Scribd
Upload
55 views

Advance Computer Architecture in 5 Pages

The document summarizes different classifications of parallel computer architectures. It describes SIMD and MIMD models, and discusses distributed memory and shared memory MIMD computers. Within shared memory systems it distinguishes between uniform memory access (UMA), non-uniform memory access (NUMA), cache-coherent NUMA (ccNUMA), non-cache-coherent NUMA (nccNUMA), and cache-only memory (COMA) models. It also notes the convergence of message passing and shared memory programming models in modern parallel systems.

Uploaded by

morvarid7980
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
55 views

Advance Computer Architecture in 5 Pages

The document summarizes different classifications of parallel computer architectures. It describes SIMD and MIMD models, and discusses distributed memory and shared memory MIMD computers. Within shared memory systems it distinguishes between uniform memory access (UMA), non-uniform memory access (NUMA), cache-coherent NUMA (ccNUMA), non-cache-coherent NUMA (nccNUMA), and cache-only memory (COMA) models. It also notes the convergence of message passing and shared memory programming models in modern parallel systems.

Uploaded by

morvarid7980
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

Advanced Computer Architecture Chapter 3

Dr.-Ing. Achim Basermann C & C Research Laboratories, NEC Europe Ltd. Rathausallee 10, 53757 Sankt Augustin [email protected] http://www.ccrl-nece.de/~basermann/

Contents
1. Introduction, von Neumann Bottleneck 2. Parallelism within a Processor 3. Classification of Parallel Architectures 4. Performance of Parallel Computers 5. Interconnection Networks 6. Cache Coherence and Cluster Interconnects 7. Vector Computers and Clusters of Vector SMPs 8. Distributed Memory Computers 9. Grid computing

Chap. 3

Classification
Parallel systems Parallel computers SIMD (Single Instruction Multiple Data): Synchronized execution of the same instruction on a set of ALU MIMD (Multiple Instruction Multiple Data): Asynchronous execution of different instructions Computer networks Networks of workstations - NOW: No dedicated highperformance network Clusters: Collection of workstations dedicated to parallel computing with a dedicated high-performance network Grids: Collection of resources, including parallel computers, building a virtual high-performance system M. Flynn, Very High-Speed Computing Systems, Proceedings of the IEEE, 54, 1966

Chap. 3

Chap. 3

MIMD computers
MIMD computers Distributed memory (multicomputers): Building blocks are nodes with private physical address space. Communication is based on messages. Shared Memory - SM (multiprocessor): System provides a shared address space. Communication is based on read/write operation to global addresses. Symmetric multiprocessors - SMP (Uniform Memory Access UMA): Centralized shared memory, accesses to global memory from all processors have same latency. Non-uniform Memory Access Systems - NUMA (Distributed Shared Memory Systems - DSM): Memory is distributed among the nodes, local accesses much faster than remote accesses.

Chap. 3

NUMA Systems
NUMA systems Cache-coherent NUMA - ccNUMA: Home location of data is fixed. Copies of shared data in the processor caches are automatically kept coherent, i.e., new values are automatically propagated. Non-cache-coherent NUMA - nccNUMA: Home location of data is fixed. Copies of shared data are independent of original location. Cache-only memory - COMA: Data migrate between memories of the nodes, i.e., home location is changed.

Chap. 3

Classification
Parallel Systems

SIMD

MIMD

Networks

Distributed Memory

Shared Memory

NOW

Grid

Cluster

UMA

NUMA

ccNUMA

nccNUMA

COMA

Chap. 3

Convergence of MP and SM Computers


Level of the programming model: Message passing (MP) available also on shared memory systems Remote memory access enables shared addresses on message passing systems Extension to the message passing abstraction Implementation based on software or hardware Shared virtual address space can be established on a distributed memory system at the page level

Chap. 3

Convergence of DM and SM Computers


Level of machine organization: Block diagrams look very similar. Shared memory Network interface was integrated with cache or memory controller. Distributed Memory (DM) Network interface is I/O device. Integration into the memory systems enabling data transfers from/into user space DMA transfers across the network

Generic Scalable Multiprocessor

Chap. 3

Mem

CA

Mem

CA

Scalable interconnection network

You might also like