py-libnuma
py-libnuma is python3 interface to numa Linux library so that you can set task affinity and memory affinity in python level for your process which can help you to improve your code's performence.
Installation
pip install --upgrade py-libnuma
Usage
py-libnuma categorize libnuma's apis into 3 groups :memory, schedule and info. You can set your tasks' cpu affinity, memory affinity and get information about your systems's hardware with these 3 modules respectively. For more information about APIs, you can refer to API.md in github
schedule
numa.schedule helps you to set cpu affinity for your process, if you have multiple numa nodes, and you want your process to be scheduled on cpus from node1 and node2, you can use numa.schedule like this
from numa import schedule
schedule.run_on_nodes(1,2)
If you want a certain process with pid to run on specific cpus, you can use numa.schedule like this
from numa import schedule
schedule.run_on_cpus(pid, 1,3,4,6)
memory
numa.memory helps you to set memory policy for your process, if you want your current process to allocate memory from multiple numa nodes to balance local and remote memory access, you can use numa.memory like this:
from numa import memory
memory.set_interleave_nodes(0,1)
or you can make your process to allocate from certain nodes by
from numa import memory
memory.set_membind_nodes(1)
Info
numa.info helps you to get information about your numa hardware, you can check hardware information by:
from numa import info
info.numa_hardware_info()
This will tell you distance between different numa nodes and node-cpu relation. You can also check cpu set for certain nodes by:
from numa import info
info.node_to_cpus(1)
or check which node is a certain cpu belongs to by:
from numa import info
info.cpu_to_node(1)
For more information about APIs, you can refer to API.md in github
Feedback
If you have any problems with using this package, feel free to create issues and you will get answered in no more than 24 hours
2 Nov 24, 2022
39 Nov 20, 2022
1 May 10, 2022
5 Jun 27, 2022
1 Jan 1, 2022
8 Dec 2, 2022
38 Dec 29, 2022
90 Dec 16, 2022
70 Dec 20, 2022
5 Apr 11, 2022
2 Jul 30, 2022
1 Nov 26, 2021
34 Nov 28, 2022
33 Aug 20, 2022
1 Nov 24, 2021
2 Oct 22, 2022
7 Apr 17, 2022
5 Nov 10, 2021
27.5k Jan 04, 2023
8 Nov 17, 2022
1 Jan 20, 2022
1 Dec 23, 2021
9 Jan 07, 2023
195 Dec 13, 2022
3 Jun 07, 2022
8 Nov 07, 2021
180 Oct 08, 2022
2 Oct 10, 2022