Make xs be a (global) shared variable:

Each subkernel increments the (global) shared variable value:

Without sharing, each subkernel has its own (local) copy of the variable:

Share a single value:

Access a single element of a shared list efficiently:

Assign a new value to a shared variable:

Only the last assignment performed remains:

Change a single element of a shared list:

Update the value of a shared variable:

Append to a shared list:

Append to an element of a shared list:

Delayed values defined on the master kernel are evaluated on the master kernel:

Delayed values defined on a subkernel are evaluated on the kernel asking for the value:

Use shared variables to synchronize input and output in an infinite search:

Find such that is prime, until the computation is manually aborted:

Use a shared variable to signal all evaluations to stop as soon as a prime has been found:

Use a shared variable to hand out tasks to do:

Monitor the progress of a calculation, counting major calculation steps:

Return results as they are found:

Watch the results appear in real time:

$KernelCount is effectively a shared variable:

Separate read and write operations are not thread-safe:

Use atomic updates for synchronization:

Alternatively, use CriticalSection to make a whole code section atomic:

Create a shared variable whose value is a large matrix:

Accessing its elements causes the whole matrix to be transferred to the subkernels repeatedly:

Use Unevaluated to allow the special code for parts of shared variables to see the variable: