Bitwise operations

anyone no of any good sites on bitwise operations not directly relating to game programming?

Meaning what exactly?

I don't think the operators themselves care whether it's a game program or an operating system (for example). They do the same thing in either case.

have you tried http://www.cprogramming.com/tutorial...operators.html

Bitwise operations are farily simple. The tutorial may cover everything! The thing that makes it hard for beginners is that cin and cout don't directly support binary, so you can't easily "see" what you are doing.

You can write your own code to display (and input) binary numbers, or you can use <bitset> to display numbers in binary format, and strtoul() to input a c-style string that represents a binary number (or a number in almost any other base).

However, most real-world programs that use bit manipulation use cin and cout with hexadecimal representation. It is very easy to convert between hex and binary. (You can learn to do it in your head.)

Bitwise operations are most-often used when you are programming "close to the hardware" (drivers or firmware), or with assembly-language programming. The state of one particular bit might represent the on/off state of an LED, or the on/off state of a switch, etc.

And, we usually don't care about the numerical value of the binary number.... We are only concerned with the value of a particular bit, or the "bit pattern". We don't care that 1111 (binary) equals 15 (decimal), or F (hex). We just want to know that all of the bits are "set".

I guess it's not the actual operations that confuse me, it's when do you use them.

Bitwise operations are used to talk directly to hardware? I guess I'm just used to dealing with a layer of abstraction,.

So, maybe an easy example of how to 'talk' to the hardware or 'addresses' so I can see how it's done, if at all possible?

Many thanx!

No. They are used (among other situations) when talking to hardware; they are not the means by which this is done. (It is done through the OS's driver interface.)

I'd say the following are the MOST common uses of bitwise operators, although there are certainly others:

1. Storing binary flags inside integer variables. A single int can hold many flags (on many systems, up to 32 of them). You could do this with 32 separate variables, but using a single int is much cheaper, and it allows you to do more complex things to the flags by using bitwise operators.

2. Talking to hardware. Hardware often takes commands in the forms of single bytes or words, each bit of which represents some flag or status to be set/retrieved from the hardware. You need bitwise operators to send and interpret these results.

3. Accessing the individual bits of a 1-bit bitmap image.

Other fun uses include bitboards in AI chess programming, encryption, etc...

Don't forget reading tightly-packed file formats like SWF.

#1 - Do you want to test if a number is odd/even? You can do this using bitwise and to test:

Code:

---

unsigned int num;

...

if( num & 1 )
{
    // It's odd
}
else
{
    // It's even
}

---

#2 - You can use the bitwise shifting operators to efficiently multiply/divide integer values by powers of 2:

Code:

---

num <<= 1;  // Same effect as num *= 2;
num <<= 3;  // Same effect as num *= 8;
num >>= 4;  // Same effect as num /= 16;

---

This shift operation can be done very quickly and is efficient (no need to actually call an expensive multiply or divide operation) although a good compiler can do the same thing when converting your code if the value being multiplied or divided by is a const/literal (and a power of 2).

#3 - You can use bitwise operators to set/clear/check the values in a variable that may be used as a flag (as mentioned by brewbuck in point #1):

Code:

---

const unsigned int ERR_FLAG1 = 1;  // 0000 0000 0000 0001 = 0x1
const unsigned int ERR_FLAG2 = 2;  // 0000 0000 0000 0010 = 0x2
const unsigned int ERR_FLAG3 = 4;  // 0000 0000 0000 0100 = 0x4
const unsigned int ERR_FLAG4 = 8;  // 0000 0000 0000 1000 = 0x8
const unsigned int ERR_FLAG5 = 16;  // 0000 0000 0001 0000 = 0x10
const unsigned int ERR_FLAG6 = 32;  // 0000 0000 0010 0000 = 0x20

... etc, etc, etc.

unsigned int flag = 0;

...

flag |= ERR_FLAG3;  // Sets the ERR_FLAG3 value stored in variable "flag"
flag &= ~ERR_FLAG1;  // Clears the ERR_FLAG1 value stored in variable "flag"

if( flag & ERR_FLAG2 )
{
    // ERR_FLAG2 value stored in variable "flag" was turned ON
}
else
{
    // ERR_FLAG2 value stored in variable "flag" was turned OFF
}

if( flag & ERR_FLAG5 && !(flag & ERR_FLAG1) )
{
    // ERR_FLAG5 was turned ON and ERR_FLAG1 was turned OFF
}
else
{
    // Either ERR_FLAG5 was OFF or ERR_FLAG1 was ON
}

---

The single "flag" variable above can be used to store many different settings which can be individually checked/set/cleared.

But #1 and #2 should really, really be left to the compiler. Especially as #1 doesn't necessarily work for negative numbers: not on 1's complement systems. Do a &#37;2, and let the compiler optimize it if it decides the platform can handle it.

if found this example at http://www.fredosaurus.com/notes-cpp...s/bitops.html:

Code:

---

int age, gender, height, packed_info;
   
. . .  // Assign values

// Pack as AAAAAAA G HHHHHHH using shifts and "or"
packed_info = (age << 8) | (gender << 7) | height;

// Unpack with shifts and masking using "and"
height = packed_info & 0x7F;  // This constant is binary ...01111111
gender = (packed_info >> 7) & 1;
age    = (packed_info >> 8);

---

which seems to explain a little bit.

I don't however understand the line:

Code:

---

height = packed_info & 0x7F;

---

Is this good practice? to purposely mask numbers in one larger number? is this much more efficient than not dong this?

also, is 2's complement the standard? how do you know what binary system is being used?

>also, is 2's complement the standard?
No, but it tends to be the most common.

>how do you know what binary system is being used?
Read the manual for your system?

I always use an unsigned type for array/vector indices, and for these it's safe to use the bitwise forms (hk_mp5kpdw should have pointed out that the "unsigned" part of the declaration is important, and the two's complement/ones' complement/sign-magnitude issue only applies to signed types).

32 bit colour blending:

Code:

---

Uint32 ColourBlend( Uint32 rgb_1, Uint32 rgb_2, Uint8 alpha_1 )
{
        Uint8 alpha_2 = 256 - alpha_1;
       
        Uint32 rb_1 = ((rgb_1 & 0x00FF00FF) * alpha_1) & 0xFF00FF00;
        Uint32 g_1  = ((rgb_1 & 0x0000FF00) * alpha_1) & 0x00FF0000;
        Uint32 rb_2 = ((rgb_2 & 0x00FF00FF) * alpha_2) & 0xFF00FF00;
        Uint32 g_2  = ((rgb_2 & 0x0000FF00) * alpha_2) & 0x00FF0000;
       
        rb_1 += rb_2;               
        g_1  += g_2;
       
        return (rb_1 | g_1) >> 8;
}

---

Quote:

---

So, maybe an easy example of how to 'talk' to the hardware or 'addresses' so I can see how it's done, if at all possible?

---

A couple of things make this difficult on a PC:*

- The current versions of Windows do not allow a user mode program to directly access the hardware. When you use the '&' (address-of) operator, you are getting a virtual address, not necessarily the true physical address in RAM. In order to access a physical address, you have to use the DDK (Driver Development Kit) to write a kernel mode driver. (I've never done this.)

- Most of the hardware information is proprietary to the hardware manufacturer. Your modem might have a particular bit in a particular register that enables the full-duplex mode. Only the modem manufacturer knows the address of the register and the purpose of the data. The modem manufacturer writes the driver so that it interfaces to Windows as specified by Microsoft, so your (user mode) program can use the modem without accessing the hardware or knowing the hardware details.

I once wrote a program to make sounds from the internal PC speaker. I had to access particular hardware addresses, and I think I had to do some bit manipulation to set-up registers. This program works on Windows 98, but it generates an "access error" with Win2K and above.



* It's not nearly as hard if you are working on an embedded system with its dedicated compiler... The Pic Microcontrollers are frequently programmed by hardware hobbyists. In this type of system, there is usually not an operating system getting in the way, and the supplied compiler library will include functions for reading/writing hardware addresses. (You might write the program on a Windows system, but it runs on a non-Windows system.)

well thank you all, i think i have a better understanding of bitwise operations and their use.