added fall-back to non-multithreading execution when using only single CPU.

This commit is contained in:
Philip Eisenlohr 2015-08-08 21:43:21 +00:00
parent 31f6d22d89
commit 57dfa6a87e
1 changed files with 10 additions and 9 deletions

View File

@ -81,15 +81,16 @@ def laguerreTessellation(undeformed, coords, weights, grains, nonperiodic = Fals
arguments = [[arg] + [seeds,squaredweights] for arg in list(undeformed)]
# Initialize workers
pool = multiprocessing.Pool(processes = cpus)
# Evaluate function
result = pool.map_async(findClosestSeed, arguments)
pool.close()
pool.join()
closestSeeds = np.array(result.get()).flatten()
if cpus > 1: # use multithreading
pool = multiprocessing.Pool(processes = cpus) # initialize workers
result = pool.map_async(findClosestSeed, arguments) # evaluate function in parallel
pool.close()
pool.join()
closestSeeds = np.array(result.get()).flatten()
else:
closestSeeds = np.zeros(len(arguments),dtype='i')
for i,arg in enumerate(arguments):
closestSeeds[i] = findClosestSeed(arg)
return grains[closestSeeds%coords.shape[0]] # closestSeed is modulo number of original seed points (i.e. excluding periodic copies)