FWIW: I get the same results on I686 Windows, as you get on I686 linux, which suggests to me that the difference if down to differences in the floating point hardware rather either the way Perl is built, or the underlying CRT math functions.

I thought for a while that it might be down to the IEEE rounding mode is use, but I tried it with all 4 modes and whilst the results do vary, the differences are far less than you are seeing: 

C:\test>junk54 MaunaLoa 636 1
0.011256        -43.244877      41833.442623
Roots: 2093.870 or 1747.905

C:\test>junk54 MaunaLoa 636 2
0.011256        -43.245005      41833.967213
Roots: 2093.804 or 1747.982

C:\test>junk54 MaunaLoa 636 3
0.011256        -43.245005      41833.442623
Roots: 2093.939 or 1747.847

C:\test>junk54 MaunaLoa 636 4
0.011256        -43.245005      41833.967213
Roots: 2093.804 or 1747.982

[download]

If I add use bignum; I get the same result regardless of rounding mode (as you'd expect with infinite precision): 

C:\test>junk54 MaunaLoa 636 0
0.011250        -43.220000      41809.395000
Roots: 2093.969 or 1747.809

[download]

But the result is still not far from the standard double precision results, and far away from your results on those other platforms.

So then I decided to "check the math".

I fed your sample data into wolfram/alpha's quadratic fit and got a = 0.01125 b=-43.33 c=41809.39500009608 (your results are close enough!)

I then walked manually through the calculation: 

a = 0.01125 
b=-43.33 
c=41809.39500009608 - 636 = 41173.39500009608

b^2 = 1867.9684

4ac = 4 * 0.01125 * 41173.39500009608 = 1852.8027750043236 
[1852.8027750043236]

2a =  0.0225

sqrt( b^2 - 4ac) = sqrt( 1867.9684 - 1852.8027750043236 ) = 3.89430674
+13438813500081381515803
[3.8943067413438813500081381515802609939934472412114543853497515883707
+998992565436459214366842601252101413353364558519950923345771453315158
+871886529115406835628748856351213011743063789464283329454353106415694
+831827916183018713201714487448856249092857883643766930086114150302755
+812386977175989440678382944531781851160460991847426853996720283630521
+154852957729948705070366158701306031797881524733027828871220198125664
+64437007666673563847997186038400782988968755635538]

x' = -b + sqrt(b^2 - 4ac) = 43.22 + 3.8943067413438813500081381515803 
+= 47.11430674134388135000813815158
[47.114306741343881350008138151580260993993447241211454385349751588370
+799899256543645921436684260125210141335336455851995092334577145331515
+887188652911540683562874885635121301174306378946428332945435310641569
+4831827916183018713201714...]

y' = -b - sqrt(b^2 - 4ac) = 43.22 - 3.8943067413438813500081381515803 
+= 39.32569325865611864999186184842
[-39.32569325865611864999186184841973900600655275878854561465024841162
+920010074345635407856331573987478985866466354414800490766542285466848
+411281134708845931643712511436487869882569362105357166705456468935843
+0516817208381698128679..]

x = x' / 2a = 47.11430674134388135000813815158 / 0.0225 = 2093.9691885
+041725044448061400702
[2093.9691885041725044448061400702338219552643218316201949044334039275
+911066336241620409527415226722315618371260647045331152148700953480673
+727639401294018081583499949171165022744136168420634814642415693618475
+3258590184970230539200761...]

y = y' / 2a = 39.32569325865611864999186184842 / 0.0225 = 1747.8085892
+736052733329716377076
[1747.8085892736052733329716377075439558225134559461575828733443738501
+866711441536157368250362551055462159406517130732446625629076824297104
+050138376483759696194277828606612755033641609357142963135362084159302
+451918759280754723857...]

[download]

The small results are done using calc.exe, the extended numbers in brackets are produced using wolfram alpha.

All of which suggests that your I686 results are correct, and the "bug" is on the other platforms.

Perhaps they're using single precision? Or at least, perhaps lookup tables for sqrt() &| pow() that are only single precision?

Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error.
"Science is about questioning the status quo. Questioning authority".
In the absence of evidence, opinion is indistinguishable from prejudice.
RIP an inspiration; A true Folk's Guy

In reply to Re: Calculation discrepancy between Perl versions by BrowserUk
in thread Calculation discrepancy between Perl versions by oko1

Title:
Use:  <p> text here (a paragraph) </p>
and:  <code> code here </code>
to format your post, it's "PerlMonks-approved HTML":


  • Posts are HTML formatted. Put <p> </p> tags around your paragraphs. Put <code> </code> tags around your code and data!
  • Titles consisting of a single word are discouraged, and in most cases are disallowed outright.
  • Read Where should I post X? if you're not absolutely sure you're posting in the right place.
  • Please read these before you post! —
  • Posts may use any of the Perl Monks Approved HTML tags:
    a, abbr, b, big, blockquote, br, caption, center, col, colgroup, dd, del, details, div, dl, dt, em, font, h1, h2, h3, h4, h5, h6, hr, i, ins, li, ol, p, pre, readmore, small, span, spoiler, strike, strong, sub, summary, sup, table, tbody, td, tfoot, th, thead, tr, tt, u, ul, wbr
  • You may need to use entities for some characters, as follows. (Exception: Within code tags, you can put the characters literally.)
            For:     Use:
    & &amp;
    < &lt;
    > &gt;
    [ &#91;
    ] &#93;
  • Link using PerlMonks shortcuts! What shortcuts can I use for linking?
  • See Writeup Formatting Tips and other pages linked from there for more info.