32-bit fixed point overflow

Question

I'm doing some 'early computing' on a 32-bit Windows PC, and looking at the limits.

Now, 2**32 is 4,294,967,296, and I find that

4294967290 + 5

is perfectly OK, and

    4294967290 + 6

quite properly overflows.

What puzzles me is that

   429496729 * 10

overflows, although the product, 4294967290, is in range.

Anyone interested?

what environment are you using? Your compiler / interpreter may be doing some inexact bounds checking. — ThomasMcLeod, Mar 02 '11 at 14:00
That's not fixed point by the way, it's integer. Fixed point is integer with scaling (for example, 100 = 1.00, 4323 = 43.23, and so on). — paxdiablo, Mar 02 '11 at 14:05

score 4 · Accepted Answer · answered Mar 02 '11 at 14:02

4

In the absense of any code, I'd guess 429496729 gets implicitly typed as signed integer, for which 4294967290 is too much.

answered Mar 02 '11 at 14:02

GSerg

You beat me to it. The result is probably negative to prove it. – phkahler Mar 02 '11 at 14:19
Thanks fellows, quite right. Actually there was no code, but the Visual C 6 debugger watch window (used for my 'probing') evaluates numeric integers as signed. Quite reasonably actually. All is well with unsigned int. – John White Mar 04 '11 at 05:49

1 Answers1