lecture 5 approaches to concurrency: the multiprocessor

12
Lecture 5 Approaches to Concurrency: The Multiprocessor

Upload: larue

Post on 06-Jan-2016

43 views

Category:

Documents


0 download

DESCRIPTION

Lecture 5 Approaches to Concurrency: The Multiprocessor. Outline. Some basics Granularity Amdahl’s Law Metrics (speedup, efficiency, redundancy) Multi-core (one thread spans an engine, multiple engines) Tightly coupled vs. Loosely coupled Interconnect Cache Coherency Memory Consistency - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Lecture 5 Approaches to Concurrency: The Multiprocessor

Lecture 5

Approaches to Concurrency: The Multiprocessor

Page 2: Lecture 5 Approaches to Concurrency: The Multiprocessor

Outline

• Some basics– Granularity– Amdahl’s Law– Metrics (speedup, efficiency, redundancy)

• Multi-core (one thread spans an engine, multiple engines)– Tightly coupled vs. Loosely coupled– Interconnect– Cache Coherency– Memory Consistency

• Examples– Early: cm*, HEP, cosmic cube– Recent: cell, power, niagara, larrabee– Tomorrow: ACMP

Page 3: Lecture 5 Approaches to Concurrency: The Multiprocessor

Outline

• Some basics– Granularity– Amdahl’s Law– Metrics (speedup, efficiency, redundancy)

• Multi-core (one thread spans an engine, multiple engines)– Tightly coupled vs. Loosely coupled– Interconnect– Cache Coherency– Memory Consistency

• Examples– Early: cm*, HEP, cosmic cube– Recent: cell, power, niagara, larrabee– Tomorrow: ACMP

Page 4: Lecture 5 Approaches to Concurrency: The Multiprocessor

Granularity of Concurrency

• Intra-instruction (Pipelining)• Parallel instructions (SIMD, VLIW)• Tight-coupled MP• Loosely-coupled MP

Page 5: Lecture 5 Approaches to Concurrency: The Multiprocessor
Page 6: Lecture 5 Approaches to Concurrency: The Multiprocessor

Metrics

• Speed-up:

• Efficiency:

• Utilization:

• Redundancy:

Page 7: Lecture 5 Approaches to Concurrency: The Multiprocessor

Outline

• Some basics– Granularity– Amdahl’s Law– Metrics (speedup, efficiency, redundancy)

• Multi-core (one thread spans an engine, multiple engines)– Tightly coupled vs. Loosely coupled– Interconnect– Cache Coherency– Memory Consistency

• Examples– Early: cm*, HEP, cosmic cube– Recent: cell, power, niagara, larrabee– Tomorrow: ACMP

Page 8: Lecture 5 Approaches to Concurrency: The Multiprocessor

Tightly-coupled vs Loosely-coupled

• Tightly coupled (i.e., Multiprocessor)– Shared memory– Each processor capable of doing work on its own– Easier for the software– Hardware has to worry about cache coherency,

memory contention

• Loosely-coupled (i.e., Multicomputer Network)– Message passing– Easier for the hardware– Programmer’s job is tougher

Page 9: Lecture 5 Approaches to Concurrency: The Multiprocessor

• Interconnection networks– Cost– Latency– Contention

• Cache Cohererency– Snoopy– Directory

• Memory Consistency– Sequential Consistency and Mutual Exclusion

Page 10: Lecture 5 Approaches to Concurrency: The Multiprocessor

Outline

• Some basics– Granularity– Amdahl’s Law– Metrics (speedup, efficiency, redundancy)

• Multi-core (one thread spans an engine, multiple engines)– Tightly coupled vs. Loosely coupled– Interconnect– Cache Coherency– Memory Consistency

• Examples– Early: cm*, HEP, cosmic cube– Recent: cell, power, niagara, larrabee– Tomorrow: ACMP

Page 11: Lecture 5 Approaches to Concurrency: The Multiprocessor
Page 12: Lecture 5 Approaches to Concurrency: The Multiprocessor