def global_iterating(arr, iters=10000):
size = arr.globalsize
time = measure_time()
for _ in range(iters):
for index in range(size):
arr[index]
time = measure_time() - time
return time/iters
def local_setting(arr, iters=10000, comm=MPI.COMM_WORLD):
comm.Barrier()
time = measure_time()
for _ in range(iters):
arr.local[0] = 1.0
time = measure_time() - time
comm.reduce(time, op=MPI.MAX, root=0)
return time / iters
def global_iterating(arr, iters=10000, comm=MPI.COMM_WORLD):
size = arr.globalsize
comm.Barrier()
time = measure_time()
for _ in range(iters):
for index in range(size):
arr[index]
time = measure_time() - time
comm.reduce(time, op=MPI.MAX, root=0)
return time/iters
def local_setting(arr, iters=10000):
time = measure_time()
for _ in range(iters):
arr.local[0] = 1.0
time = measure_time() - time
return time / iters
def local_slicing(arr, iters=10000):
time = measure_time()
for _ in range(iters):
arr.local[::2]
time = measure_time() - time
return time / iters
import numpy as np
import scipy.cluster.vq as cluster
from operations import gen_blobs, measure_time
if __name__ == '__main__':
t_time = measure_time()
print("Type,Obs,Feats,Clusters,Time")
for obs_power in range(4, 11):
num_obs = 2**obs_power
for feats_power in range(1, 11):
for clust_power in range(1, obs_power):
features = 2**feats_power
k = 2**clust_power
seed = np.zeros((k, features))
observations, labels = gen_blobs(num_obs, features, k)
for _ in range(5):
time = measure_time()
centroids, labels = cluster.kmeans2(observations,
k,
minit='points')
time = measure_time() - time
print("scipy,%d,%d,%d,%.9f" % (num_obs, features, k, time))
total_time = measure_time() - t_time
print(total_time)
from mpi4py import MPI
import mpids.MPInumpy as mpi_np
import mpids.MPIscipy.cluster as mpi_cluster
from operations import gen_blobs, measure_time
if __name__ == '__main__':
comm = MPI.COMM_WORLD
rank = comm.Get_rank()
n_procs = comm.Get_size()
runs = int(sys.argv[1])
obs_power = int(sys.argv[2])
local_size = 2**obs_power
num_obs = n_procs * local_size
k = 2
features = 2
observations, labels = gen_blobs(num_obs, features, k)
mpi_obs = mpi_np.array(observations, dist='b', dtype=np.float64)
for _ in range(runs):
comm.Barrier()
time = measure_time()
centroids, labels = mpi_cluster.kmeans(mpi_obs, k)
time = measure_time() - time
comm.reduce(time, op=MPI.MAX, root=0)
if rank == 0:
print("mpi_scipy,%d,%d,%d,%d,%.9f" %
(n_procs, local_size, features, k, time))
del centroids, labels
