You need to run the test 1000 times. Like:
time1 = time()
for i in range(1000):
c_result = (a + b) * (a + b) * (a / 2.0)
time2 = time()
print(“Execution time with numpy: “, time2 – time1, “s”)

You will see OpenCL performs better than NumPy (but NumPy is still pretty good, IMHO). On my machine:

(‘Execution time of test without OpenCL: ‘, 9.549367904663086, ‘s’)
(‘Execution time with numpy: ‘, 0.01725006103515625, ‘s’)
===============================================================
(‘Platform name:’, ‘Intel(R) OpenCL’)
(‘Platform profile:’, ‘FULL_PROFILE’)
(‘Platform vendor:’, ‘Intel(R) Corporation’)
(‘Platform version:’, ‘OpenCL 1.1 LINUX’)
—————————————————————
(‘Device name:’, ‘ Intel(R) Core(TM) i7-2630QM CPU @ 2.00GHz’)
(‘Device type:’, ‘CPU’)
(‘Device memory: ‘, 5864, ‘MB’)
(‘Device max clock speed:’, 2000, ‘MHz’)
(‘Device compute units:’, 8)
Execution time of test: 0.000382 s
Results OK
===============================================================
(‘Platform name:’, ‘NVIDIA CUDA’)
(‘Platform profile:’, ‘FULL_PROFILE’)
(‘Platform vendor:’, ‘NVIDIA Corporation’)
(‘Platform version:’, ‘OpenCL 1.1 CUDA 4.2.1’)
—————————————————————
(‘Device name:’, ‘GeForce GT 540M’)
(‘Device type:’, ‘GPU’)
(‘Device memory: ‘, 2047, ‘MB’)
(‘Device max clock speed:’, 1344, ‘MHz’)
(‘Device compute units:’, 2)
Execution time of test: 0.000624512 s
Results OK

you have to compare to numpy operations, not pure python on CPU.
with numpy you can write directly
c_result = (a + b) * (a+b) * (a/2)

my results:
(‘Execution time of test without OpenCL: ‘, 11.889606952667236, ‘s’)
(‘Execution time of test within numpy ‘, 0.00010013580322265625, ‘s’)
===============================================================
(‘Platform name:’, ‘NVIDIA CUDA’)
(‘Platform profile:’, ‘FULL_PROFILE’)
(‘Platform vendor:’, ‘NVIDIA Corporation’)
(‘Platform version:’, ‘OpenCL 1.1 CUDA 4.0.1’)
—————————————————————
(‘Device name:’, ‘Quadro NVS 290’)
(‘Device type:’, ‘GPU’)
(‘Device memory: ‘, 255L, ‘MB’)
(‘Device max clock speed:’, 918, ‘MHz’)
(‘Device compute units:’, 2)
Execution time of test: 0.0360267 s
Results OK

@David – thanks for that link. I’ve been planning on a tutorial about timing and what kind of things benefit from OpenCL. I’ll cover it after I writeup my Part 2 (OpenGL particles)

I think I figured out the problem (thanks to a nice benchmark program I found online:http://git.tiker.net/pyopencl.git/blob_plain/HEAD:/examples/benchmark-all.py). Here’s the result.

Thanks again for the great work. I’m impressed at how easy it was to get this up and running. Nice job!

Cheers,

David.

(‘Execution time of test without OpenCL: ‘, 9.6666891574859619, ‘s’)
===============================================================
(‘Platform name:’, ‘Apple’)
(‘Platform profile:’, ‘FULL_PROFILE’)
(‘Platform vendor:’, ‘Apple’)
(‘Platform version:’, ‘OpenCL 1.0 (Dec 26 2010 12:52:21)’)
—————————————————————
(‘Device name:’, ‘ATI Radeon HD 6750M’)
(‘Device type:’, ‘GPU’)
(‘Device memory: ‘, 512L, ‘MB’)
(‘Device max clock speed:’, 150, ‘MHz’)
(‘Device compute units:’, 5)
Execution time of test: 0.0126587 s
Results OK
===============================================================
(‘Platform name:’, ‘Apple’)
(‘Platform profile:’, ‘FULL_PROFILE’)
(‘Platform vendor:’, ‘Apple’)
(‘Platform version:’, ‘OpenCL 1.0 (Dec 26 2010 12:52:21)’)
—————————————————————
(‘Device name:’, ‘Intel(R) Core(TM) i7-2720QM CPU @ 2.20GHz’)
(‘Device type:’, ‘CPU’)
(‘Device memory: ‘, 6144L, ‘MB’)
(‘Device max clock speed:’, 2200, ‘MHz’)
(‘Device compute units:’, 8)
Execution time of test: 0.00191466 s
Results OK

Thanks for the great tutorial. This is a great toolkit and very easy to use. The only problem I have is that using pyopencl is *slower* than numpy! I’ve been struggling for the last few hours to find a combination of data size/kernel that would show the power of OpenCL. So far no luck..

Any thoughts? (Btw I’m using pyopencl on a MacBook Pro 2.2GHz, running python2.6, and the latest version of pyopencl).

Thanks,

David.

The enthought python distro is an easy way to install a stack of packages (numpy, mathplot etc etc) no opencl in there yet.

Would love to get this working on gumstix overo and pandaboard.

Its a shame no python on iOS.

Thank you