Lecture2
Download as PPTX, PDF0 likes698 views
This document discusses different types of parallelism in computing including data parallelism and task parallelism. It describes how vector instructions can perform the same operation on multiple data elements simultaneously, like adding two arrays element by element. Pipelining is described as overlapping the execution stages of instructions to improve performance similar to an assembly line. Limitations of pipelining include being limited by the slowest stage and penalties for mispredicted branches in deep pipelines. Super scalar execution and multiple pipelines can help alleviate these bottlenecks by issuing multiple instructions per cycle.
Lecture2
- 1. CS-416 Parallel and Distributed SystemsJawwadShamsiLecture #2 19th January 2010
- 2. How Much of ParallelismDecomposition: The process of partitioning a computer program into independent pieces that can be run simultaneously (in parallel).Data ParallelismTask Parallelism
- 3. Scalar InstructionSingle Instruction at a timeLoad R1 @ 1000Vector InstructionVector operationsC(1:100)=A(1:100)+B(1:100)Super computerVector Instructions with pipelined floating point arithematic
- 4. PipeliningPipelining overlaps various stages of instruction execution to achieve performance. At a high level of abstraction, an instruction can be executed while the next one is being decoded and the next one is being fetched. This is akin to an assembly line for manufacture of cars.
- 5. Pipeline-limitationsPipelining, however, has several limitations. The speed of a pipeline is eventually limited by the slowest stage. For this reason, conventional processors rely on very deep pipelines (20 stage pipelines in state-of-the-art Pentium processors). However, in typical program traces, every 5-6th instruction is a conditional jump! This requires very accurate branch prediction. The penalty of a misprediction grows with the depth of the pipeline, since a larger number of instructions will have to be flushed.
- 6. Pipelining and Superscalar Execution One simple way of alleviating these bottlenecks is to use multiple pipelines. The question then becomes one of selecting these instructions.
- 7. Super Scalar ExecutionIssue multiple instructions during the same CPU cycleHow?By simultaneously dispatching multiple instructions to redundant unitsnot a separate CPU core but an execution resource within a single CPU e.g. ALU
- 8. Difference Between Vector Instruction and Pipelining
- 9. Parallel Processing Computer speed / output is increased through parallel processingIncrease hardwareIncrease costTech advancementsIncrease in cost is minimal
- 10. Types of Parallelism - explained
- 11. Data ParallelismSame code segment runs concurrently on each processorEach processor is assigned its own part of the data to work on