Parsl is a flexible and scalable parallel programming library for Python which is being developed at the University of Chicago. It augments Python with simple constructs for encoding parallelism. For more information about Parsl, please refer to its documentation.


Parsl can be installed with Conda for use on Summit by running the following from a login node:

$ module load python/3.6.6-anaconda3-5.3.0
$ conda create --name parsl-py36 python=3.6
$ source activate parsl-py36
$ conda config --add channels conda-forge
$ conda install parsl

Hello world!

The following instructions illustrate how to run a “Hello world” program with Parsl on Summit.

Parsl needs to be able to write to the working directory from compute nodes, so we will work from within the member work directory and assume a project ID ABC123:

$ mkdir -p ${MEMBERWORK}/abc123/parsl-demo/
$ cd ${MEMBERWORK}/abc123/parsl-demo/

To run an example “Hello world” program with Parsl on Summit, create a file called with the following contents, but with your own project ID in the line specified:

from parsl.addresses import address_by_interface
from parsl.config import Config
from parsl.executors import HighThroughputExecutor
from parsl.launchers import JsrunLauncher
from parsl.providers import LSFProvider

from parsl import python_app

import parsl


config = Config(
    executors = [
            label = 'Summit_HTEX',
            address = address_by_interface('ib0'),
            worker_port_range = (50000, 55000),
            provider = LSFProvider(
                launcher = JsrunLauncher(),
                walltime = '00:10:00',
                nodes_per_block = 1,
                init_blocks = 1,
                max_blocks = 1,
                worker_init = 'source activate parsl-py36',
                project = 'abc123', # replace this line
                cmd_timeout = 30

def hello ():
    import platform
    return 'Hello from {}'.format(platform.uname())


Now, run the program from a shell or script:

$ python3

There will be a flood of output to stdout, but the lines that indicate successful execution will look something like the following:

2021-06-28 16:10:46 parsl.dataflow.dflow:431 [INFO]  Task 0 completed (launched -> exec_done)
Hello from uname_result(system='Linux', node='a01n14', release='4.14.0-115.21.2.el7a.ppc64le', version='#1 SMP Thu May 7 22:22:31 UTC 2020', machine='ppc64le', processor='ppc64le')

Congratulations! You have now run a Parsl job on Summit.