to PySCF package

The electronic structure Python module PySCF provided an interface to run Block code. You need create a pyscf config file /path/to/pyscf/future/dmrgscf/settings.py and add the following settings in it:

BLOCKEXE = "/path/to/Block/block.spin_adapted"
BLOCKEXE_COMPRESS_NEVPT = "/path/to/serially/compiled/Block/block.spin_adapted"
BLOCKSCRATCHDIR = "/path/to/scratch"
MPIPREFIX = "mpirun"

BLOCKEXE is the parallel Block program. Most DMRG calculations (DMRG-CASCI, DMRG-CASSCF etc) will call this parallel executable through mpirun interface. BLOCKEXE_COMPRESS_NEVPT points to the serially compiled Block executable. It is only needed by the compressed perturber NEVPT2 method. Note the serial version of Block is slightly different to the parallel Block with compiling flag USE_MPI=yes. Specifically, NEVPT2 implementation in the serial code is not available in the parallelized code. Although this Block executable file is not MPI-parallelized, the DMRG-NEVPT2 program are efficiently parallelized in a different manner. Note the parameter MPIPREFIX should be adjusted according to your job scheduler, eg:

# For OpenPBS/Torque
MPIPREFIX = "mpirun"
# For SLURM
MPIPREFIX = "srun"

If calculation is carried out on interactive node, eg with 4 processors, the setting looks like:

MPIPREFIX = "mpirun -n 4"

With the Block-PySCF interface, a simple DMRG-SCF calculation can be input in Python interpereter:

>>> from pyscf import gto, scf, dmrgscf
>>> mf = gto.M(atom='C 0 0 0; C 0 0 1', basis='ccpvdz').apply(scf.RHF).run()
>>> mc = dmrgscf.dmrgci.DMRGSCF(mf, 6, 6)
>>> mc.run()

DMRG-NEVPT2 calculation can be applied:

>>> from pyscf import mrpt
>>> mrpt.NEVPT(mc).compress_approx().run()

Run Block in cmdline

The standalone serial code can be executed running:

$ block.spin_adapted input.dat > output.dat

input.dat is the input file and the output of the program is piped into the output file output.dat.

The MPI parallel mode can be called running:

$ mpirun -np 4 block.spin_adapted input.dat > output.dat

Testjobs

Tests are placed in the directory ./Block/dmrg_tests:

$ cd dmrg_tests
$ ./runtest

The tests require Python to be installed on the system.