import pycuda.autoinit
import pycuda.driver as drv
from pycuda import gpuarray
from pycuda import cumath
from pycuda.elementwise import ElementwiseKernel
import numpy as np

start = drv.Event()
end = drv.Event()

N = 10**6

a = 2*np.ones(N,dtype=np.float64)

start.record()
np.exp(a)
end.record()
end.synchronize()
secs = start.time_till(end)*1e-3
print "Numpy",secs

a_gpu = gpuarray.to_gpu(a)
b_gpu = gpuarray.zeros_like(a_gpu)

kernel = ElementwiseKernel(
   "double *a,double *b",
   "b[i] = exp(a[i]);",
    "kernel")

start.record() # start timing
kernel(a_gpu,b_gpu)
end.record() # end timing
end.synchronize()
secs = start.time_till(end)*1e-3
print "Kernel",secs

start.record()
cumath.exp(a_gpu)
end.record()
end.synchronize()
secs = start.time_till(end)*1e-3
print "Cumath", secs

Numpy    0.0218774719238
Kernel   0.143336898804
Cumath   0.14694732666

Numpy    0.0205841598511
Kernel   0.138252410889
Cumath   0.00203606390953