Legendre prime counting function: Difference between revisions
m
Further clarification of comments on the Legendre prime counting task...
GordonBGood (talk | contribs) (→{{header|Erlang}}: add Crystal non-memoized versions as per Nim...) |
GordonBGood (talk | contribs) m (Further clarification of comments on the Legendre prime counting task...) |
||
Line 27:
As to "the Legendre method is generally much faster than sieving up to n.", while the number of operations for the Legendre algorithm is about a factor of `log n` squared less than the number of operations for odds-only Sieve of Eratosthenes (SoE) sieving, those operations are "divide" operations which are generally much slower than the simple array access and addition operations used in sieving and the SoE can be optimized further using wheel factorization so that the required time for a given range can be less for a fully optimized SoE than for the common implementation of the Legendre algorithm; the trade off is that a fully optimized SoE is at least 500 lines of code, whereas the basic version of the Legendre prime counting algorithm is only about 40 to 50 lines of code (depending somewhat on the language used).
Also note that the Legendre prime counting function was never used practically at the time it was invented other than to demonstrate that it would find the count of primes to a trivial range only knowing the primes up to the square root of that range and there were too many operations (especially long integer division operations) to actually use it for any reasonably range even with this optimization (about 250 thousand divisions to count primes to ten million), but the follow-on work by Meissel in the 1800's definitely would have used this optimization and others in order to hand calculate the number of primes to a billion (1e9) in about ten years. Even with this optimization, Meissel would have had to hand calculate over five million divisions, so certainly used other Look Up Tables (LUT's) although certainly not caching of Phi/φ values in order to reduce the work to something possible in this amount of time. A "TinyPhi" LUT table for the first six primes of thirteen and less would have reduced the amount of work Meissel did to about 600 thousand divisions, but even that would have been perhaps too much and it is very likely that he also used "partial sieving" techniques, although that would have meant that as well as a table of the primes up to a million, he would have also needed 161 other tables of that range to a million sieved by the primes up to 13, 17, 19, to 997; however, that extra work in building these tables (which might have been done mechanically) would pay off in reducing the number of divisions to about seven thousand so the divisions become a minor problem possible to do over months and the majority of the time would be spent producing the partial sieving tables up to a million.
The reason that Meissel refined the Legendre method would have been that, even applying all of the optimizations including "partial sieving", he would still have had to do about three and a half million divisions to count the primes to a billion even if the number of primes and "partial sieve tables" only needed to be known to about 32 thousand, where his "Meissel" algorithm reduced the number of divisions to only a few thousand as per the above. Without a computer, he could never have completed the calculation of the number of primes to a billion using an optimized Legendre algorithm where he could using his modification. However, modern computers make (reasonably) quick work of integer divisions so that optimized algorithms of the Legendre type become moderately useful although at the cost of memory use as compared to Meissel type algorithms.
<br>
<br>
| |||