scalapack4py 0.2.0

Python wrapper for dynamically loaded ScaLAPACK and BLACS libraries

Dynamic ScaLAPACK wrapper for Python

Python wrapper for dynamically loaded ScaLAPACK and BLACS libraries

from scalapack4py import ScaLAPACK4py, parprint, ordprint
from ctypes import cast, py_object, CDLL, RTLD_GLOBAL

scalapack_lib = CDLL('libscalapack-openmpi.so.2.0', mode=RTLD_GLOBAL)
sl = ScaLAPACK4py(scalapack_lib)


descr = sl.wrap_blacs_desc(descr)
locshape = (descr.locrow, descr.loccol)
data = np.ctypeslib.as_array(data, shape=locshape)
sl.scatter(data_src, descr, data)

Or

# run as:  mpiexec -n 2 python3 -u test_scatter_complex.py
import numpy as np, os
from mpi4py import MPI
from scalapack4py import ScaLAPACK4py
from ctypes import CDLL, RTLD_GLOBAL, POINTER, c_int, c_double

sl = ScaLAPACK4py(CDLL('libscalapack-openmpi.so.2.0', mode=RTLD_GLOBAL))

n = 5
dtype=np.complex128
a = np.arange(n*n, dtype=dtype).reshape((n,n), order='F') * (MPI.COMM_WORLD.rank+1) if MPI.COMM_WORLD.rank==0 else None

print (a)

MP, NP = 2,1

ctx = sl.make_blacs_context(sl.get_default_system_context(), MP, NP)
descr = sl.make_blacs_desc(ctx, n, n)
print("descr", descr, descr.locrow, descr.loccol)

b = np.zeros((descr.locrow, descr.loccol), dtype=dtype)

sl.scatter_numpy(a, POINTER(c_int)(descr), b.ctypes.data_as(POINTER(c_double)), b.dtype)
print (b)


c = sl.gather_numpy(POINTER(c_int)(descr), b, (n, n))
print (c)

A streamlined interface to scalapack4py is provided by NPScal, which replicates many of the arithmatic operators implemented for ndarrays

from numpy.random import RandomState
from scalapack4py.scalapack4py import ScaLAPACK4py
from scalapack4py.npscal import NPScal
from scalapack4py.npscal import CTXT_Register, DESCR_Register, BLACSDESCRManager, BLACSContextManager
from scalapack4py.npscal.index_utils.npscal_select import diag
from scalapack4py.npscal.utils import find_squarest_grid
from mpi4py import MPI

from ctypes import RTLD_GLOBAL, POINTER, c_int, c_double, CDLL
import numpy as np
import os

sl = ScaLAPACK4py(CDLL(scalapack_libpath, mode=RTLD_GLOBAL))

n = 16
dtype=np.float64

b = np.arange(n*n, dtype=dtype).reshape((n,n))
dist_b = NPScal(gl_array=b, ctxt_tag="main", descr_tag="default", lib=sl)

b_matmul = b @ b
dist_b_matmul = dist_b @ dist_b