[Numpy-discussion] Purpose for bit-wise and'ing the initial mersenne twister key?

Sturla Molden sturla@molden...
Fri Feb 13 09:25:37 CST 2009


On 2/13/2009 4:07 PM, Michael S. Gilbert wrote:

> I'm not "obsessed" with speed, but a 2x improvement is quite
> significant.

Honestly, I don't care about 2x differences here. How many milliseconds 
do you save?

The PRNG in SciPy is fast enough for 99% of any use I can conceive. I 
have yet to see a program speed-limited by the PRNG. Projects that are 
speed limited by the PRNG could take the hassle to compile their own. 
And if speed is that important, perhaps one should not use the Mersenne 
Twister at all?

http://groups.google.com/group/comp.lang.c/msg/e3c4ea1169e463ae?dmode=source
http://groups.google.com/group/sci.crypt/msg/12152f657a3bb219?dmode=source



S.M.





More information about the Numpy-discussion mailing list