Even when using atomics in the most naive way it was possible to speed up gridding by a factor of 2.81 on a quadcore machine. This makes gridding faster than disk IO. Since we're doing these two in parallel it means the program's runtime is bound to disk IO.
Even when using atomics in the most naive way it was possible to speed up gridding by a factor of 2.81 on a quadcore machine. This makes gridding faster than disk IO. Since we're doing these two in parallel it means the program's runtime is bound to disk IO.