Question

I’ve got two streams of pixels and basically need to do a custom xor on them for a final resulting picture. It’s working great – only problem is is that it’s taking the simulator about 4 seconds to parse through the code. I know there’s got to be a way to optimize this routine – but after a few days of testing out my ideas (to no avail) I’m resorting to getting some help!

Here’s my code – thanks in advance for any suggestions!

             //rawPic1Data and rawPic2Data is a stream of unsigned chars that ultimately came from a UIImage

                for (int i = 0 ; i < (bufferLength); i=i+4)
                {

                    sred = (int)(rawPic1Data[i + 0]);
                    sgreen = (int)(rawPic1Data[i + 1]);
                    sblue = (int)(rawPic1Data[i + 2]);

                    rred = (int)(rawPic2Data[i + 0]);
                    rgreen = (int)(rawPic2Data[i + 1]);
                    rblue = (int)(rawPic2Data[i + 2]);

                    fred = 0;
                    fgreen = 0;
                    fblue = 0;
                    falpha = 0;

                    if (((sred == 102) && (sgreen == 0) && (sblue == 153)) || ((rred == 102) && (rgreen == 0) && (rblue == 153)))
                    {
                        fred = 102; fgreen = 0; fblue = 153; falpha = 255;
                    }
                    else if (((sred == 153) && (sgreen == 51) && (sblue == 204)) || ((rred == 153) && (rgreen == 51) && (rblue == 204)))
                    {
                        fred = 153; fgreen = 51; fblue = 204; falpha = 255;
                    }

                     //...repeat the elseifs for another 12 colors.  (14 total)
                }

            //Use the f values for my final output... 

Answer 1

One thing you can do is to combine the three values from each stream into one larger variable. This will allow you to perform a comparison on all three at once, so there will be one-third as many comparisons in your code.

sred = (int)(rawPic1Data[i + 0]);
sgreen = (int)(rawPic1Data[i + 1]);
sblue = (int)(rawPic1Data[i + 2]);
register unsigned sval = (sred << 16) | (sgreen << 8) | (sblue);

rred = (int)(rawPic2Data[i + 0]);
rgreen = (int)(rawPic2Data[i + 1]);
rblue = (int)(rawPic2Data[i + 2]);
register unsigned rval = (rred << 16) | (rgreen << 8) | (rblue);

fred = 0;
fgreen = 0;
fblue = 0;
falpha = 0;

if(sval == ((102<<16)|(153) || rval == ((102<<16)|(153)) {
    fred = 102; fgreen = 0; fblue = 153; falpha = 255;
} else if(sval == ((153<<16)|(51<<8)|(204))) || rval = ((153<<16)|(51<<8)|(204)))) {
    fred = 153; fgreen = 51; fblue = 204; falpha = 255;
} ...

An optimization like this sped up my test program by 36.5%, with only 5 test cases.