<- previous    index    next ->

Lecture 31c, Parallel PDE


Now we attack solving the PDE on multicore and multiprocessor
computer architectures.

The basic concept is to use worker task to do the parallel
computation and a master control task, the main procedure,
to keep the workers synchronized. These examples, in several
languages, use the barrier method of synchronization.

The first example is converting  fem_check44_la.adb  to
a parallel version. We developed  psimeq, parallel simultaneous
equations solver in lecture 3b
Thus we need to choose the number of processors, NP, to pass
to psimeq.

Then we need to parallelize the longest computation, creating
the stiffness matrix, as covered in lecture 32

Care must be taken to insure no worker tasks set any variable that
might by modified by any other worker task or the master task.

Parallelizing Ada, FEM example

fem_check44_pla.adb fem_check44_pla_ada.out The other files needed to compile and execute include: psimeq.ada laphi.ads laphi.adb gaulegf.adb real_arrays.ads real_arrays.adb Ada barrier software

Parallelizing Java, FEM example

The first example is converting fem_check44_la.adb to a parallel version. We developed psimeq, parallel simultaneous equations solver in lecture 3b Thus we need to choose the number of processors, NP, to pass to psimeq. Then we need to parallelize the longest computation, creating the stiffness matrix, as covered in lecture 32 Care must be taken to insure no worker tasks set any variable that might by modified by any other worker task or the master task. fem_check44_pla.java fem_check44_pla_java.out
    <- previous    index    next ->

Other links

Go to top