Hi. To find the answer you can see “Parallel execution in Abaqus/Standard,” Section 3.5.2 of the Abaqus Analysis User's Manual.

Hi Danniel,

I ran into similar situation recently and based on my searching, standard do not seem to provide the opportunity to specify domain partitions as Explicit does. And it also seems the parallel in Standard is only for the linear solver. Do you find a solution to this? If so, could you share an update here, which I believe would be very beneficial to other researchers and me as well. Thank you!

Best,

Xiang

Hi Xiang,

I did not find a method of selecting or modifying the parallel domains in Standard. One of my lab mates tried tricking the selection by changing things such as element numbering, and adding various constraints and connections in place of the normal mesh connectivity. In the end there doesn't seem to be a way to control parallel domains.

We use standard in parallel with our crystal plasticity UMATs. I believe everything is nonlinear. Because of the way we write our UMATs we treat each CPU core as a MPI host. You could use something like the following python code in your environment file to achieve this:

import os, socket, time

os.environ['ABA_BATCH_OVERRIDE'] = '1'

os.environ['ABA_MPI_SKIP_BUNCH_NODES'] = "1"

cpus = globals().get('cpus', locals().get('cpus', 1))

memory=str(90/cpus) + '%'

mp_host_list = []

for i in range(cpus):

mp_host_list.append([socket.gethostname(), 1])

del i

Best,

Dan