Rapid color replacement for in-memory BMP

A function that relatively quick replaces one color with another, in a memory-buffered BMP.
I am trying to improve the performance as much as it is possible. The previous implementation of this function was even faster.. but then I ran into some padding complications.

```
union
{
unsigned long ulColor;
unsigned char byteColor[4];
} oldColor;
union
{
unsigned long ulColor;
unsigned char byteColor[4];
} newColor;

typedef unsigned char BYTE;
typedef unsigned short int WORD;
typedef unsigned long int DWORD;
typedef unsigned long long int DDWORD;

DWORD
bitfox_find_xy
(DWORD WIDTH, DWORD xp, DWORD yp)
{
DWORD w = WIDTH;
DWORD channels = 3;
DWORD bpp = 8;
DWORD single = (channels*bpp)/8;
DWORD offset = (2+sizeof(BMP)+sizeof(DIB));
DWORD rowsize = w*single;

DWORD pixAddress;

if (rowsize % 4 != 0) rowsize += 4 - (rowsize % 4); // account for padding
pixAddress = offset + yprowsize + xpsingle; // finding the address
return pixAddress;
}
DDWORD
bitfox_color_replace_data
(BYTE *buff, BYTE old_r, BYTE old_g, BYTE old_b, BYTE new_r, BYTE new_g, BYTE new_b)
{
#define OFFSET_OF_SIZE 0x2
#define OFFSET_OF_PIXELS 0xA
#define OFFSET_OF_WIDTH 0x12
#define OFFSET_OF_HEIGHT 0x16

DWORD buffSize = (DWORD)&buff[OFFSET_OF_SIZE];
DWORD buffPixels = (DWORD)&buff[OFFSET_OF_PIXELS];
DWORD buffWidth = (DWORD)&buff[OFFSET_OF_WIDTH];
DWORD buffHeight = (DWORD)&buff[OFFSET_OF_HEIGHT];

DDWORD pixels_replaced = 0;
DDWORD ln, col;

oldColor.byteColor[0] = old_b; newColor.byteColor[0] = new_b;
oldColor.byteColor[1] = old_g; newColor.byteColor[1] = new_g;
oldColor.byteColor[2] = old_r; newColor.byteColor[2] = new_r;

for(ln = 0; ln < *buffHeight; ln++)
{
for(col = 0; col <*buffWidth; col++)
{
DDWORD offset = bitfox_find_xy(*buffWidth, col, ln);

if(!memcmp(buff + offset, oldColor.byteColor, 3))

Computing rowsize inside bitfox_find_xy (that is, for each and every pixel)seems wasteful: row size never changes during the bitfox_color_replace_data invocation, and shall only be computed once.

StackExchange Code Review Q#80293, answer score: 2