optimized prime factor finder for 12c+ - Printable Version +- HP Forums (https://archived.hpcalc.org/museumforum) +-- Forum: HP Museum Forums (https://archived.hpcalc.org/museumforum/forum-1.html) +--- Forum: Old HP Forum Archives (https://archived.hpcalc.org/museumforum/forum-2.html) +--- Thread: optimized prime factor finder for 12c+ (/thread-172173.html) optimized prime factor finder for 12c+ - Don Shepherd - 09-20-2010 Well, Dave Britten started this ball rolling here with a prime factor finder for the 32sii and 20s based upon an original program for the 67. Tim Wessman used this as a model for the 30b in message number 30 of that thread. I made a few modifications here for the 32s, a very nice machine. But I always wondered if this basic algorithm could be done on the 12c, and especially the fast 12c+. At first glance, it would appear unlikely because of the heavy reliance on subroutines, which the 12c does not have. But the more I looked at the code, the more I believed that it should be possible to somehow implement this algorithm on the 12c+. Yesterday I figured out how to do it, essentially using indirect addressing via the Rcl CFj command with the cash flow registers. The code is listed below. The 12c+ is rather slower than the 30b for this application: the 30b determines the primality of 300,000,007 in 9 seconds versus the 12c+ 38 seconds. But this algorithm represents a significant improvement over a brute-force algorithm that eliminates only multiples of 2 from the trial factor pool; that version takes 57 seconds on the 12c+. The fact that an algorithm that is so subroutine-intensive can be done on the 12c+ at all is a testament to the greatness of the design of the 12c. No wonder it is the most successful calculator HP has ever produced. Edited on 9/20/2010 to implement Katie Wasserman's suggestion to use Nj in addition to CFj so that no preloading of registers is necessary, a brilliant suggestion. ```Prime factors program for 12c+ Eliminates multiples of 2, 3, and 5 from trial factor pool These are the trial factor increment values that are stored via cfj and nj in R2 to R7: 6,2,6,4,2,4,2,4,2,2,1,2 Enter number to factor, R/S R/S after each factor displayed 0 indicates done maximum number to factor = 999,999,999 Register usage: R2 - R7 (and corresponding nj) = trial factor increments, begins at R7 R0 - number to factor R1 - current trial factor n - used to control indirect addressing i - flag for returning to the right location from the routine at line 56 Mem command = p-71 r-11 01 sto 0 number to factor 02 clr sigma clears R1 for trial factor use 03 2 load trial factor increment values into cfj/nj 04 sto n registers R2 to R7 05 6 06 cfj 07 2 08 nj 09 6 10 cfj 11 4 12 nj 13 2 14 cfj 15 4 16 nj 17 2 18 cfj 19 4 20 nj 21 2 22 cfj 23 nj 24 1 25 cfj 26 2 27 nj 28 0 loop begins here to get next series of trial factor increment values 29 sto i flag so you return to the correct line number from routine at line 56 30 rcl nj get the next trial factor increment value 31 goto 56 check to see if this is a factor 32 1 33 sto i flag to return to the correct line number from line 56 routine 34 rcl cfj get the next trial factor increment value 35 goto 56 check to see if this is a factor 36 rcl n when n gets to 2, an iteration of loop is done 37 3 38 x<=y 39 goto 28 loop not done, so continue loop with next increment 40 rcl 0 if num to factor is 1, you are done factoring 41 ln will be 0 if num to factor is 1 42 x=0 43 goto 00 display 0 and exit program 44 6 reset indirect pointer to 6 for second 45 sto n and all subsequent loop iterations 46 rcl 0 number to factor 47 rcl 1 current trial factor 48 enter 49 x no x-squared key on 12c so this does it 50 x<=y continue loop until you get to 51 goto 28 the square root of the number to factor 52 rcl 0 the final factor is in R0 now so display 53 R/S it, then display 0 and stop 54 0 55 goto 00 56 sto+1 beginning of routine to see if this is a factor, increment trial divisor 57 rcl 0 current number to factor 58 rcl 1 current trial divisor 59 / 60 frac frac part will be 0 if R1 is a factor of R0 61 x=0 62 goto 67 if a factor 63 rcl i 0=return to line 32, 1=return to line 36 64 x=0 65 goto 32 66 goto 36 67 rcl 1 factor found, so update number to factor 68 sto/0 update number to factor 69 r/s display factor wait for user to press r/s 70 0 don't increment trial factor, might be same factor again 71 goto 56 see if current trial factor is again a factor ``` Edited: 21 Sept 2010, 1:00 p.m. Re: optimized prime factor finder for 12c+ - Dave Britten - 09-22-2010 Very cool. I shied away from doing it on a 12c because of the lack of subroutines. Getting it running on the algebraic 20s was enough of a challenge. :) I might have to try this on my standard 12c and see how much more slowly it runs. Re: optimized prime factor finder for 12c+ - Don Shepherd - 09-22-2010 Yeah, given the extensive use of subroutines in the algorithm, for a long time I thought it was just not possible to consider it on the 12c. But when I thought about the indirect addressing using Cfj, I thought there might be a way, and I found one! But it originally required loading all 12 of the trial factor increment values prior to running the program, which was definitely not cool. Then Katie said that if you use Nj in addition to Cfj you would need half as many registers and the extra program space that would free up would probably be adequate to initialize the registers you do need, and she was right. What a brilliant suggestion. Katie also had the idea of stuffing all 12 increment values into one register using the 10-digit number and 2-digit exponent. That would free up all kinds of registers for program lines, but the mechanics of decoding that single register each time, including the exponent, would probably impact the execution speed, and I thought the Nj was a better method, so I did that. It runs very quickly on the 12c+, not as fast as the 30b, but certainly acceptable. I've also got it running on my original 12c from 12 years ago, and it is ssssllllloooooowwwww on that machine, predictably. But the challenge was to do that algorithm at all on the 12c, and I am happy with the result. Yes, I did appreciate your algebraic solution on the 20s, that was very clever. And I do like the display on that machine, very clear. So thanks Dave for resurrecting this old 67 code. I've also got it running on my 65, which handles the subroutines very well, of course. Don