A parallel computer is about how things are connected together the network that connects the parallel computers together and they can be connected in a number of different ways we’re going to take separate processrors and connect them together to make a parallel computer
It’s all about the need for speed It’s about trying to solve a problem faster by using more than one processor that is by using parallelism and parallel computers.
Parallel Computing: Using multiple processors in parallel to solve problems more quickly than with a single processor
;
- Shared Memory(SMP) or Multicore
A shared memory multiprocessor (SMP*) by connecting multiple processors to a single memory system.
A multiprocessor contains multiple processors(cores) on a single chip.
Technically, SMP stands for “Symmetric Multi-Processor”
- High Performance Computing(HPC) or distributed memory
A distrubuted memory multiprocessor has processors with their own memories connected by a high speed network
Also called a cluster
A High Performance Computing(HPC) system contains 100s or 1000s of such processors(nodes)
- Single Instruction Multiple Data(SIMD)
A single Processor multiple data(SIMD) computer has multiple processors(or functional units) that perform the same operation on multiple data elements at once.
Most single processors have SIMD units with ~2-8 way parallelism
Graphics processing units(GPUs) use this as well.
Concurrency vs. Parallelism Concurrency: A condition of a system in which multiple tasks are logically active at one time. Parallelism: A condition of a system in which multiple tasks are actually active at one time.
Concurrent Applications: An application for which computations logically execute simultaneously due to the semantics of the application. The problem is fundamentally concurrent.
Parallel Application: An application for which the computations actually execute simutanenously in oder to compute a problem in less time. The problem doesn’t inherently require concurrency…you can state it sequenttially.
