A wrapper around mpi4py that offers simple scattering of iterable objects.
Project description
A wrapper around mpi4py that offers simple scattering of iterable objects.
This is useful for embarassingly parallel, SPMD, type tasks that simply need to work on a list of things.
example usage:
parfor_test.py
# import the parfor function; note
# that this will automatically initialize MPI
# also import pprint for parallel-friendly printing
from simplempi.parfor import parfor, pprint
# define a list to loop over
my_list = list(range(10))
# define a function that does something with each item in my_list
def func(i):
return i**2
# loop in parallel over my_list
for i in parfor(my_list):
result = func(i)
pprint(f"{i}**2 = {result}")
Running this with mpirun on 4 processors shows that the list of 10 numbers gets scattered as evenly as possible across all 4 processors; it also shows that the order of evaluation in the for loop is not well-defined (which is okay for embarassingly parallel code like this):
$ mpirun -n 4 python parfor_test.py
(rank 1/4): 0**2 = 0
(rank 1/4): 4**2 = 16
(rank 1/4): 8**2 = 64
(rank 3/4): 2**2 = 4
(rank 3/4): 6**2 = 36
(rank 4/4): 3**2 = 9
(rank 4/4): 7**2 = 49
(rank 2/4): 1**2 = 1
(rank 2/4): 5**2 = 25
(rank 2/4): 9**2 = 81
Alternatively, one can use the object-oriented interface:
simpleMPI_test.py
import simplempi
#Initialize MPI
smpi = simplempi.simpleMPI()
#Make a list of things (20 numbers in this case)
testList = range(20)
#Scatter the list to all processors (myList differs among processes now)
myList = smpi.scatterList(testList)
#Print the list contents (as well as the rank of the printing process)
smpi.pprint(myList)
Running this with mpirun on 6 processors shows that the list of 20 numbers gets scattered as evenly as possible across all 6 processors:
$ mpirun -n 6 python simpleMPI_test.py
(rank 1/6): [0, 6, 12, 18]
(rank 2/6): [1, 7, 13, 19]
(rank 4/6): [3, 9, 15]
(rank 6/6): [5, 11, 17]
(rank 5/6): [4, 10, 16]
(rank 3/6): [2, 8, 14]
Install
python3 -m pip install simplempi
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
File details
Details for the file simplempi-0.1.5.tar.gz
.
File metadata
- Download URL: simplempi-0.1.5.tar.gz
- Upload date:
- Size: 5.2 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/5.1.1 CPython/3.8.18
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | f6632b19243938d847e4e0c2a68c9970cee32149c872c691c7e9b78736c4a9e8 |
|
MD5 | 073e27bd0a968fe1597e2d50609269d2 |
|
BLAKE2b-256 | 86723ff6782c15ed24e9b990edd4bcb5824fe4ed153cb9c0968072b42b984134 |
File details
Details for the file simplempi-0.1.5-py3-none-any.whl
.
File metadata
- Download URL: simplempi-0.1.5-py3-none-any.whl
- Upload date:
- Size: 5.6 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/5.1.1 CPython/3.8.18
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 4f9c112e9aac4f60d292c1080a344129f2679852f5ec6977b745d9ffaf68e7ae |
|
MD5 | c3636cb6535d07cb6aaa641d9a6c56c5 |
|
BLAKE2b-256 | 86e5f98e465d14904bf5a15b26572f7c1121a508d42836e7b97333f1f21a1b57 |