I want to use the Mersenne Twister implementation from the CUDA SDK.
All I want is to call a kernel which creates a requested amouth (e.g. 20MB) of random
numbers and put it into the device memory.
Taking a look at the Mersenne Twister from the SDK is confusing me…
Is there no simple function (kernel) which I can call to get random numbers? :wacko:
The Mersenne Twister from CUDA SDK isn’t that hard to understand and use. But for now it confuses me too :) I’m thinking about changing it so that it can be used in each block independently. Is there someone more advanced who could be so kind to explain what are N_PER_RNG, RAND_N, PATH_N, MT_RNG_COUNT? This would be helpful in understanding the algorithm. External Media