Messages in this thread | | | Date | Wed, 13 Aug 2003 19:43:19 +0100 | From | Jamie Lokier <> | Subject | Re: i810_rng.o on various Dell models |
| |
Jeff Garzik wrote: > Intel's RNG was slow anyway, compared to the AMG and now VIA RNGs, so > you didn't want it anyway :) I really like VIA's "xstore", which is an > RNG implemented via a CPU instruction. That way you don't need a kernel > driver at all, really.
From an electronic point of view, assuming similar technology, and circuits of similar size, I would expect a slower source to be able to give getter randomness than a very fast source.
Whether it does depends on the implementation, of course.
I don't know if "xstore" can execute on every clock cycle. But imagine if it could execute once every 1GHz cycle, and if there was only one RNG circuit, not lots of them in parallel. Then it would seem difficult indeed to ensure that successive outputs of the RNG were strongly independent. Whereas the same circuit, integrating over a signal 10^6 times to return a value every 1ms, would return better results but fewer of them.
That said, it might be better to read lots of results from "xstore" and feed them into a known algorithm like random.c, than to read just a few results from a slow Intel black box and trust the quality of its source.
-- Jamie - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |