Bizarre math bug

[Daniel Corbier]

In the Visual Basic help file, it explains that the base argument
can be negative as long as the exponent argument is an integer
value. Looking at help files from other compilers, I see that
they do it the same way for equivalent Power functions. Although
the PB help file does not have an explanation for the ^ operator,
I assume that it has always used this same standard. Does
someone from PB care to shed some light on how the ^ operator is
implemented internally in PB? The mysterious result that I got
only appears under certain circumstances, so it appears not to
be the intended compiler behavior.

------------------
Daniel Corbier
UCalc Fast Math Parser
http://www.ucalc.com

[Charles Dietz]

Daniel,

I have conducted many tests, and my results show that the exponent
is converted to the nearest integer so that:

(-2)^2.501 = -8
(-2)^2.499 = 4

And my test show that it doesn't matter what kind of data elenment
receives the answer. I believe that the problem has to do with the
automatic register assignments. For example:

#COMPILE EXE
FUNCTION PBMAIN
   DIM a AS EXT, b AS EXT, c AS EXT, d AS EXT
   MSGBOX STR$(-1 + (-2)^3.45)   'wrong ans
   MSGBOX STR$((-2)^3.45 - 1)    'right ans 
END FUNCTION

whereas...

#COMPILE EXE
FUNCTION PBMAIN
   DIM a AS DOUBLE, b AS DOUBLE, c AS DOUBLE, d AS DOUBLE
   MSGBOX STR$(-1 + (-2)^3.45)   'right ans
   MSGBOX STR$((-2)^3.45 - 1)    'right ans 
END FUNCTION

But if #REGISTER NONE is added, or if less than four unused variables
are allocated, the result is always right.

Therefore, it would appear to me that the problem rests totally on
how the registers are automatically assigned by the compiler.



------------------


[This message has been edited by Charles Dietz (edited August 25, 2000).]

[Lance Edmonds]

Charles, you have done some great detective work here. I've asked R&D to review this topic, guys, so thanks for the info! We'll see what "develops"



------------------
Lance
PowerBASIC Support
mailto:[email protected][email protected]</A>

[Lance Edmonds]

R&D report that it is due to an FPU register stack overflow. It is now on the list. Thanks again guys!



------------------
Lance
PowerBASIC Support
mailto:[email protected][email protected]</A>

[Gregery D Engle]

Originally posted by Tim Wisseman:
>>
Is the problem not with raising a negative number to a power?
This is how i would have done it.
MSGBOX STR$ (-1 +((ABS(-1) ^2)/1))
so that the negation is removed before raising to the power of 2.
Answer 0
>>

My tests show that the problem only hits when you raise
a negative number to a power. Using ABS() is a good work
around to the problem. Your code will work fine.

Tim

Its been years since I've done this but a negative power is an
imaginary number if I remember correctly.

sqr(-2) = sqr(2)i

I might be incorrect on the formula but I think that is somewhat
right.

------------------
-Greg

[Marcus Moore]

Originally posted by Lance Edmonds:
Thanks guys... R&D tell me that this problem has been identified and will be sorted out in the next update to the compiler.

Also, remind them to create a patch for those who can't afford an update, and just want what they already have to work. Thank you.


Sincerely,

Marcus

[Tim Wisseman]

>>
Its been years since I've done this but a negative power is an
imaginary number if I remember correctly.

sqr(-2) = sqr(2)i

I might be incorrect on the formula but I think that is somewhat
right.
>>

Yes, the square root of a negative requires complex math
and imaginary numbers. Negative square roots requires much
more then the SQR() function.

However, powers are just good old floating point math.

(-2) ^ 2 = (-2) * (-2) = 4

2 ^ (1) = 2
2 ^ (0) = 1
2 ^ (-1) = 0.5

. . . and we are just dealing with powers when it comes
to this problem. When a program made with PowerBasic raises
a negative number to a power and things are "just right" the
FPU stack gets messed up.

SQR() is not envolved with this.


------------------

[Lance Edmonds]

Originally posted by Marcus Moore:
Also, remind them to create a patch for those who can't afford an update, and just want what they already have to work. Thank you.

Updates are usually free, Upgrades are not. There is expected to be an update to the current version of the compiler, but at this time, no information, pricing, or details are available..

------------------
Lance
PowerBASIC Support
mailto:[email protected][email protected]</A>

[Marcus Moore]

Originally posted by Lance Edmonds:
Updates are usually free, Upgrades are not. There is expected to be an update to the current version of the compiler, but at this time, no information, pricing, or details are available..

If updates are free, why would "pricing" apply? By the way, where are past updates posted? Forgive me, I'm new around here. Thanks.


Sincerely,

Marcus


------------------

[Lance Edmonds]

Did you note my emphasis on the word "usually"? Unless we start giving away products, upgrades and new purchases will continue to involve charges. However, in the past, updates (which are diffent from upgrades) have been issued for free to registered customers.

The SUPPORT section of this web site details the current edition/build of our products, which is determined by the file's date & time (the DAY indicates the release build number, and the time indicates the version number).

If your product is older than the relevant date shown in the SUPPORT page, then you can send a request to mailto:[email protected][email protected]</A> with your serial number and your zip code (US residents), and they will send you the relevant update. If (and the emphasis is on IF) there are any costs involved, Sales will sort it out with you at that time.

I hope this explains things...
http://www.powerbasic.com/support



------------------
Lance
PowerBASIC Support
mailto:[email protected][email protected]</A>