Parallel Computing for Data Science

4.8. EXAMPLE: TRANSFORMING AN ADJACENCY MATRIX 99

We then call rollmean() on this thread’s chunk, and write the results into

this thread’s section of mas:

mas [ 1 , m y f i r s t : mylast ] <−

rollmean ( x [ m y f i r s t : ( mylast+k −1)] , k )

When all the threads are done executing the above line, we will be ready

to combine the results. But how will we know when they’re done? That’s

where the barrier comes in. We call barr() to make sure everyone is done,

and then designate one thread to then combine the results found by the

threads:

barr ( ) # make s u r e a l l t h r e a d s have w r i t t e n to mas

i f ( myinfo$ i d == 1) {

r s l t s [ 1 , 1 ]