triqs.utility.mpi_mpi4py

Functions

all_reduce(x[, comm, op])
barrier([poll_msec])	Use asynchronous synchronization, otherwise mpi.barrier uses up all the CPU time during the run of subprocess.
bcast(x[, root])
is_master_node()
master_gets_host_names()	find the host name of all the nodes
myprint_err(x)
myprint_out(x)
recv([source])
report(*x, **opt)	print on the master, do nothing on the nodes
send(val, dest)
slice_array(A)	Given an array A, it returns a VIEW of a slice over the first dim on the node
slice_inf(imin, imax)
slice_sup(imin, imax)