This is the mail archive of the gcc-help@gcc.gnu.org mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Strange conversion to int64

2010/1/27 Tony Bernardin <sarusama@gmail.com>:
>
> I was expecting to get -64 in every case. I don't understand why
> (int64) = -1 (int32) * 64 (uint32)
> is different from
> (int64) = -1 (int32) * 64 (uint64)
>
> Also why is
> (int64) = -1 (int32) * 64 (uint32)
> is different from
> (int32) = -1 (int32) * 64 (uint32)
>

The difference is in whether the multiplicand or the product is being
promoted, and thus whether things are being sign- or zero-extended.

Here's why (and how) you get 4294967232:

    (int32)-1 * (uint32)64

Then promote both multiplicands to unsigned

    (uint32)4294967295 * (uint32)64

Perform the unsigned 32-bit multiplication

    (uint32)4294967232

Promote to 64-bits, using zero-extension since it's unsigned

    (uint64)4294967232

Then coerce it to signed

    (int64)4294967232

HTH,
~ Scott
Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]