Question

I started working with C# a few weeks ago and I’m now in a situation where I need to build up a “bit set” flag to handle different cases in an algorithm. I have thus two options:

    enum RelativePositioning
    {
        LEFT = 0,
        RIGHT = 1,
        BOTTOM  = 2,
        TOP = 3,
        FRONT = 4,
        BACK = 5
    }

    pos = ((eye.X < minCorner.X ? 1 : 0) << (int) RelativePositioning.LEFT)
        + ((eye.X > maxCorner.X ? 1 : 0) << (int) RelativePositioning.RIGHT)
        + ((eye.Y < minCorner.Y ? 1 : 0) << (int) RelativePositioning.BOTTOM)
        + ((eye.Y > maxCorner.Y ? 1 : 0) << (int) RelativePositioning.TOP)
        + ((eye.Z < minCorner.Z ? 1 : 0) << (int) RelativePositioning.FRONT)
        + ((eye.Z > maxCorner.Z ? 1 : 0) << (int) RelativePositioning.BACK);

Or:

    enum RelativePositioning
    {
        LEFT = 1,
        RIGHT = 2,
        BOTTOM  = 4,
        TOP = 8,
        FRONT = 16,
        BACK = 32
    }

    if (eye.X < minCorner.X) { pos += (int) RelativePositioning.LEFT;   }
    if (eye.X > maxCorner.X) { pos += (int) RelativePositioning.RIGHT;  }
    if (eye.Y < minCorner.Y) { pos += (int) RelativePositioning.BOTTOM; }
    if (eye.Y > maxCorner.Y) { pos += (int) RelativePositioning.TOP;    }
    if (eye.Z > maxCorner.Z) { pos += (int) RelativePositioning.FRONT;  }
    if (eye.Z < minCorner.Z) { pos += (int) RelativePositioning.BACK;   }

I could have used something as ((eye.X > maxCorner.X) << 1) but C# does not allow implicit casting from bool to int and the ternary operator was similar enough. My question now is: is there any performance improvement in using the first version over the second?

Thank you
Tommaso

Answer 1

The inline if operator (?, :) will generate nearly the same IL as the standard if list in the second example. The only difference you will see here are the particular operations the processor will be doing, and I can bet that ADD is quicker than SHL.
Since you’re going to be adding the results anyway, I would opt for the second example (plus it makes it far easier to read).

EDIT
I just checked the IL of both examples, and it goes against what I have said above.
The first example generates far less IL (34 lines less) so you’ll have to run a performance test to really determine if it is faster too.