PRNG's
category: code [glöplog]
quote Wikipedia: "If a PRNG's internal state contains n bits, its period can be no longer than 2^n results."
Internal state, what does that mean?
If I use two 8-bit registers in calculation this period 2^16 should be true, because two 8-bit registers equals 16-bits in total, even if they only perform 8-bit arithmetic...
Internal state, what does that mean?
If I use two 8-bit registers in calculation this period 2^16 should be true, because two 8-bit registers equals 16-bits in total, even if they only perform 8-bit arithmetic...
yes.
as for intitial condition the best seed is the value zero.
yea las, i was confused at first, but had to think through it.
any ideas for testing. i was thinking of doing the Monte Carlo method, but i heard its not 100% accurate. However I want to start with something easy..
What do you need your PRNG for? What do you want to test exactly?
http://en.wikipedia.org/wiki/Diehard_tests
If you need a very fast, small prng that you can also quickly reset it seeds value while it still passes the diehard tests I can warmly recommend the xor shift randomizer.
i think that is what i am using. i think its a linear congruential generator with xor bit-twiddling i am using.
thanks for the paper anyway! everything i can read about that i understand is usefull.
Probably not needed but you can simplify the xor shift method to an add/shuffle and reduce the code complexity by 5x
based on one's and zero's i made this 1D and 2D random walks. it has 2^16 period and is pretty small in size for generation.
i chopped up the image (its too damn big) you only see the first steps.
1D walk.
based on the same prng i made this walk in 2D.
2D walk, but this are based on some value from the generator using one of the variables as choosing the bits to compare. show some pretty strange output behaviour.
i chopped up the image (its too damn big) you only see the first steps.
1D walk.
based on the same prng i made this walk in 2D.
2D walk, but this are based on some value from the generator using one of the variables as choosing the bits to compare. show some pretty strange output behaviour.
forgot to mention in the first image you see every bit in the 8-bit register. they are based on the one's or zero's. i.e. 0 goes left, 1 goes right on each step or vice versa.
Another way to visualize the bit distribution is filling a bitplane.
BTW, does those result really differ much from doing?
x += x<<8 | x>>8;
BTW, does those result really differ much from doing?
x += x<<8 | x>>8;
T21: not the first and second image. the third image are a bit different when it comes to choosing what direction the particle go.
prng for color shades:
for same prng based on bitvalues:
for same prng based on bitvalues:
Just to be clear, I was mentioning
x += x<<8 | x>>8;
as the actual number generator :)
x += x<<8 | x>>8;
as the actual number generator :)
T21: please geive some output of that if you cn
I also used this in the poison & dot ball thread
2 example of bit plotted
Longer string (262k bit lenght)
oh, nice that last one
subject matter and visual stimuli highly applauded, thanks gents
Why does it look so dithered?
Graga: Because it's very likely that the function's codomain is not Lipschitz continuous?
