Peter Johansson <troj...@gmail.com> writes: > Not true. I gave a recipe how to do it in my previous email.
It seems I did not express my intent clearly. I want to initialize the state from unknown data. By which I mean that I do not know if the data is suitable for initialization without modification, but I do not want a bad sequence (short period, low-dimensional correlation, etc.) if I provide bad initial data. Your recipe does not validate the input data. It does what I assume gsl_rng_fread does, except from memory instead of a file. It is not what I meant to ask for. But thank you for taking the time to answer anyway.