Char size (in bytes) independent code

The code I am currently working on assumes that a char is 1 byte long, and it depends on this assumption. If I wanted to make sure 8 bits are used could bit fields be a good alternative? Would it slow down my program if I had a bitfield on a char like so:

Code:

---

unsigned char one:8;

---

Even though unsigned char is already 8 bits (on most machines), would it effect my program if it where 8 bits to begin with?

>The code I am currently working on assumes that a char is 1 byte long
A char is 1 byte, always. That's a guarantee made by the C++ standard.

>If I wanted to make sure 8 bits are used
You realize that a byte doesn't have to be eight bits, right? It does have to be at least eight bits though, and if your code assumes an octet then the best solution may be to refuse compilation for systems where CHAR_BIT is not 8:

Code:

---

#include <climits>

#if CHAR_BIT != 8
#error I don't like you
#endif

---

Ya I thought it was only one byte but I read some articles and got confused... Thanks

So would my code slow down the program even on systems where 1 byte is 8 bits?

Quote:

---

>The code I am currently working on assumes that a char is 1 byte long
A char is 1 byte, always. That's a guarantee made by the C++ standard.

---

Now I realize that its not the char itself but the actual byte, I meant to say that my program assumes 1 byte is 8 bits.

>So would my code slow down the program even on systems where 1 byte is 8 bits?
Very likely, but it doesn't matter if the performance hit isn't measurably different when you profile the program. Try it and see.

By the way, what are you doing that requires that each byte be an octet? Is possible that you could avoid that portability issue entirely through careful design.

It's this encryption algorithm I've been working on that moves blocks of bits around in a certain fassion and does some other tricks. When I first came up with the idea I wasn't aware of all the portability issues, but in the end I think it will be great because it can be used for network security across different types of systems because its also endian independent.