Portability question: int32s wrapping around?

[Eulex]

I'm writing a program that translates code written in another programming language into C.
I've found out that I can optimize the outputted code in this way:

long a, b, c, d, e;

a = b * c + d * e

Thing is, what if b * c is a positive value that doesn't fit in a long, and d * e is a negative value, that when added with b * c produces a value that fits in a long (ie it produces valid output). On Intel processors b * c would "wrap", and when I then add d * e to it, it would "wrap back", producing the same output as if I had written a = (long) ((float) b * (float) c + (float) d * (float) e). But I don't know how processors of other brands would react to this. Would they produce the same output?

Regards, Alexander Toresson

[Salem]

> Would they produce the same output?
There's not guarantee of anything.

C only defines wraparound for the unsigned types (incrementing an unsigned value with all 1's set gets you back to all 0's set)

Some processors can (and will) throw hardware exceptions if you try what you are proposing.