Anonymous user
Talk:Summarize and say sequence: Difference between revisions
→Caching
Line 34:
And set up a ring buffer of the last three terms as a cache and also did not compute the series for starting values that had permuted digits of a starting number that had come before.<br>
I have no version without those possible speed-ups, but the code ran in a minute or two for a million starting numbers and maybe Perl would run in an acceptable time as, in general, Perl 5 and Python have similar run times for similar algorithms I find. --[[User:Paddy3118|Paddy3118]] 05:06, 25 August 2011 (UTC)
:: The perl code already caches everything, so the benefit of a last-three buffer is not obvious. As for doing permutations of seeds, most numbers run into cached sequences pretty fast anyway, so that's not going to help much, either. Calculating only one of a permutation group runs another risk: if there's a sequence A->B->B where A and B are permutations, the result for seed A and seed B may be inconsistent. This doesn't happen for base 10 below 1000000, though (last three numbers of sequence from seed 0 is such an example, which is what, 16 digits long?).
:: To put things into perspective, the perl code runs one million in about a minute, but my attention span is only 10 seconds or so, so it felt too long. I really wanted to be able to cache only sorted strings which would give about 10 times speed up and reduce memory usage, but dealing with the A->B->B stuff above was a real dog. --[[User:Ledrug|Ledrug]] 07:30, 25 August 2011 (UTC)
| |||