How to save up memory while using Multiprocessing in Python? -

-

i've got function takes node id of graph input , calculate in graph(without altering graph object), saves results on filesystem, code looks this:

... # graph file being loaded g = loadgraph(gfile='data/graph.txt') # list of nodeids being loaded nodeids = loadseeds(sfile='data/seeds.txt')  import multiprocessing mp # parallel part of code print ("entering parallel part ..") num_workers = mp.cpu_count() # 4 on machine p = mp.pool(num_workers) # _myparallelfunction(nodeid) {calculate nodeid in g , save file} p.map(_myparallelfunction, nodeids) p.close() ...

the problem when load graph python takes lots of memory(about 2g, it's big graph thousands of nodes actually), when starts go parallel part of code(the parallel map function execution) seems every process given separate copy of g , run out of memory on machine(it's got 6g ram , 3g swap), wanted see there way give each process same copy of g memory hold 1 copy of required? suggestions appreciated , in advance.

if dividing graph smaller parts not work, may able find solution using this or multiprocessing.sharedctypes, depending on kind of object graph is.


Comments

Post a Comment

Popular posts from this blog

Java 3D LWJGL collision -

-
i making 3d java game lwjgl library, , wondering how add collision detection, player not go through models. i using obj models. here objloader class, loads models: package renderengine; import java.io.bufferedreader; import java.io.file; import java.io.filenotfoundexception; import java.io.filereader; import java.util.arraylist; import java.util.list; import models.rawmodel; import org.lwjgl.util.vector.vector2f; import org.lwjgl.util.vector.vector3f; public class objloader { public static rawmodel loadobjmodel(string filename, loader loader){ filereader fr = null; try { fr = new filereader(new file("res/"+filename+".obj")); } catch (filenotfoundexception e) { system.err.println("couldn't load file!"); e.printstacktrace(); } bufferedreader reader = new bufferedreader(fr); string line; list<vector3f> vertices = new arraylist<vector3f...
Read more

spring - SubProtocolWebSocketHandler - No handlers -

-
i have ugly error during deploying spring app on jboss. 18:11:16,025 error [org.apache.catalina.core.containerbase.[jboss.web].[default-host].[/consumer]] (msc service thread 1-7) exception sending context initialized event listener instance of class org.springframework.web.context.contextloaderlistener: org.springframework.context.applicationcontextexception: failed start bean 'subprotocolwebsockethandler'; nested exception java.lang.illegalargumentexception: no handlers @ org.springframework.context.support.defaultlifecycleprocessor.dostart(defaultlifecycleprocessor.java:176) [spring-context-4.1.3.release.jar:4.1.3.release] @ org.springframework.context.support.defaultlifecycleprocessor.access$200(defaultlifecycleprocessor.java:51) [spring-context-4.1.3.release.jar:4.1.3.release] @ org.springframework.context.support.defaultlifecycleprocessor$lifecyclegroup.start(defaultlifecycleprocessor.java:346) [spring-context-4.1.3.release.jar:4.1.3.release] @ org.s...
Read more

methods - python can't use function in submodule -

-
my folder structure this: pythonstuff/ program.py modulea/ __init__.py foo.py bar.py here's code bar.py : def hello(): print("hello, world!") here's code program.py : #!/usr/bin/env python3 modulea import bar bar.hello() i run $ python3 program.py somehow error: file "program.py", line 3, in <module> bar.hello() attributeerror: 'module' object has no attribute 'hello' edit: __init__.py file empty. edit2: after trying realized had bar.py in root directory contained hello() method. bar.py in modulea/ directory empty. add it import os __all__ = [] module in os.listdir(os.path.dirname(__file__)): if module != '__init__.py' , module[-3:] == '.py': __all__.append(module[:-3]) the init .py files required make python treat directories containing packages; done prevent directories common name, such string, unintentionally hiding v...
Read more