Bin data to sinwave

Is there a easy and computationally efficient way to group data according to a sin wave? So say I have 20,000 data points, that I want to group into bins following a sinwave, i.e. the data gets grouped by smaller amounts (say ~20) data points in the middle but into larger amounts on the edge (~100) where the sinwave is changing less as a function of time.

And I need to do this all very computationally efficient since I want to do it A LOT in real time. My current method I am doing seems overly complex, I feel like this should be easier.

[Jan]

It depends. What kind of data are you talking of? Where do they come from? Are you talking about one period of a sin wave or does it contain more waved lengths? What is "the middle"? What is the edge? Actually I think that the derivative of the sin has a maximum at 0, but this would mean that 0 is not one of your "edges".