Friday, June 13, 2008

Going Parallel With Parallel Extension For .Net 3.5

Now a days, almost all computers being used have multiple cores, for an example technologies like dual core.But, although you have dual core, generally you write programs that run on a single core, this does not utilize the total power of the multiple CPU or core that you have.
This is exactly the power what Microsoft Parallel Extensions for .NET Framework 3.5 (a.k.a PFX).gives you.
With PFX, you can write flexible programs that would take advantage of utilizing multiple cores in the environment it is deployed in, giving you high performance and scalabilty.
PFX comes in 2 "Flavours", Task Parallel Library (TPL) and PLINQ.
TPL gives you general APIs that can be used to make your program run in parallel, for an example , take the for loop below
for(int i = 0; i < 1000000; i++ )
{ //Do something that is not dependent on other iterations }

the for loop does 1000000 iterations, and does something that is not dependent on other iterations. Running this program in a mulit core machine , will only utilize one CPU and not at it's best and eventually the response time is also high.
This for loop can be easily made to run in parallel, as the iterations are not dependent on each other.
With PFX, you can re-write your program like this

Parallel.For(0, 1000000, i => (/*Do somthing */));

This program would run faster as PFX assigns chunks of iterations to the mulitple CPUs. You would also see that the almost 100% of CPUs power is used if you are on a dual core.
Parallel is a class in the PFX that gives you high level static APIs for looping and invoking in parallel, you can have a look at the docs if you want more... :)
PLINQ on the other hand, is a way to make your queries run in parallel, this applies to in-memory objects and xml. Queries that hit databases is not a candidate for parallisam and hench cannot be used with PLINQ.You can find some introductory samples and videos at the MSDN Parallel Computing section.

No comments:

Post a Comment