1. RE: bigmath
- Posted by Matthew Lewis <matthewwalkerlewis at YAHOO.COM> Nov 11, 2002
- 375 views
> From: Kat [mailto:kat at kogeijin.com] > Oh. I didn't get any feedback from you on it, can you send me > what caused > errors, please, so i can track them down? That was plainly labeled > prerelease, i hope no one was relying on it. Sorry for not getting back to you, I was a little too focused (and frustrated) on getting the guts of my lib to work. Here's a division example. Actually, it errors out on the line where you try to strip out the decimal before returning (BTW, 31600001 and 31600039 are both prime, and 998561264000039 is their product--you can actually test this one with atoms) : 998561264000039 / 31600001 = bigmath: 31600039 Kat: 3 998561264000039 / 31600039 = bigmath: 31600001 Kat: 3 Subtraction (I haven't looked too closely at this, but the digits are mostly right--there seems to be some periodicity to the discrepancies): 992843955998047072709347710023082885682999253386849704500005269311072880491 - 998561264000039 = bigmath:99284395599804707270934771002308288568299925338684970450000427074980 8880452 Kat: 991843855988046072609337709022982875681999153376848704499994270749808880452 Off X X X X X X XX XX X X X X X XXXXX I didn't test against non-integers (most of the stuff I did uses integers). Matt Lewis
2. RE: bigmath
- Posted by Kat <kat at kogeijin.com> Nov 11, 2002
- 382 views
On 11 Nov 2002, at 15:30, Matthew Lewis wrote: > > > > From: Kat [mailto:kat at kogeijin.com] > > > Oh. I didn't get any feedback from you on it, can you send me > > what caused > > errors, please, so i can track them down? That was plainly labeled > > prerelease, i hope no one was relying on it. > > Sorry for not getting back to you, I was a little too focused (and > frustrated) on getting the guts of my lib to work. > > Here's a division example. Actually, it errors out on the line where you > try to strip out the decimal before returning (BTW, 31600001 and 31600039 > are both prime, and 998561264000039 is their product--you can actually test > this > one with atoms) : > > 998561264000039 / 31600001 = > bigmath: 31600039 > Kat: 3 I see some of the problem, i got this from value(050561234000039) 0,5.0561234e+013 which naturally makes no sense. Needs more error proofing, oh well. Kat
3. RE: bigmath
- Posted by Matthew Lewis <matthewwalkerlewis at YAHOO.COM> Nov 18, 2002
- 371 views
> From: aku saya [mailto:akusaya at gmx.net] > Maybe this is a bug but I am unsure... More of a feature :) OK, so you start out with 56 digits: 1 2 3 4 5 12345678901234567890123456789012345678901234567890123456 > d1 = "12345678901234567895555555555555555555555555555555555555" > d2 = "23" I assume you left it at base 10, meaning that you have 75 total digits of precision. Whenever multiplying or dividing, I recommend using at least twice as many digits as you think you'll need--in this case, at least 112, or base 15. I tried it with 20 and everything worked. I think I may switch division/mod operations over to string math to simplify this sort of thing. It may slow some things down, but will make getting accurate results easier. Matt Lewis