Sieve of Eratosthenes: Difference between revisions

Sets h$ to 1 for higher multiples of 2 via <code>FILL$</code>, later on augments each <code>STEP</code> accordingly; replaces Floating Pt array p(z) with string variable h$(z) to sieve out all primes < z=441 (<code>l</code>=21) in under 1K, so that h$ is fillable to its maximum (32766), even on a 48K ZX Spectrum if translated back.

40 FOR n=3 TO l STEP 2

While slower than the optimised Sieve of Eratosthenes before it, the Sieve of Sundaram above has a compatible compression scheme that's more convenient than the conventional one used beforehand. It is therefore applied below along with Euler's alternative optimisation in a reversed implementation that lacks backward-compatibility to ZX80 BASIC. Although this program uses features & limits applicable to the QL, it can be rewritten more efficiently for 1K ZX80s, as they allow integer variables to be parameters of <code>FOR</code> statements (& as their 1K of static RAM is equivalent to L1 cache, even in <code>FAST</code> mode). That's left as an exercise for ZX80 enthusiasts, who for o%=14 should be able to generate all primes < 841, i.e. 3 orders of (base 2) magnitude more than found by the program listed under Sinclair ZX81 BASIC. In QL SuperBASIC, o% may at most be 127--generating all primes < 65,025 (almost 2x the upper limit for indices & integer variables used to calculate them ~2x faster than for floating point as used in line 30).

10 INPUT "Enter highest value of on-diagonal index: ";o%