Cast integral array to class array

Question

I'm using a library passes me a pointer to an array of int32_T. I need to do some math on these numbers, but the data is actually in Q23.8 format.

I created a class overloading the basic math operators, but I'm stuck with how to cast the int32_t array pointer to my new class -- the datatypes are different sizes. Is there a way to do force my new class to match the size of int32_t, or is there another way to get this done?

Note: I want to avoid copying each item of this large array, as I'm developing for a resource constrained system.

In your class have a constructor that takes a int32_t. You could then cast a single element of the array to your class in order to operate on it. — Aumnayan
– Aumnayan, Commented Jun 3, 2015 at 20:59

PlushBeaver · Accepted Answer · 2015-06-03 21:59:52Z

First of all, please note that if you are not constrained by performance, only by memory consumption, the solution suggested by @Aumnayan may be preferable for simplicity and portability.

In order to force your objects to match the size int32_t and alignment of Q23.8 you can use bit-fields:

struct Q_23_8
{
    int32_t fractional :  8;
    int32_t integral   : 23;
};

static_assert(sizeof(Q_23_8) == sizeof(int32_t), "Sizes differ!");

Please note that the order of fields depends on target CPU endianness; the sample is for little-endian. Also, to ensure the size of your class you may need to use your compiler's packing and alignment control capabilities (e g. #pragma pack).

Having int32_t encoded[] you can interpret it as an array of Q_23_8 as follows:

Q_23_8* decoded = reinterpret_cast<Q_23_8*>(encoded);

Here is a complete code sample: http://ideone.com/X177fQ. I strongly suggest reading more about bit-fields, their portability, and limitations.

Collectives™ on Stack Overflow

Cast integral array to class array

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related