HP Forum Archive 14

[ Return to Index | Top of Index ]

Deep Pi - 42S - Some Results Message #1 Posted by Bill (Smithville, NJ) on 10 May 2004, 1:12 p.m. The two HP-42S's are busy runing the deep pi program and printing their results. The following is the first 24 hours: Note: FAST refers to Fast Speed, NORM refers to the standard speed. Also, I was using a HP-41CX to print the date and time every 10 minutes. I lost some time marks when my timer printing program didn't adjust for rolling over to a new day at midnight. So there's several hours where the 42S printed results, but I don't know the time frame. This is noted as NO-TIME in the list. The times given are from the start time to the next ten minute time mark. Thus the actual time will be less that that indicate - up to a maximum of 10 minutes less. Digit Fast Norm 1 10 Min 20 Min 7 20 Min 40 Min 13 40 Min 1 Hr 20 Min 19 1 Hr 0 Min No Time 25 1 Hr 30 Min No Time 31 No Time No Time 37 No Time No Time 43 No Time 7 Hr 20 Min 49 No Time 9 Hr 20 Min 55 No Time 11 Hr 40 Min 61 No Time 14 Hr 20 Min 67 8 Hr 10 Min 17 Hr 20 Min 73 9 Hr 50 Min 20 Hr 40 Min 79 11 Hr 40 Min 24 Hr 30 Min 85 13 Hr 40 Min 28 Hr 00 Min 91 15 Hr 50 Min 97 18 Hr 20 Min 103 21 Hr 0 Min 109 23 Hr 50 Min 115 27 Hr 00 Min 121 30 Hr 20 Min It looks like it'll take a long time to reach the 1000th digit. Wonder if the batteries will hold out. The Fast Mode is a little more than double the speed. At least with this program.

Guys, aren't these times awful? Message #2 Posted by Gene on 10 May 2004, 1:29 p.m., in response to message #1 by Bill (Smithville, NJ) Perhaps I'm confused. (Wouldn't be the first time). These times are awful. There was an HP41 program (that would work fine on the 42S) that found 1000 digits of PI in under 10 hours. What am I missing here? Gene 12345

Re: Guys, aren't these times awful? Message #3 Posted by Bill (Smithville, NJ) on 10 May 2004, 1:54 p.m., in response to message #2 by Gene Hi Gene, Well.... Hugh did say his program wasn't optimized :) I was mainly running it to see what dfference the fast mode made on the 42S in program execution and battery consumption. Do you have copy of the HP-41 program? Did it just use standard 41 functions or did it make use of synthetic fuctions? If standard functions, then the 42S should execute it quicker. Just re-read your posting and you do say it would run on the 42S. Could you post the program and I'll give it try. Thanks.

Re: Guys, aren't these times awful? Message #4 Posted by Gene on 10 May 2004, 2:09 p.m., in response to message #3 by Bill (Smithville, NJ) PPC V8N2P4 (March/April 1980) A graph of PI calculation times is found in the above PPC Journal. The program was written by Ron Knapp. It was published before the graph date, but not sure when. He later revised it. 1000 digits in 8 hours, I think. Since the 42S in fast mode is about 4-8X faster than a vanilla 41c, that should be between 1-2 hours for 1000 digits.

Program is PPC Journal V7N5P9 Message #5 Posted by Gene on 10 May 2004, 2:15 p.m., in response to message #4 by Gene It can be found on "that other site". Gene P.S. I know the other program might could compute digits 100-110 without doing the first 100 digits, but since Knapp's program can give you all 110 of those digits in much less time, it's more efficient.

42S Fast Mode Message #6 Posted by Trent Moseley on 10 May 2004, 3:32 p.m., in response to message #4 by Gene Gene, I have the 42S Owner's Manual right in front of me. I don't see anything in the index about a fast mode. What am I missing? tm

Re: 42S Fast Mode Message #7 Posted by Bill (Smithville, NJ) on 10 May 2004, 3:41 p.m., in response to message #6 by Trent Moseley Hi Trent, Early versions of the HP-42S Rom could be programmed into a "Fast" mode by using the built-in debugger. See the following page: http://www.rskey.org/gene/hpgene/hp42fast.htm

Re: 42S Fast Mode Message #8 Posted by Gene on 10 May 2004, 3:50 p.m., in response to message #7 by Bill (Smithville, NJ) The 42S rom was changed around late 1989 (?) and this no longer worked. It uses battery power but makes games/programs a joy to run. Old 42S calculators are worth more as a result. Gene

Re: 42S Fast Mode Message #9 Posted by Veli-Pekka Nousiainen on 11 May 2004, 4:03 a.m., in response to message #8 by Gene I can confirm taht with Serial 3330S... [VPN] )-`:

That means you have an updated 42S rom Message #10 Posted by Gene Wright on 11 May 2004, 7:52 a.m., in response to message #9 by Veli-Pekka Nousiainen sorry, but a "size error" means you have one that does not work in fast mode. I have never seen/heard of a 42S that was capable of fast mode with a SN after 3100sXXXX. All of them are 1990 vintage or earlier. Happy hunting...it's worth it! Gene

Re: That means you have an updated 42S rom Message #11 Posted by Trent Moseley on 11 May 2004, 2:37 p.m., in response to message #10 by Gene Wright Thanks for information. My 42S must be one of the last to be made, serial 3515S... tm

Re: That means you have an updated 42S rom Message #12 Posted by Massimo (Italy) on 11 May 2004, 7:01 p.m., in response to message #11 by Trent Moseley One of the ones I have around here is an "ID027..." replacement unit. I think those were among the the last to be built. Massimo

Re: Guys, aren't these times awful? Message #13 Posted by Bram on 10 May 2004, 2:12 p.m., in response to message #2 by Gene If I remeber well, the special thing about deep-pi was that it computes a bunch of decimals without computing the ones before them. I can imagine that it takes extra computing time to do so. The program you are referring to may be more "straight forweard" and computes decimals in order. Hence, perhaps no real practical use for deep-pi, but that wasn't the point here.

Re: Deep Pi - 42S - Timing Correction Message #14 Posted by Bill (Smithville, NJ) on 10 May 2004, 2:28 p.m., in response to message #1 by Bill (Smithville, NJ) Note - The times are the total time to that digit INCLUDING the time to calculate ALL the previous digits. So, 8 HR to Digit 67 also includes calculating and printing the previous digits up to 67. I don't think my original post was real clear on this.

Re: Deep Pi - 42S - Some Results Message #15 Posted by hugh steers on 10 May 2004, 10:13 p.m., in response to message #1 by Bill (Smithville, NJ) hi guys, thanks for posting your results. you're right, this is very bad indeed. 24 hours and only just a hundred digits or so. i was hoping we'd get to 1000 in that time. the algorithm is O(n^1.5) so this is bad news for more digits. firstly, its unlikely that this or similar algorithm will beat the 8 hours for 1000 whatever it does. this is because the method is fundamentally slower than a straight fixed size calculation. i have a few ideas on how to speed up the current method, but it will be a *2 or so if im lucky. i think there is a more recent algorithm that has slightly better order, but its more complicated and may, in fact, be slower for these numbers. i'll look into it. in the meantime, im most grateful that you've got these results as it gives an indication into whether this approach stands a change at all of being feasible (or not).

[ Return to Index | Top of Index ]

Go back to the main exhibit hall