Faster graphics performance with Java using the Join/Fork framework?

Some time ago, I had stopped with my experiments with a multi-threaded pure Java Graphics2D rendering pipeline, because I couldn’t find a way to efficiently fork and join fine grained worker threads.

Today I stumbled over the paper A Java Fork/Join Framework by Doug Lea, which looks very promising.

I am not sure yet, whether this is going to solve all of my problems.

My rendering pipeline is currently rendering into an int[] array. Where each pixel is located at array[x+y*width]. I don’t know, whether having such an int[] array leads to I/O contention between the CPU cores. Maybe I should switch to an array of arrays, a int[][] array, where each pixel is located at array[x][y]. This is the same design that Antigrain uses.