for i in xrange(iter):
t1 = time.time()
gmm_cpu.fit(samples)
#a = calcA_cpu(gmm_cpu.weights_, gmm_cpu.means_, gmm_cpu.covars_)
#cl.enqueue_copy(clQueue, gmm.dA, a).wait()
#gmm.score(dSrc, dOut)
elapsed += time.time()-t1
print elapsed/iter
hOut = np.empty((hSrc.shape), np.float32)
dOut = cl.Buffer(context, cm.READ_WRITE | cm.COPY_HOST_PTR, hostbuf=hOut)
elapsed = 0
t1 = t2 = 0
for i in xrange(iter):
gmm.has_converged = False
t1 = time.time()
#gmm.fit(samples)
w,m,c = gmm.fit(dSamples, nSamples, retParams=True)
#gmm.fit(dOut, nSamples)
#gmm.score(dSrc, dOut)
elapsed += time.time()-t1
print elapsed/iter
#to estimate wmc for data already on gpu
elapsed = 0
t1 = t2 = 0
for i in xrange(iter):
t1 = time.time()
gmm_cpu.fit(samples)
