One last thing before I go to bed for the night:

I did some brief benchmarks using the following function:

>>> def dobench(size):
...     v = p.Vector(size, 0.1)
...     w = p.Vector(size, 0.2)
...     p.Statement(v+w).execute()
...     p.backend_finish()
...     t1 = time.time()
...     for n in range(10000):
...             p.Statement(v+w).execute()
...     p.backend_finish()
...     t2 = time.time()
...     return (t2 - t1)

(where I have "import pyviennacl as p"). On my GeForce 610M, I get the
following results:

dobench(2**1)    0.6708228588104248
dobench(2**2)    0.6711068153381348
dobench(2**3)    0.6693098545074463
dobench(2**4)    0.6727585792541504
dobench(2**5)    0.6788749694824219
dobench(2**6)    0.6779205799102783
dobench(2**7)    0.6860549449920654
dobench(2**8)    0.7044177055358887
dobench(2**9)    0.7262992858886719
dobench(2**10)   0.7909643650054932
dobench(2**11)   0.9179358482360840
dobench(2**12)   1.2310407161712646
dobench(2**13)   1.7771134376525880
dobench(2**14)   2.2906632423400880
dobench(2**15)   3.8429172039031982
dobench(2**16)   7.0529527664184570
dobench(2**17)  15.2666733264923100
dobench(2**18)  28.5626835823059080
dobench(2**19)  54.6134884357452400

Units are time in seconds (for 10,000 statement constructions and
executions). If I read my old graph (from my proposal[1]) correctly,
this is about the same shape and absolute performance that I got with my
"naive" implementation, which suggests that the limit here is my
hardware, not my code, which is good. Of course, I haven't tested more
complex expressions yet.

[1]
https://github.com/tsmithe/viennacl-dev/blob/pyviennacl/GSoC/pyviennacl-puzzle.pdf


Best,

Toby


------------------------------------------------------------------------------
See everything from the browser to the database with AppDynamics
Get end-to-end visibility with application monitoring from AppDynamics
Isolate bottlenecks and diagnose root cause in seconds.
Start your free trial of AppDynamics Pro today!
http://pubads.g.doubleclick.net/gampad/clk?id=48808831&iu=/4140/ostg.clktrk
_______________________________________________
ViennaCL-devel mailing list
ViennaCL-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/viennacl-devel

Reply via email to