Here are the ways to reap the benefits of multicore processors:

• We have to divide the application into multiple tasks that can be run concurrently.

• We should choose at least as many threads as the number of cores, provided the problem is large enough to benefit from those many threads.

• For computationally intensive applications, we should limit the number of threads to the number of cores.

• For IO-intense applications, the time spent blocking influences the number of threads we’d create.

• Estimate the number of threads using the following formula:

Number of threads = Number of Available Cores / (1 - Blocking Coefficient)

where 0 ≤ blocking coefficient < 1.

• We should slice the problem into several parts so there is enough work for cores and they’re utilized well.

• We must avoid shared mutable state and instead use either isolated mutability or shared immutability.

• We should make good use of the modern threading API and thread pools.