This PR improves over #38 by using view to share the memory between linear and grid shaped data.
The performance, as measured in #38 is
22.779962 seconds (8.79 M allocations: 112.412 GiB, 17.40% gc time, 16.43% compilation time)
which is better than both proposed alternatives.
I belive that this approach is the most efficient way to proceed forward, and that a correct implementation of the force can make this significantly better.
This PR improves over #38 by using view to share the memory between linear and grid shaped data. The performance, as measured in #38 is
which is better than both proposed alternatives.
I belive that this approach is the most efficient way to proceed forward, and that a correct implementation of the force can make this significantly better.