Hi Thomas,
I think we can fix the time needed to slice a model.
Actually, its simple to parallize the task over a couple of processors.
25.6454157829 necessary to perform slice with a 1 processor(s).
5.20338606834 necessary to perform slice with a 8 processor(s).
Pretty good, right, that's a nice 5 fold speed up ( could be faster
async ).
Still the example is pretty trivial. I haven't figured out how to
memory map / share a shape between processes, so I'm recreating a
sphere for every slice ( duhhh... )
So, a bit of an oversimplified example, but proves the point of how
simple it really is to do this kind of thing.
Oh, its still 33 lines ;')
Cheers,
-jelle
from OCC.gp import *
from OCC.BRepPrimAPI import *
from OCC.BRepAlgoAPI import *
from OCC.TopOpeBRepTool import *
from OCC.BRepBuilderAPI import *
from OCC.Geom import *
from OCC.TopoDS import *
import time, processing, numpy
def slice( z ):
import os
# Create Plane defined by a point and the perpendicular direction
print 'slicing index:', z, 'sliced by process:', os.getpid()
shape = BRepPrimAPI_MakeSphere ( 60. ).Shape() # Create the shape
to slice
plane = gp_Pln( gp_Pnt( 0., 0., z ), gp_Dir( 0., 0., 1. ) )
face = BRepBuilderAPI_MakeFace( plane ).Shape()
# Computes Shape/Plane intersection
section = BRepAlgoAPI_Section( shape, face )
if section.IsDone():
return section
else:
return None
def run( n_procs ):
Zmin, Zmax, deltaZ = -100., 100., 0.001
#shape = BRepPrimAPI_MakeSphere (60.).Shape() # Create the shape
to slice
P = processing.Pool( n_procs )
init_time = time.time() # for total time computation
result = P.map( slice, numpy.arange( Zmin, Zmax + 0.01,
deltaZ ).tolist() )
total_time = time.time() - init_time
print "%s necessary to perform slice with a %s processor(s)." %
( total_time, n_procs )
time.sleep( 5 )
run( 1 ); run( 8 )
_______________________________________________
Pythonocc-users mailing list
Pythonocc-users@gna.org
https://mail.gna.org/listinfo/pythonocc-users
