Never been particularly proficient with pack, but it looks like your approach falls apart with the restricted character set. OTOH, that's an impressive result for [20000,26,20]. I imagine optimizing your unpack could yield some impressive optimization.20000 26 20 Rate hdb RobertCraven kennethk xx +x hdb 2.57/s -- -8% -55% -84 +% RobertCraven 2.81/s 9% -- -51% -82 +% kennethk 5.77/s 124% 106% -- -64 +% xxx 15.8/s 515% 464% 174% - +- 20000 4 20 Rate hdb RobertCraven xxx kenneth +k hdb 2.82/s -- -11% -46% -54 +% RobertCraven 3.18/s 13% -- -40% -48 +% xxx 5.28/s 87% 66% -- -13 +% kennethk 6.08/s 115% 91% 15% - +- 200000 26 5 s/iter kennethk xxx kennethk 2.05 -- -32% xxx 1.40 47% -- 200000 4 5 s/iter xxx kennethk xxx 24.6 -- -92% kennethk 2.07 1090% --
And, FYI, the windowing approach is about 30x slower than other options with only 1 letter.
#11929 First ask yourself `How would I do this without a computer?' Then have the computer do it the same way.
In reply to Re^3: suffix array efficiency
by kennethk
in thread suffix array efficiency
by RobertCraven
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |