2011-12-16, 04:10 AM
I'm aware that this topic deserves a whole book's worth of discussion, but my implementation should be relatively simple, so I'm wondering if I could dive right into it without understanding the underpinnings of multicore programming.
Objective: Having a large amount of data in row-column/table/matrix format, I want to split it up into equal P parts (P = # of cores) and send them to each respective core to perform a summation operation (or equivalently, the inner product of vectors).
The operation should be as simple as
![[Image: %25255BUNSET%25255D.gif]](https://lh4.googleusercontent.com/-2EvHayGhf0c/Tur6guInDcI/AAAAAAAAAN8/HWjkHaYC5JY/s192/%25255BUNSET%25255D.gif)
where
x_i is a vector of length n
mu is a vector containing all the arithmetic means of all the x_i's
Question: How exactly do I "send" each operation to a specific core and have it carried out? How do I handle the returned values?
Objective: Having a large amount of data in row-column/table/matrix format, I want to split it up into equal P parts (P = # of cores) and send them to each respective core to perform a summation operation (or equivalently, the inner product of vectors).
The operation should be as simple as
where
x_i is a vector of length n
mu is a vector containing all the arithmetic means of all the x_i's
Question: How exactly do I "send" each operation to a specific core and have it carried out? How do I handle the returned values?