I love this kind of thing. I'm surprised to hear that we're doing so much work to get our random numbers; most roguelikes don't bother.
I agree that we can do better; I think we should look for something (1) faster and (2) with a smaller state. The period shouldn't matter, since although we're generating a lot of numbers, we're doing so in a context where actually USING the repeats will be completely impossible. (For this reason, it's unfortunate that we're using a fancy long-period generator, since it has all the wrong tradeoffs -- slow, big state, and low-quality numbers.) I rather admire the first RNGs the author of that paper proposes; they look to me like they meet all the requirements, and better yet they're VERY short code. I don't know if I'd care to make things any more complicated than that. Yes, we CAN throw in an encryption algorithm; but why? This isn't rocket science. (BTW, in my youth I wrote the RNG that Biskup used for ADOM (he tweaked it a little, and replaced the 'unsigned' variables with 'signed' ones, which is why that one version came out where all the rooms were dark and all the items were daggers). It was a simple RC4 stream. If you want, I can do it again -- although I wouldn't bother if I were making the decision. RC4 has very simple code and a simple state, only 256 permuted bytes and 2 index bytes; but those RNGs are far simpler.) -Wm ------------------------------------------------------------------------------ Create and publish websites with WebMatrix Use the most popular FREE web apps or write code yourself; WebMatrix provides all the features you need to develop and publish your website. http://p.sf.net/sfu/ms-webmatrix-sf _______________________________________________ Crawl-ref-discuss mailing list Crawl-ref-discuss@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/crawl-ref-discuss
