By sorting the data into classes of offset, we can restrict
the size of the spiral to the maximum offset of each class.
The spiral data movement allows any azimuth distribution
of the input traces. However, because a trace contributes to the stack
only along its original source-geophone line, some processors
are left idle during the spiral trip of the trace. Figure
shows the working processors in grey and the idle processors in white.
In this example, the computer load, defined as the ratio of the number
of working processors to the total number of processors, is thirty percent.
For larger offsets, the load balance is expected to be worse because the
number of working processors increases in proportion to the offset length
whereas the total number of processors increases in proportion to
the square of the offset.
The unoptimized use of processing power makes the algorithm
inefficient for 3-D land data.