[1.1.2] | 1: Multicore & Parallel Systems Flashcards
What type of processing does Von Neumann Architecture use?
• SISD: Single Instruction, Single Data
• An instruction must be executed (Full FDE cycle) before the next is fetched.
What is parallel processing?
• The simultaneous processing of data to increase speee.
What 4 methods allow for parallel processing?
• Pipelining
• GPUs
• Multicore Processing
• Distributed Computing
Define pipelining.
• Prevention of idling by fetching an instruction as the previous is decoded and the one prior is executed.
Define a Multicore system.
• A single hardware component with 2 or more independent cores.
• MIMD processing: Multiple Instruction, Multiple Data
What form of processing does a GPU use?
• SIMD: Single Instruction, Multiple Data
• It performs one instruction on a large data set.
Compare the instruction & data set usage for CPUs & GPUs.
• Instructions: Complex, Small/Simple
• Data: •Small sets, Large Sets
What is distributed computing?
• Allowing for multiple computers to work together across a network to solve problems.
• Can reach supercomputer levels of performance but not consistently.
List 3 limitations of parallel processing.
• Not infinitely scalable (efficiency gains decrease as more components are introduced)
• Not all problems can be resolved in parallel.
• Software must be written to take advantage of parallel processing.
List the advantages (3) and disadvantages (2) of parallel processing.
Advantages:
• Faster handling of large data sets
• Not limited by bottlenecking
• Can make maximum use of the CPU
Disadvantages:
• Not all data is suitable for parallel processing
• More expensive hardware