civilmh :
Thanks, is it related to the program that I want to run on a specified OS, as well or what? could you please tell me which Microsoft windows is better for parallel computing. by the way, is there any OS that uses grid computing?
what's the meaning of parallel processing? I think it means two Cpus starts simultaneously, not after the first cpu reaches to its utmost, after that the second cpu enter to the processing! which of thoughts is true??
It would be very difficult for me to adequately summarize about 50 years worth of R&D here so I'll try and keep it concise.
Parallel computing has only a little in common with parallel execution. Linear programs that can be vectorized by a good compiler will be executed in parallel if possible. Similarly, programs which are compiled for parallism can still run on a processor with only a single logical processor.
Operating systems address "logical processors" which may or may not have certain functionality. In a symmetric/multi-core/multi-processor system the number of logical processors increases (number of logical threads per core * number of cores per processor * number of processors per system). Thus, a quad core processor without symmetric multiprocessing (aka Hyperthreading) such as the Core 2 Quad family has 4 execution cores each with 1 logical thread which provides for a total of 4 logical processors. Similarly, a dual core processor with symmetric multiprocessing such as most of the Intel i3 family has two cores each with 2 logical threads which provides for a total of 4 logical processors.
Similarly, using a multi-socket system further multiplies this. A server with two Core 2 Quad processors will have a total of 8 logical processors.
The operating system is tasked with properly balancing all of its workloads across all of the logical processors, with only minimal regard to their physical location. Naturally it is preferable to schedule workloads on two logical processors which belong to different cores rather than two logical processors which belong to the same core (hyperthreading) because those would be sharing execution resources. The component tasked with this job is called the scheduler
http://en.wikipedia.org/wiki/Scheduling_(computing)
The scheduler is tasked with trying to find the optimal way to distribute workloads and is responsible for ensuring that real time applications receive real time responses.
Facebook
Twitter
Instagram