Talk:Rare numbers: Difference between revisions
→30 mins not 30 years
m (→30 mins not 30 years: fixed typo in the data (04+40 --> 44), corrected a misspelling.) |
|||
Line 58:
:: Rubbish or not, is there anything on the referenced (Gupta's) website that is incorrect? The properties and observations is what the REXX solution used (and others have as well) to calculate ''rare'' numbers, albeit not as fast as your algorithm. I have no idea how long Shyam Sunder Gupta's program(s) executed before it found eight rare numbers (or how much virtual memory it needed). Is the '''F#''' algorithm suitable in finding larger ''rare'' numbers? I suspect (not knowing '''F#''') that virtual memory may become a limitation. Eight down, seventy-six more to go. -- [[User:Gerard Schildberger|Gerard Schildberger]] ([[User talk:Gerard Schildberger|talk]]) 18:18, 18 September 2019 (UTC)
:::So is the task now to replicate [[http://www.shyamsundergupta.com/R20.htm]]? This may not be reasonable for a RC task as I now discuss.<br> Why have you "no idea how long Shyam Sunder Gupta's program(s) executed before it found eight rare numbers"? On the webpage it says "the program has been made so powerful that all numbers up to 10^14 can be just checked for Rare numbers in less than a minute". This implies that it can search 10^13 in 5 secs. I believe this. The problem with the webpage is not that it is wrong, but that it is disingenuous. I estimate that the following is achievable:<br>10^13 -> 5 secs;<br>10^15 -> 60 secs;<br>10^17 -> 20 mins;<br>10^19 -> 7 hours;<br>10^21 -> 6 days;<br>10^23 -> 4 months.<br>I would say that 10^17 is reasonable for a RC task and is in line with the timings given on the webpage. Those who have recently obtained an 8th. generation i7 might want to observe that there is an obvious multithreading strategy and might want to prove that Goroutines are more than a bad pun on coroutines, as a suitable punishment for making me envious. (Warning, from my experience of i7s it might be wise to take it back to the shop and have water cooling installed before attempting to run it for a day and a half on full throttle). I remain to be convinced that these benchmarks are achieved by the Fortran, Ubasic programs on the webpage, or can be achieved in this task using the methods described on the webpage.<br>It is necessary to distinguish the algorithm I describe above from the F# implementation on the task page. The algorithm can be written to require very little memory. Obviously the F# as it stands calculates all candidates before checking them and this list grows with increasing number of digits. It is more than adequate for the current task, and I anticipate little difficulty in accommodating reasonable changes to make the task less trivial as layed out above--[[User:Nigel Galloway|Nigel Galloway]] ([[User talk:Nigel Galloway|talk]]) 14:01, 20 September 2019 (UTC)
== the 1<sup>st</sup> REXX version ==
| |||