Timeline for Random number generator that returns unique 64-bit numbers in sorted order
Current License: CC BY-SA 3.0
19 events
when toggle format | what | by | license | comment | |
---|---|---|---|---|---|
Nov 20, 2023 at 8:28 | vote | accept | Thomas Mueller | ||
Jul 31, 2017 at 8:19 | history | tweeted | twitter.com/StackStats/status/891936451365138432 | ||
Jul 11, 2017 at 22:18 | answer | added | whuber♦ | timeline score: 3 | |
Jul 9, 2017 at 9:26 | history | edited | Thomas Mueller | CC BY-SA 3.0 |
added 3260 characters in body
|
Jul 7, 2017 at 19:31 | history | edited | Thomas Mueller | CC BY-SA 3.0 |
added 290 characters in body
|
Jul 7, 2017 at 19:27 | comment | added | Thomas Mueller | @Kodiologist do you mean sort on disk? That many numbers is 8 TB. First, I don't have such a disk (well, it can be done on Amazon EC2 I know). Second, how to ensure there are no duplicates (you can't simply re-run the test; the probability of duplicates is simply too large. You would have to generate more numbers, re-test,...). And it takes a long time. 2 hours just to read the keys, without sorting. So, no, I wouldn't want to sort. | |
Jul 7, 2017 at 19:20 | comment | added | Thomas Mueller | @whuber I need exactly $10^{12}$. This is to replicate a test they did for BBHash where they generated exactly that many numbers. In their code, the numbers are not random, so it's not a realistic test. | |
Jul 7, 2017 at 19:08 | comment | added | Kodiologist | My guess is that the fastest way to do this is to sort the random variates in parallel, as they're produced. | |
Jul 7, 2017 at 18:56 | comment | added | whuber♦ | Do you need to generate exactly $10^{12}$ values or do you only need to select each value independently with probability $10^{12}/2^{64}$? | |
Jul 7, 2017 at 16:36 | history | edited | Thomas Mueller | CC BY-SA 3.0 |
added 330 characters in body
|
Jul 7, 2017 at 15:33 | answer | added | Craig | timeline score: 0 | |
Jul 7, 2017 at 14:39 | history | reopened | whuber♦ | ||
Jul 7, 2017 at 14:29 | history | edited | Thomas Mueller | CC BY-SA 3.0 |
added 526 characters in body; edited title
|
Jul 7, 2017 at 13:38 | history | closed | whuber♦ | Needs details or clarity | |
Jul 7, 2017 at 13:38 | comment | added | whuber♦ | The nature of this question is obscure. Much of it seems to focus on dealing with high-precision arithmetic. Some of it refers to "blocks," which meaning is undefined; and none of it explains what properties these "random numbers" need to have: should they be independent? Uniform? Constrained within a predefined interval? What might "use the probability that the current value matches" mean? It sounds like some aspects of this situation are interesting, but if you would like an answer than please make edits that clarify what you need. | |
Jul 7, 2017 at 13:10 | history | edited | Thomas Mueller | CC BY-SA 3.0 |
added 215 characters in body
|
Jul 7, 2017 at 9:15 | history | edited | Thomas Mueller | CC BY-SA 3.0 |
deleted 4 characters in body
|
Jul 7, 2017 at 9:04 | review | First posts | |||
Jul 7, 2017 at 9:12 | |||||
Jul 7, 2017 at 9:02 | history | asked | Thomas Mueller | CC BY-SA 3.0 |