Thread: Program running slower after optimization?

Hello,

I programmed a set of functions for bit output. Trying to make my code run even faster, I came across two lines of code that looked like this:

Code:

*put_buffer |= bits >> (count - current_bit);
 count -= current_bit;

I replaced it with:

Code:

 count -= current_bit;
*put_buffer |= bits >> count;

And did the same for two similar pieces of code.

Than I ran my program in Time profiler (I'm coding on a Mac and using Xcode) only to find out that - on average - my program spent about 30% of time more in the function that contained the above changes.

Any ideas why that might be happening? With that three changes, I did remove three subtractions for each function call and the function was called 10,000,000 times in each of my testing runs.

First, was the revised program run as a release, or did you accidentally run it as a debug version?

Would there be cache miss on the second example, for some reason?

What differences does the generated assembly level code show? I'm not able to help much with assembly (to put it mildly), but when you get down to this level of optimization, I believe it becomes essential to consult it.