Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
_INTEGER64 calculation
#1
Hi all

Just a quick query.

Why does this code produce these results


billion&& = 10000000000
PRINT billion&&
billionCalc&& = 100000 * 100000
PRINT billionCalc&&


10000000000
1410065408

Is it something to do with the type that would be used for the 2 x 100000 values are not _INTEGER64, so it miscalculates them, or am I just not getting it.
Reply
#2
Overflow in the evaluation phase, is my guess.

Try: billionCalc&& = 10000&& * 10000&&

Define the constant value types. Wink
Reply
#3
Works on my system! feels odd putting suffix on the ends of numbers.
b = b + ...
Reply
#4
(07-25-2022, 10:12 PM)bplus Wrote: Works on my system! feels odd putting suffix on the ends of numbers.

It might feel a little odd at first, but once you think about the compilation process of your code, it makes perfect sense.

Let's take any old equation:  X = A + B + C

*HOW* do we get a value for X, from the above??

Our computer only adds two numbers at a time, so it'd tend to first add A + B, and store that value in a temporary variable.  That variable is then added to C, and repeating computer logic, it's stored in a temp variable. With no other math to do, that temp variable is then finally assigned to X.

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

So, think of X = A + B

Step 1:  Add A + B, store the result in a temp variable.
Step 2: Assign X the value of that temp variable.

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

In an attempt to maximize speed and minimize memory usage, our compiler takes a shot in the dark at how much memory we need to do those calculations.  Generally speaking, most compilers will try and default to using 32-bit math registers for this type of calculation -- which in the case of 100,000 * 100,000 is beyond the limits of what a 32-bit variable can hold without overflowing.

By making one of our constants become an INT64 type, rather than an unsigned type where the compiler tries to guess at what works best, we're telling the compiler that it's going to need to fire up and use the 64-bit math registers and such, rather than the 32-bit ones.

Code: (Select All)
billion&& = 10000000000
Print billion&&
billionCalc&& = 100000 * 100000
Print billionCalc&&

l& = 10000000000
Print l&

As you can see from the simple code above, trying to store the same value in l& as we did in billion&&, overflows giving the exact same answer as our billionCalc&& had.

All you have to do is make certain that at least one  of the variables you're sending into the math registers is 64-bit, and you'll be certain to get back 64-bit results.  Make either of those 100000 become 100000&&, and the issue goes away.

Generally speaking, whenever dealing with anything which might result in greater than 32-bit values, you're going to want to make certain that all your constant values are 64-bit manually.  If you get in the habit of adding those type definition symbols, you'll save yourself overflow issues like this from ever happening with your code.  

It's a good habit to learn to acquire ASAP, IMHO.  Wink
Reply
#5
Thanks for your responses. 

Before I posted my question I did try putting the INT64 suffix and as expected it did display the correct calculation, but as bplus stated, it does feel a bit odd, but obvious thing to do.

Great detailed explanation SMcNeill, thanks for clearing that up.[b][url=https://qb64phoenix.com/forum/member.php?action=profile&uid=6][/url][/b]
Reply
#6
I am wondering
what is real purpose of such a big number evaluations?
jerking off or what?
Reply
#7
what is difference in
i32
i64
if both are integers
what is ?

number 100 as i32 and i64
both are 100
but processor if is 64 size it differently
what i normal...and some users thinking they are krcks...he heh
Reply
#8
(07-26-2022, 03:00 PM)aurel Wrote: I am wondering
what is real purpose of such a big number evaluations?
jerking off or what?

Integr64 can't help you. You need FreeBASIC for that! Big Grin

Some people I guess like to carry out large calculations, and for non-integers, large digit decimal fractions, but frankly, string math routines are the way I approach these situations.

One hurdle in string math is what to do with equations like 1 / 3 * 3. Without a way to hold a remainder, the function returns .99999999 repeatedly to as many digits as the user assigns.

Pete
Reply
#9
(07-27-2022, 04:02 AM)Pete Wrote:
(07-26-2022, 03:00 PM)aurel Wrote: I am wondering
what is real purpose of such a big number evaluations?
jerking off or what?

Integr64 can't help you. You need FreeBASIC for that! Big Grin

Some people I guess like to carry out large calculations, and for non-integers, large digit decimal fractions, but frankly, string math routines are the way I approach these situations.

One hurdle in string math is what to do with equations like 1 / 3 * 3. Without a way to hold a remainder, the function returns .99999999 repeatedly to as many digits as the user assigns.

Pete

This should fix your wagon: https://qb64phoenix.com/forum/showthread...62#pid4162
b = b + ...
Reply




Users browsing this thread: 1 Guest(s)